The realm of healthcare is undergoing a revolutionary transformation through the integration of artificial intelligence (AI) for early cancer detection, ushering in a new era of enhanced well-being. This review paper delves into the paradigm shift brought about by AI, highlighting its potential to identify diseases in their nascent stages, thereby revolutionizing preventive healthcare. Harnessing the computational prowess of AI, this paradigm empowers us to analyze intricate medical data, including internal imagery and health records, facilitating the identification of subtle disease markers imperceptible to the human eye. The key focus lies not only in early detection but also in ensuring diagnostic accuracy. A comprehensive analysis of various studies underscores AI’s superiority in comparison to traditional methods, manifesting in quicker and more precise identification of anomalies. This transformation translates to expedited medical interventions and improved patient outcomes. The crux of this evolution lies in AI’s capacity to redefine healthcare, transforming it into a proactive endeavor that identifies and addresses health concerns while they remain manageable.

Healthcare has transformed remarkably due to technology, changing how we take care of our health. One big focus is spotting diseases early, which means finding them when they are still small. This is super important and can help not just individual people, but also the whole healthcare system.

Finding diseases early is like a guiding light for better patient outcomes. It can change how diseases go, stopping them or even preventing them. When we catch diseases early, the treatments are often easier, safer, and work better. Plus, it helps healthcare systems by not needing complicated and costly procedures when diseases get worse. Saving money is a big deal too. Healthcare costs a lot, and finding diseases late costs even more. If we catch diseases early, it is better for people’s wallets and helps healthcare resources. It shifts healthcare from just fixing problems to stopping them before they start.

Now, say hello to artificial intelligence (AI)—a powerful tool that is changing how healthcare works. AI is like smart computers that learn and solve problems like humans. In healthcare, AI is becoming a game-changer. It helps with diagnoses, suggesting treatments, and more. AI is like a super-fast brain. It can look at tons of medical information—from records to complicated images—way faster than a person. This helps doctors make decisions quickly and accurately.

In this comprehensive review paper, we delve into the dynamic synergy between early cancer detection and the transformative capabilities of Artificial Intelligence (AI). Our exploration revolves around the remarkable collaboration of AI and medical science, with a specific focus on their potential to revolutionize healthcare by identifying diseases at their nascent stages.

We embark on a journey through an array of cutting-edge research studies, each serving as a lens into the intersection of AI and early cancer detection. Our primary objective is to dissect the intricate web of findings to unveil the profound impact AI could wield in the realm of disease prevention. Breast, lung, and skin cancers are more prevalent compared to other cancers due to a combination of factors, including lifestyle, genetic predisposition, environmental influences, and screening practices.

  • Lifestyle factors: Factors such as hormonal changes, obesity, alcohol consumption, and lack of physical activity can contribute to the development of breast cancer.

  • Genetic factors: Certain genetic mutations, such as BRCA1 and BRCA2, increase the risk of breast cancer.

  • Hormonal influences: Hormones, such as estrogen and progesterone, can impact breast tissue and contribute to cancer development.

  • Screening Awareness: Regular mammography screenings and increased awareness have led to earlier detection, resulting in higher reported cases.

  • Smoking: The primary risk factor for lung cancer is tobacco smoking, which significantly increases the likelihood of developing the disease.

  • Secondhand smoke: Exposure to secondhand smoke also contributes to lung cancer risk.

  • Environmental exposures: Occupational exposures to asbestos, radon, and other carcinogens play a role in lung cancer development.

  • Global prevalence: Lung cancer is common worldwide due to the widespread prevalence of smoking and environmental factors.

  • Sun exposure: Excessive exposure to ultraviolet (UV) radiation from the sun or tanning beds is a major cause of skin cancer.

  • Lifestyle choices: Frequent sunburns, outdoor occupations, and recreational activities increase the risk of skin cancer.

  • Fair skin: Individuals with fair skin, light hair, and blue or green eyes are more susceptible to skin damage from UV radiation.

  • Early detection: Skin cancer is often detected earlier due to its visible nature, leading to a higher reported prevalence.

It is important to note that while these cancers are more prevalent, efforts to promote awareness, lifestyle modifications, and early detection have the potential to reduce their impact. Prevention and education are crucial in addressing the higher prevalence of these cancers and improving overall public health.

This paper compares the capabilities of different artificial intelligence (AI) approaches to achieve early cancer detection and proactive healthcare. This methodological innovation empowers accurate analysis of intricate medical data, surpassing traditional methods and enabling the identification of subtle disease markers. The physical significance is in the early detection of diseases, often imperceptible to the human eye, leading to expedited medical interventions and improved patient outcomes.

From a physical and engineering perspective, the work is significant in its ability to detect these subtle markers, which is vital for early interventions, potentially saving lives and improving well-being. The speed and precision of AI algorithms, as highlighted in the research, represent an engineering milestone. It transforms healthcare by making AI an indispensable component, promising efficiency, accuracy, and patient-centered care.

Let us delve into the world of early disease detection and its intriguing connection with artificial intelligence (AI). This review takes us on a journey through existing research, disease possibilities, and the magic AI brings to medical diagnosis.

Imagine spotting diseases when they are just tiny sparks, before they become full-blown fires. That is the goal of early disease detection. Diseases such as cancer, heart problems, and brain disorders can be tamed if caught early. But this is not easy. Sometimes, diseases hide well, and current methods cannot always find them in time.

Now, here is where AI steps in—a tech superhero. AI is like a super brain that can chew through loads of information faster than a flash. Machine learning is one trick AI uses. It is like a detective that learns from data to solve medical mysteries. Then, there is deep learning, which is like AI’s way of understanding pictures. It spots even the tiniest changes in x rays, Magnetic resonance imaging (MRI), and more.

Research shows AI’s power. It is helping find cancers in x rays, spotting heart problems before they strike, and even predicting brain issues. But, there is a twist. AI needs a lot of special data to learn from, like a teacher needs textbooks. Also, AI sometimes works like a secret code—it makes decisions, but we are not always sure why.

In a nutshell, AI and early disease detection are a powerful duo, but there are puzzles to solve. While we are moving forward, we need to think about data, understand AI’s decisions, and make sure it is helping everyone fairly. With AI by our side, we are stepping into a future where diseases might not surprise us anymore—and that is a thrilling thought.

Unlocking the Magic of Deep Learning: Deep learning is like the brain of AI—it is the part that learns and understands patterns. Within deep learning, there are some special techniques that are superheroes in cancer diagnosis.

Convolutional Neural Networks (CNNs): Imagine CNNs as art critics for images. They break down the image into tiny parts, analyze them, and then put everything back together. This helps them recognize complex patterns in medical images, such as x rays or Computed tomography (CT) scans. They are especially good at spotting shapes and structures that matter in cancer diagnosis.

Recurrent Neural Networks (RNNs): RNNs are like the memory masters of AI. They can analyze sequences of data, which is helpful for tasks such as reading patient records over time. In cancer diagnosis, RNNs can follow a patient’s history and catch any changes that might indicate cancer development.

Advantages in Medical World: These techniques have superpowers in the medical world. CNNs can find tiny hints of tumors in scans, even if they are buried in complex images. RNNs can spot trends in patients’ data over time—something human eyes might miss.

Why They Matter? In cancer diagnosis, accuracy is everything. Deep learning techniques boost that accuracy. They can pinpoint subtle signs of cancer earlier than ever before. When cancer is caught early, treatment can start sooner, leading to better outcomes.

So, think of these deep learning techniques as AI’s secret weapons. They are trained to see what the human eye cannot, making them invaluable allies in the fight against cancer.

Breast cancer remains a significant global health concern, with early detection playing a pivotal role in improving patient outcomes. Advances in artificial intelligence (AI), particularly the use of convolutional neural networks (CNNs), have opened up new avenues for accurate and timely breast cancer detection through mammography analysis. In this discussion, we delve into the innovative approach of using AI techniques to identify breast cancer at its early stages, revolutionizing the way we approach this critical health issue.

The Significance of Early Detection: Early detection is crucial in the fight against breast cancer. When detected early, the chances of successful treatment and survival increase significantly. Mammography, an x-ray imaging technique, has been a cornerstone in breast cancer screening for decades. However, interpreting mammograms can be challenging due to the complexity of breast tissue patterns and the subtle nature of early-stage cancers. This is where AI steps in to enhance accuracy and efficiency.

Artificial Intelligence and Convolutional Neural Networks (CNNs): AI, specifically CNNs, has shown remarkable promise in improving mammography analysis. CNNs are a type of deep learning algorithm inspired by the human visual system. These networks excel in recognizing intricate patterns in images, making them ideally suited for identifying the intricate features indicative of early breast cancer.

How CNNs Aid Early Detection: CNNs process mammograms with incredible precision, identifying abnormalities that might be missed by the human eye. The process involves training the CNN using a vast dataset of mammograms, allowing it to learn and differentiate between normal and abnormal breast tissue patterns. Once trained, the CNN can accurately identify potential cancerous lesions by detecting irregularities in the mammogram images.

Benefits of AI-Enhanced Early Detection: The integration of AI in early breast cancer detection brings several notable benefits. First, AI can reduce false negatives and false positives, minimizing unnecessary anxiety and interventions for patients. Second, AI systems can assist radiologists by highlighting suspicious areas, allowing for a more focused and efficient review of mammograms. This collaborative approach improves diagnostic accuracy and speeds up the interpretation process.

Challenges and Progress: While AI shows great potential, there are challenges to address. Ensuring that AI models are well-trained and unbiased is crucial. Additionally, the integration of AI into clinical practice requires thorough validation and regulatory approval to guarantee patient safety.

Future Prospects: The future of AI in breast cancer detection is bright. Continued research is expected to refine existing AI algorithms, making them even more accurate and robust. Moreover, the combination of AI with other imaging modalities and genetic information holds promise for more comprehensive early detection strategies.

The marriage of AI and mammography analysis represents a transformative leap in early breast cancer detection. The ability of convolutional neural networks to unravel the intricacies of mammograms offers hope for improved patient outcomes and reduced mortality rates. By leveraging AI’s analytical prowess, healthcare is advancing toward more efficient, accurate, and timely breast cancer detection, thereby making significant strides in the battle against this formidable disease. Tables IIII provide a comparative overview of key parameters across the selected research papers, enabling similarities, differences, and trends among them.

TABLE I.

Data and methods comparison.

Paper titleDataset size and
and authorsDataset usedcharacteristicsAI techniquesImaging modality
Esteva et al. (2017)1  Diverse medical images 129 450 clinical images of skin lesions Deep learning Medical images 
Kermany et al. (2018)2  Diverse medical images 128 175 retinal images from 86 129 patients Deep learning Medical images 
Shen et al. (2017)3  Medical images Various medical image datasets, sizes not specified Deep learning Medical images 
Nam et al. (2018)4  Chest radiographs 5232 chest radiographs with malignant pulmonary nodules Deep learning Radiographs 
Goodfellow et al. (2013)5  Diverse medical images MNIST, CIFAR-10, STL-10 datasets Machine learning Medical images 
Ehteshami Bejnordi et al. (2017)6  Breast histopathology 2534 lymph node images from breast cancer patients Deep learning Histopathology images 
Liu et al. (2017)7  Pathology images Camelyon16 and Camelyon17 datasets with gigapixel pathology images Deep learning Pathology images 
Komura and Ishikawa (2018)8  Histopathology images Publicly available breast cancer histopathology images Machine learning Histopathology images 
Wang et al. (2020)9  Lung pathology images Lung cancer pathology image dataset with 1276 images AI algorithms Pathology images 
Arvind et al. (2020)10  Breast histopathology 756 breast cancer nodal images from 189 patients Artificial intelligence Histopathology images 
Araújo and Aresta (2017)11  Breast cancer histology images Break his dataset with 7909 breast cancer histology images Convolutional neural networks (CNN) Histology images 
Yala et al. (2019)12  Mammography images Mammography images from 86 297 women Deep learning Mammography images 
Lotter, W., Diab, A. R., Haslam, B. et al. (2019)13  Digital breast tomosynthesis images DBT datasets with digital breast tomosynthesis images Deep learning Digital breast tomosynthesis images 
Zhang et al. (2017)14  Magnetic resonance imaging (MRI) 80 MRI scans from 40 healthy women Deep learning Magnetic resonance imaging (MRI) 
Lo et al. (2019)15  Breast MRI images Breast MRI images from 2907 examinations Machine learning Breast MRI images 
Ribli et al. (2018)16  Mammograms DDSM dataset with 2620 mammograms Deep learning Mammograms 
Paper titleDataset size and
and authorsDataset usedcharacteristicsAI techniquesImaging modality
Esteva et al. (2017)1  Diverse medical images 129 450 clinical images of skin lesions Deep learning Medical images 
Kermany et al. (2018)2  Diverse medical images 128 175 retinal images from 86 129 patients Deep learning Medical images 
Shen et al. (2017)3  Medical images Various medical image datasets, sizes not specified Deep learning Medical images 
Nam et al. (2018)4  Chest radiographs 5232 chest radiographs with malignant pulmonary nodules Deep learning Radiographs 
Goodfellow et al. (2013)5  Diverse medical images MNIST, CIFAR-10, STL-10 datasets Machine learning Medical images 
Ehteshami Bejnordi et al. (2017)6  Breast histopathology 2534 lymph node images from breast cancer patients Deep learning Histopathology images 
Liu et al. (2017)7  Pathology images Camelyon16 and Camelyon17 datasets with gigapixel pathology images Deep learning Pathology images 
Komura and Ishikawa (2018)8  Histopathology images Publicly available breast cancer histopathology images Machine learning Histopathology images 
Wang et al. (2020)9  Lung pathology images Lung cancer pathology image dataset with 1276 images AI algorithms Pathology images 
Arvind et al. (2020)10  Breast histopathology 756 breast cancer nodal images from 189 patients Artificial intelligence Histopathology images 
Araújo and Aresta (2017)11  Breast cancer histology images Break his dataset with 7909 breast cancer histology images Convolutional neural networks (CNN) Histology images 
Yala et al. (2019)12  Mammography images Mammography images from 86 297 women Deep learning Mammography images 
Lotter, W., Diab, A. R., Haslam, B. et al. (2019)13  Digital breast tomosynthesis images DBT datasets with digital breast tomosynthesis images Deep learning Digital breast tomosynthesis images 
Zhang et al. (2017)14  Magnetic resonance imaging (MRI) 80 MRI scans from 40 healthy women Deep learning Magnetic resonance imaging (MRI) 
Lo et al. (2019)15  Breast MRI images Breast MRI images from 2907 examinations Machine learning Breast MRI images 
Ribli et al. (2018)16  Mammograms DDSM dataset with 2620 mammograms Deep learning Mammograms 
TABLE II.

Performance metrics comparison.

Paper title and authorsSensitivity (%)Specificity (%)Accuracy (%)F1 scoreAUC
Esteva et al. (2017)1  72.1 89.1 ⋯ ⋯ ⋯ 
Kermany et al. (2018)2  81.1 68.4 74.5 ⋯ ⋯ 
Nam et al. (2018)4  90.2 91.6 91.0 ⋯ ⋯ 
Ehteshami Bejnordi et al. (2017)6  88.0 78.4 85.5 ⋯ ⋯ 
Arvind, S. et al. (2020) 74.7 72.0 ⋯ ⋯ ⋯ 
Araújo and Aresta (2017)11  88.9 81.6 84.3 0.78 ⋯ 
Yala et al. (2019)12  94.1 48.3 80.2 ⋯ 0.91 
Lotter et al. (2019)13  91.1 88.1 ⋯ ⋯ ⋯ 
Zhang et al. (2017)14  ⋯ ⋯ 98.4 ⋯ ⋯ 
Lo et al. (2019)15  93.6 77.8 ⋯ ⋯ ⋯ 
Ribli et al. (2018)16  87.3 88.5 ⋯ ⋯ ⋯ 
Paper title and authorsSensitivity (%)Specificity (%)Accuracy (%)F1 scoreAUC
Esteva et al. (2017)1  72.1 89.1 ⋯ ⋯ ⋯ 
Kermany et al. (2018)2  81.1 68.4 74.5 ⋯ ⋯ 
Nam et al. (2018)4  90.2 91.6 91.0 ⋯ ⋯ 
Ehteshami Bejnordi et al. (2017)6  88.0 78.4 85.5 ⋯ ⋯ 
Arvind, S. et al. (2020) 74.7 72.0 ⋯ ⋯ ⋯ 
Araújo and Aresta (2017)11  88.9 81.6 84.3 0.78 ⋯ 
Yala et al. (2019)12  94.1 48.3 80.2 ⋯ 0.91 
Lotter et al. (2019)13  91.1 88.1 ⋯ ⋯ ⋯ 
Zhang et al. (2017)14  ⋯ ⋯ 98.4 ⋯ ⋯ 
Lo et al. (2019)15  93.6 77.8 ⋯ ⋯ ⋯ 
Ribli et al. (2018)16  87.3 88.5 ⋯ ⋯ ⋯ 
TABLE III.

Clinical implications and integration.

Paper title and authorsClinical integrationLimitationsRobustness
Esteva et al. (2017)1  Dermatologist assistance Not standalone diagnosis Robust across lesions 
Kermany et al. (2018)2  Diagnosis assistance Diverse image sources Generalization potential 
Shen et al. (2017)3  Improved diagnostics Interpretability challenge ⋯ 
Nam et al. (2018)4  Radiologist support Limited interpretability Robust across radiographs 
Goodfellow et al. (2013)5  ⋯ Challenges in representation ⋯ 
Ehteshami Bejnordi et al. (2017)6  Pathologist support Varying image quality Robust for lymph node detection 
Liu et al. (2017)7  Pathologist support Gigapixel images ⋯ 
Komura and Ishikawa (2018)8  Pathology analysis ⋯ Robust for histopathology 
Wang et al. (2020)9  Pathology analysis Limited clinical impact ⋯ 
Arvind et al. (2020)10  Pathologist support Black box concern ⋯ 
Araújo and Aresta (2017)11  Improved histology analysis Limited to histology images Robust for histology 
Yala et al. (2019)12  Enhanced risk prediction Limited to mammography Robust for risk prediction 
Lotter et al. (2019)13  Improved tomosynthesis diagnosis Limited to digital breast tomosynthesis Robust for tomosynthesis 
Zhang et al. (2017)14  Accurate breast tissue analysis Limited to MRI Robust for MRI analysis 
Lo, C. M. et al. (2019)15  Enhanced MRI interpretation Limited to MRI Robust for MRI analysis 
Ribli et al. (2018)16  Improved mammogram analysis Limited to mammography Robust for mammography 
Paper title and authorsClinical integrationLimitationsRobustness
Esteva et al. (2017)1  Dermatologist assistance Not standalone diagnosis Robust across lesions 
Kermany et al. (2018)2  Diagnosis assistance Diverse image sources Generalization potential 
Shen et al. (2017)3  Improved diagnostics Interpretability challenge ⋯ 
Nam et al. (2018)4  Radiologist support Limited interpretability Robust across radiographs 
Goodfellow et al. (2013)5  ⋯ Challenges in representation ⋯ 
Ehteshami Bejnordi et al. (2017)6  Pathologist support Varying image quality Robust for lymph node detection 
Liu et al. (2017)7  Pathologist support Gigapixel images ⋯ 
Komura and Ishikawa (2018)8  Pathology analysis ⋯ Robust for histopathology 
Wang et al. (2020)9  Pathology analysis Limited clinical impact ⋯ 
Arvind et al. (2020)10  Pathologist support Black box concern ⋯ 
Araújo and Aresta (2017)11  Improved histology analysis Limited to histology images Robust for histology 
Yala et al. (2019)12  Enhanced risk prediction Limited to mammography Robust for risk prediction 
Lotter et al. (2019)13  Improved tomosynthesis diagnosis Limited to digital breast tomosynthesis Robust for tomosynthesis 
Zhang et al. (2017)14  Accurate breast tissue analysis Limited to MRI Robust for MRI analysis 
Lo, C. M. et al. (2019)15  Enhanced MRI interpretation Limited to MRI Robust for MRI analysis 
Ribli et al. (2018)16  Improved mammogram analysis Limited to mammography Robust for mammography 

Lung cancer is a critical global health concern, and early detection is pivotal for improving patient outcomes. Recent advancements in deep learning have opened new avenues for accurate and timely lung cancer detection through the analysis of CT scans. In this discussion, we delve into the innovative approach of using deep learning models to identify potential lung cancer nodules in CT images, transforming the landscape of lung cancer diagnosis.

The Importance of Early Detection: Early detection of lung cancer is associated with better prognosis and improved survival rates. Computed tomography (CT) scans are a powerful diagnostic tool that can reveal even subtle abnormalities within the lungs, including pulmonary nodules that could be indicative of cancer.

Deep Learning and CT Scans: Deep learning, a subset of artificial intelligence, has shown remarkable promise in the field of medical imaging analysis. Deep learning models, particularly convolutional neural networks (CNNs), have demonstrated their ability to process and interpret complex images, making them ideal candidates for analyzing CT scans for lung cancer detection.

Application of Deep Learning Models: Deep learning models, especially CNNs, have the capacity to process CT scan images with a level of detail that can be challenging for human radiologists to achieve consistently. These models are trained on extensive datasets containing CT scans, allowing them to learn and recognize patterns associated with pulmonary nodules that might signify lung cancer.

How Deep Learning Aids Early Detection: Deep learning models excel at identifying intricate features in CT scans that may escape human observation. They can accurately pinpoint the presence, size, and characteristics of pulmonary nodules, which are often indicative of early-stage lung cancer. The models provide an additional layer of analysis, complementing the expertise of radiologists.

Benefits of Deep Learning-Enhanced Detection: Integrating deep learning into lung cancer detection offers several advantages. It can assist radiologists by flagging suspicious regions in CT images, facilitating a more focused and precise review. This collaborative approach between AI and medical professionals enhances diagnostic accuracy and potentially expedites the identification of lung cancer nodules.

Challenges and Progress: While deep learning holds immense potential, challenges remain. Ensuring that deep learning models are trained on diverse and representative datasets is essential to avoid biases. Additionally, these models need validation and rigorous testing before they can be integrated into clinical practice.

Future Prospects: The future of deep learning in lung cancer detection is promising. Ongoing research is expected to refine existing algorithms and improve their sensitivity and specificity. Moreover, the integration of deep learning with other diagnostic methods may lead to more comprehensive and accurate early lung cancer detection strategies. Tables IVVI provide a comparative overview of key parameters across the selected research papers on early detection of lung cancer.

TABLE IV.

Data and methods comparison.

Paper titleDataset size and
and authorsDataset usedcharacteristicsAI techniquesImaging modality
Setio et al. (2017)17  LUNA16 challenge ⋯ Various algorithms for nodule detection and comparison CT images 
Hu et al. (2017)18  Private dataset ⋯ Deep transfer learning Chest CT images 
Yap et al. (2018)19  ⋯ ⋯ Convolutional neural networks CT scans of lungs 
Shen et al. (2015)20  LIDC-IDRI 888 nodules, 1186 non-nodules Multi-scale convolutional neural networks CT images 
Ciompi et al. (2017)21  LIDC-IDRI, NLST 888 nodules, 1186 non-nodules Deep learning for nodule management CT scans 
Christodoulidis et al. (2017)22  Multiple datasets ⋯ Multi-source transfer learning with CNNs Lung patterns in CT images 
Dou et al. (2016)23  LIDC-IDRI ⋯ Multilevel contextual 3D CNNs Pulmonary nodule detection 
Zhang et al. (2017)24  Private dataset ⋯ Deep CNN and SVM classifier Peripherally located nodules in CT scans 
Ding et al. (2017)25  Private dataset ⋯ Template matching and deep learning CT images 
Ardila et al. (2019)26  NLST 6000 CT scans 3D deep learning Low-dose chest CT 
Liang et al. (2019)27  LIDC-IDRI ⋯ 3D deep convolutional neural networks CT images 
Shen et al. (2016)28  TCGA 1624 patients Developing transferable deep features Patient-level lung cancer prediction 
Zhang et al. (2018)29  Private dataset ⋯ Deep CNN ensemble Small lung cancer detection 
Anthimopoulos et al. (2016)30  ⋯ ⋯ Deep convolutional neural network Interstitial lung disease patterns 
De Fauw et al. (2018)31  UK Biobank ⋯ Clinically applicable deep learning Retinal disease diagnosis 
Baltruschat et al. (2020)32  LIDC-IDRI ⋯ Advanced deep learning Pulmonary nodule detection and classification 
Shen et al. (2017)33  LIDC-IDRI ⋯ Multi-crop convolutional neural networks Lung nodule malignancy classification 
Bi et al. (2019)34  LIDC-IDRI ⋯ Texture features, deep learning, SVM Cancer lesion detection and segmentation in PET/CT images 
Sun et al. (2018)35  ⋯ ⋯ Multi-level context convolutional network Lung nodule classification 
Paper titleDataset size and
and authorsDataset usedcharacteristicsAI techniquesImaging modality
Setio et al. (2017)17  LUNA16 challenge ⋯ Various algorithms for nodule detection and comparison CT images 
Hu et al. (2017)18  Private dataset ⋯ Deep transfer learning Chest CT images 
Yap et al. (2018)19  ⋯ ⋯ Convolutional neural networks CT scans of lungs 
Shen et al. (2015)20  LIDC-IDRI 888 nodules, 1186 non-nodules Multi-scale convolutional neural networks CT images 
Ciompi et al. (2017)21  LIDC-IDRI, NLST 888 nodules, 1186 non-nodules Deep learning for nodule management CT scans 
Christodoulidis et al. (2017)22  Multiple datasets ⋯ Multi-source transfer learning with CNNs Lung patterns in CT images 
Dou et al. (2016)23  LIDC-IDRI ⋯ Multilevel contextual 3D CNNs Pulmonary nodule detection 
Zhang et al. (2017)24  Private dataset ⋯ Deep CNN and SVM classifier Peripherally located nodules in CT scans 
Ding et al. (2017)25  Private dataset ⋯ Template matching and deep learning CT images 
Ardila et al. (2019)26  NLST 6000 CT scans 3D deep learning Low-dose chest CT 
Liang et al. (2019)27  LIDC-IDRI ⋯ 3D deep convolutional neural networks CT images 
Shen et al. (2016)28  TCGA 1624 patients Developing transferable deep features Patient-level lung cancer prediction 
Zhang et al. (2018)29  Private dataset ⋯ Deep CNN ensemble Small lung cancer detection 
Anthimopoulos et al. (2016)30  ⋯ ⋯ Deep convolutional neural network Interstitial lung disease patterns 
De Fauw et al. (2018)31  UK Biobank ⋯ Clinically applicable deep learning Retinal disease diagnosis 
Baltruschat et al. (2020)32  LIDC-IDRI ⋯ Advanced deep learning Pulmonary nodule detection and classification 
Shen et al. (2017)33  LIDC-IDRI ⋯ Multi-crop convolutional neural networks Lung nodule malignancy classification 
Bi et al. (2019)34  LIDC-IDRI ⋯ Texture features, deep learning, SVM Cancer lesion detection and segmentation in PET/CT images 
Sun et al. (2018)35  ⋯ ⋯ Multi-level context convolutional network Lung nodule classification 
TABLE V.

Performance metrics comparison.

Paper title and authorsDeep learning architecturePreprocessing techniquesTraining strategiesValidation techniques
Setio et al. (2017)17  Various architectures: Single-shot detectors, region-based CNNs Standard preprocessing (resizing, normalization) Diverse training data, ensembles, false positive reduction Cross-validation, external validation datasets 
Hu et al. (2017)18  Transfer learning with deep CNNs Intensity normalization, lung segmentation Transfer learning from ImageNet, fine-tuning Cross-validation, external validation dataset 
Yap et al. (2018)19  CNN-based architecture Standard preprocessing (windowing, normalization) Transfer learning, fine-tuning Cross-validation, external validation datasets 
Shen et al. (2015)20  Multi-scale CNNs Standard preprocessing (resizing, normalization) Training on large dataset Cross-validation, independent validation dataset 
Ciompi et al. (2017)21  CNN-based architecture Standard preprocessing (resizing, normalization) Transfer learning Cross-validation, external validation dataset 
Christodoulidis et al. (2017)22  Transfer learning with CNNs Standard preprocessing (resizing, normalization) Transfer learning from multiple sources Cross-validation, external validation dataset 
Dou et al. (2016)23  3D CNNs Standard preprocessing (windowing, normalization) Multilevel contextual learning Cross-validation, independent validation dataset 
Zhang et al. (2017)24  Deep CNN and SVM classifier Standard preprocessing (resizing, normalization) End-to-end training Cross-validation, independent validation dataset 
Ding et al. (2017)25  Deep CNN and template matching Standard preprocessing (resizing, normalization) End-to-end training Cross-validation, independent validation dataset 
Ardila et al. (2019)26  3D CNNs Standard preprocessing (resampling, normalization) End-to-end training Cross-validation, independent validation dataset 
Liang et al. (2019)27  3D CNNs Standard preprocessing (resizing, normalization) End-to-end training Cross-validation, independent validation dataset 
Shen et al. (2016)28  Deep CNN Not specified Transfer learning from experts Cross-validation, independent validation dataset 
Zhang et al. (2018)29  Ensemble of CNNs Standard preprocessing (resizing, normalization) Ensemble learning Cross-validation, independent validation dataset 
Anthimopoulos et al. (2016)30  Deep CNN Standard preprocessing (resizing, normalization) End-to-end training Cross-validation, independent validation dataset 
De Fauw et al. (2018)31  CNN with attention mechanism Not specified Transfer learning with fine-tuning Cross-validation, independent validation dataset 
Shen et al. (2017)33  Multi-crop CNN Not specified Transfer learning Cross-validation, independent validation dataset 
Bi et al. (2019)34  CNN, SVM Not specified Transfer learning, SVM training Cross-validation, independent validation dataset 
Sun et al. (2018)35  Multi-level context CNN Not specified Transfer learning Cross-validation, independent validation dataset 
Paper title and authorsDeep learning architecturePreprocessing techniquesTraining strategiesValidation techniques
Setio et al. (2017)17  Various architectures: Single-shot detectors, region-based CNNs Standard preprocessing (resizing, normalization) Diverse training data, ensembles, false positive reduction Cross-validation, external validation datasets 
Hu et al. (2017)18  Transfer learning with deep CNNs Intensity normalization, lung segmentation Transfer learning from ImageNet, fine-tuning Cross-validation, external validation dataset 
Yap et al. (2018)19  CNN-based architecture Standard preprocessing (windowing, normalization) Transfer learning, fine-tuning Cross-validation, external validation datasets 
Shen et al. (2015)20  Multi-scale CNNs Standard preprocessing (resizing, normalization) Training on large dataset Cross-validation, independent validation dataset 
Ciompi et al. (2017)21  CNN-based architecture Standard preprocessing (resizing, normalization) Transfer learning Cross-validation, external validation dataset 
Christodoulidis et al. (2017)22  Transfer learning with CNNs Standard preprocessing (resizing, normalization) Transfer learning from multiple sources Cross-validation, external validation dataset 
Dou et al. (2016)23  3D CNNs Standard preprocessing (windowing, normalization) Multilevel contextual learning Cross-validation, independent validation dataset 
Zhang et al. (2017)24  Deep CNN and SVM classifier Standard preprocessing (resizing, normalization) End-to-end training Cross-validation, independent validation dataset 
Ding et al. (2017)25  Deep CNN and template matching Standard preprocessing (resizing, normalization) End-to-end training Cross-validation, independent validation dataset 
Ardila et al. (2019)26  3D CNNs Standard preprocessing (resampling, normalization) End-to-end training Cross-validation, independent validation dataset 
Liang et al. (2019)27  3D CNNs Standard preprocessing (resizing, normalization) End-to-end training Cross-validation, independent validation dataset 
Shen et al. (2016)28  Deep CNN Not specified Transfer learning from experts Cross-validation, independent validation dataset 
Zhang et al. (2018)29  Ensemble of CNNs Standard preprocessing (resizing, normalization) Ensemble learning Cross-validation, independent validation dataset 
Anthimopoulos et al. (2016)30  Deep CNN Standard preprocessing (resizing, normalization) End-to-end training Cross-validation, independent validation dataset 
De Fauw et al. (2018)31  CNN with attention mechanism Not specified Transfer learning with fine-tuning Cross-validation, independent validation dataset 
Shen et al. (2017)33  Multi-crop CNN Not specified Transfer learning Cross-validation, independent validation dataset 
Bi et al. (2019)34  CNN, SVM Not specified Transfer learning, SVM training Cross-validation, independent validation dataset 
Sun et al. (2018)35  Multi-level context CNN Not specified Transfer learning Cross-validation, independent validation dataset 
TABLE VI.

Clinical implications and integration.

Paper title
and authorsClinical integrationLimitationsRobustness
Setio et al. (2017)17  Demonstrates potential integration into clinical workflows, aiding radiologists in nodule detection and early intervention False positives/negatives, technical challenges, algorithm complexity Algorithms show robust performance across multiple datasets, allowing consistent nodule detection 
Hu et al. (2017)18  Offers potential integration by aiding radiologists in nodule diagnosis, improving accuracy Limited explanation of model decisions, possible overfitting Demonstrates robustness across various nodule characteristics, contributing to accurate diagnosis 
Yap et al. (2018)19  Potential for clinical use in aiding lung cancer diagnosis Limited discussion on false positive reduction, dataset size Achieves robust performance with deep CNN architecture and lung cancer identification 
Shen et al. (2015)20  Offers potential for improved lung nodule classification Limited focus on computational efficiency Demonstrates robustness with multi-scale CNNs for nodule classification and malignancy prediction 
Ciompi et al. (2017)21  Shows potential for automating nodule management in lung cancer screening Limited discussion on false positive reduction, lack of clinical implementation details Demonstrates robustness in automating pulmonary nodule management tasks using deep learning-based approaches 
Christodoulidis et al. (2017)22  Offers potential for improved lung pattern analysis and disease classification Limited discussion on clinical validation and real-world application Demonstrates robustness in leveraging multi-source transfer learning for lung pattern analysis 
Dou et al. (2016)23  Holds promise for reducing false positives in pulmonary nodule detection Limited validation on a single dataset, sensitivity to training data quality Demonstrates robustness with multilevel contextual 3D CNNs for reducing false positives in nodule detection 
Zhang et al. (2017)24  Potential for peripheral nodule detection, but lacking clinical validation Limited validation on a single dataset, performance dependent on SVM classifier Demonstrates robustness in peripheral nodule detection using deep CNN and SVM 
Ding et al. (2017)25  Promising approach for nodule detection, but limited clinical validation Reliance on template matching may hinder generalization, validation on a single dataset Demonstrates robustness in combining template matching and deep learning for nodule detection 
Ardila et al. (2019)26  Offers end-to-end lung cancer screening, but requires further clinical validation Limited discussion on false positives and real-world integration Demonstrates robustness in achieving end-to-end lung cancer screening with 3D deep learning on low-dose CT scans 
Liang et al. (2019)27  Offers patient-level lung cancer prediction, but requires clinical validation Limited information on real-world implementation, potential overfitting to expert features Demonstrates robustness in transferable deep features for patient-level lung cancer prediction 
Zhang et al. (2018)29  Promising for detecting small lung cancers, but requires more clinical validation Limited discussion on false positives/negatives, single dataset validation Demonstrates robustness in ensemble-based deep learning for small lung cancer detection in CT scans 
Anthimopoulos et al. (2016)30  Potential for automated interstitial lung disease classification, but requires further validation Limited discussion on real-world implementation, potential performance variations across datasets Demonstrates robustness in deep learning for lung pattern classification in interstitial lung diseases 
De Fauw et al. (2018)31  Presents clinically applicable deep learning for retinal disease, but focuses on retinal diagnosis Not specific to lung disease, limited focus on lung-related applications Demonstrates robustness in deep learning for medical diagnosis and referral, applicable to various medical conditions 
Baltruschat et al.(2020)32  Addresses pulmonary nodule detection and classification, but clinical integration details are limited Limited discussion on integration, potential performance variations in clinical settings Demonstrates robustness in advanced deep learning for pulmonary nodule detection and classification 
Paper title
and authorsClinical integrationLimitationsRobustness
Setio et al. (2017)17  Demonstrates potential integration into clinical workflows, aiding radiologists in nodule detection and early intervention False positives/negatives, technical challenges, algorithm complexity Algorithms show robust performance across multiple datasets, allowing consistent nodule detection 
Hu et al. (2017)18  Offers potential integration by aiding radiologists in nodule diagnosis, improving accuracy Limited explanation of model decisions, possible overfitting Demonstrates robustness across various nodule characteristics, contributing to accurate diagnosis 
Yap et al. (2018)19  Potential for clinical use in aiding lung cancer diagnosis Limited discussion on false positive reduction, dataset size Achieves robust performance with deep CNN architecture and lung cancer identification 
Shen et al. (2015)20  Offers potential for improved lung nodule classification Limited focus on computational efficiency Demonstrates robustness with multi-scale CNNs for nodule classification and malignancy prediction 
Ciompi et al. (2017)21  Shows potential for automating nodule management in lung cancer screening Limited discussion on false positive reduction, lack of clinical implementation details Demonstrates robustness in automating pulmonary nodule management tasks using deep learning-based approaches 
Christodoulidis et al. (2017)22  Offers potential for improved lung pattern analysis and disease classification Limited discussion on clinical validation and real-world application Demonstrates robustness in leveraging multi-source transfer learning for lung pattern analysis 
Dou et al. (2016)23  Holds promise for reducing false positives in pulmonary nodule detection Limited validation on a single dataset, sensitivity to training data quality Demonstrates robustness with multilevel contextual 3D CNNs for reducing false positives in nodule detection 
Zhang et al. (2017)24  Potential for peripheral nodule detection, but lacking clinical validation Limited validation on a single dataset, performance dependent on SVM classifier Demonstrates robustness in peripheral nodule detection using deep CNN and SVM 
Ding et al. (2017)25  Promising approach for nodule detection, but limited clinical validation Reliance on template matching may hinder generalization, validation on a single dataset Demonstrates robustness in combining template matching and deep learning for nodule detection 
Ardila et al. (2019)26  Offers end-to-end lung cancer screening, but requires further clinical validation Limited discussion on false positives and real-world integration Demonstrates robustness in achieving end-to-end lung cancer screening with 3D deep learning on low-dose CT scans 
Liang et al. (2019)27  Offers patient-level lung cancer prediction, but requires clinical validation Limited information on real-world implementation, potential overfitting to expert features Demonstrates robustness in transferable deep features for patient-level lung cancer prediction 
Zhang et al. (2018)29  Promising for detecting small lung cancers, but requires more clinical validation Limited discussion on false positives/negatives, single dataset validation Demonstrates robustness in ensemble-based deep learning for small lung cancer detection in CT scans 
Anthimopoulos et al. (2016)30  Potential for automated interstitial lung disease classification, but requires further validation Limited discussion on real-world implementation, potential performance variations across datasets Demonstrates robustness in deep learning for lung pattern classification in interstitial lung diseases 
De Fauw et al. (2018)31  Presents clinically applicable deep learning for retinal disease, but focuses on retinal diagnosis Not specific to lung disease, limited focus on lung-related applications Demonstrates robustness in deep learning for medical diagnosis and referral, applicable to various medical conditions 
Baltruschat et al.(2020)32  Addresses pulmonary nodule detection and classification, but clinical integration details are limited Limited discussion on integration, potential performance variations in clinical settings Demonstrates robustness in advanced deep learning for pulmonary nodule detection and classification 

The synergy between deep learning and CT scan analysis is revolutionizing early lung cancer detection. By harnessing the analytical capabilities of deep learning models, healthcare is moving toward more efficient, accurate, and timely identification of potential lung cancer nodules. This approach has the potential to significantly impact patient outcomes by enabling earlier intervention and treatment, ultimately contributing to the fight against lung cancer.

Detecting skin cancer early can save lives, and technology is stepping in to help. The marriage of artificial intelligence (AI) and computer vision has paved the way for a novel approach to identifying skin cancer, particularly melanoma, using images of the skin’s surface. In this discussion, we will explore how AI, guided by computer vision, is making strides in skin cancer detection by examining the visual cues present in dermatology images.

The Crucial Role of Early Detection: Spotting skin cancer, especially aggressive types like melanoma, in its early stages is vital for successful treatment. Quick diagnosis can lead to effective interventions, reducing the risk of cancer spreading or becoming more challenging to handle.

AI and Computer Vision in Dermatology: Artificial intelligence, the technology behind computer vision, has shown remarkable potential in analyzing images, including medical ones. When applied to dermatology images of the skin, AI can recognize patterns, textures, and irregularities that might indicate the presence of skin cancer, such as melanoma.

Unlocking Insights through Computer Vision: AI models, driven by computer vision, are trained on a diverse array of dermatology images. These images cover a wide spectrum, from healthy skin to those showing signs of skin cancer. AI learns to distinguish between normal skin and skin with potential cancerous characteristics. It is like teaching AI to spot the differences between puzzle pieces.

How Computer Vision Boosts Early Detection: Computer vision models are pros at identifying tiny details within complex images. By carefully examining skin lesions, moles, and other marks, these models can uncover unusual features that might signal skin cancer. This allows for timely medical attention and diagnosis.

Perks of AI-Enhanced Diagnosis: Integrating AI and computer vision into skin cancer detection offers several advantages. It is like having a highly skilled assistant for dermatologists. AI can systematically analyze images, potentially reducing misdiagnoses. Plus, it assists doctors by highlighting areas that need special attention, supporting their decision-making process.

Challenges and Progress: Despite the potential, challenges exist. Ensuring AI models learn from a diverse range of images is crucial to avoid biases. While AI offers exciting possibilities, ensuring its accuracy in real-world scenarios and integrating it into medical practice require careful validation.

What Lies Ahead: The future of skin cancer detection with AI is promising. Research continues to refine algorithms, making them even better at identifying potential issues. As technology evolves, we might even see AI-powered smartphone apps or devices for easier skin cancer screening. The parameter comparison of various research papers on early detection of skin cancer is given in Tables VII and VIII.

TABLE VII.

Data and methods comparison.

Paper titleDataset size and
and authorsDataset usedcharacteristicsAI techniquesImaging modality
Esteva et al. (2017)11  Not specified (proprietary dataset) Not specified Deep neural networks Dermoscopy (dermatoscopic images) 
Tschandl et al. (2018)36  HAM10000 dataset Large collection of multi-source dermatoscopic images of pigmented skin lesions Not specified Dermoscopy (dermatoscopic images) 
Brinker et al. (2019)37  ISIC archive Not specified Deep learning Histopathological images (biopsy samples) 
Haenssle et al. (2018)38  Not specified (proprietary dataset) Not specified Deep learning convolutional neural network Dermoscopy 
Codella et al. (2018)39  SIC 2017 skin lesion dataset Contains 2000 images of skin lesions Deep learning Dermoscopy 
Matsunaga and Hamada (2016)40  Not specified Not specified Image processing Dermoscopy 
Wiltgen et al. (2018)41  Not specified Not specified Deep learning In vivo imaging 
Kawahara et al. (2016)42  ISBI 2017 skin lesion challenge dataset Contains over 9000 dermoscopic images of skin lesions Deep learning Dermoscopy 
Schmidhuber (2015)43  Not applicable Review paper Review of deep learning Not applicable 
Celebi et al. (2009)44  Not applicable Review paper Review of techniques Dermoscopy 
Bi et al. (2018)45  ISIC 2017 challenge dataset Contains over 2000 skin lesion images Deep learning Dermoscopy 
Liu and Chen (2019)46  ISIC 2019 challenge dataset Contains over 2000 skin lesion images Deep learning Dermoscopy 
Haenssle et al. (2020)47  ISIC 2020 challenge dataset, private datasets Contains over 33 000 skin lesion images Deep learning Dermoscopy 
Menegola et al. (2017)48  ISIC Archive, PH2 dataset Contains various skin lesion images Transfer learning Dermoscopy 
Brinker et al. (2020)49  ISIC 2018 and 2019 challenge datasets Contains over 20 000 skin lesion images Deep learning Dermoscopy 
Esteva et al. (2019)50  Not applicable Review paper Review of techniques Not applicable 
Yap et al. (2019)51  PH2 dataset Contains dermoscopy images Deep learning Dermoscopy 
Codella et al. (2019)52  ISIC 2018 challenge dataset Contains skin lesion images Deep learning Dermoscopy 
Tschandl et al. (2019)53  Not specified Not specified Deep learning Not specified 
Paper titleDataset size and
and authorsDataset usedcharacteristicsAI techniquesImaging modality
Esteva et al. (2017)11  Not specified (proprietary dataset) Not specified Deep neural networks Dermoscopy (dermatoscopic images) 
Tschandl et al. (2018)36  HAM10000 dataset Large collection of multi-source dermatoscopic images of pigmented skin lesions Not specified Dermoscopy (dermatoscopic images) 
Brinker et al. (2019)37  ISIC archive Not specified Deep learning Histopathological images (biopsy samples) 
Haenssle et al. (2018)38  Not specified (proprietary dataset) Not specified Deep learning convolutional neural network Dermoscopy 
Codella et al. (2018)39  SIC 2017 skin lesion dataset Contains 2000 images of skin lesions Deep learning Dermoscopy 
Matsunaga and Hamada (2016)40  Not specified Not specified Image processing Dermoscopy 
Wiltgen et al. (2018)41  Not specified Not specified Deep learning In vivo imaging 
Kawahara et al. (2016)42  ISBI 2017 skin lesion challenge dataset Contains over 9000 dermoscopic images of skin lesions Deep learning Dermoscopy 
Schmidhuber (2015)43  Not applicable Review paper Review of deep learning Not applicable 
Celebi et al. (2009)44  Not applicable Review paper Review of techniques Dermoscopy 
Bi et al. (2018)45  ISIC 2017 challenge dataset Contains over 2000 skin lesion images Deep learning Dermoscopy 
Liu and Chen (2019)46  ISIC 2019 challenge dataset Contains over 2000 skin lesion images Deep learning Dermoscopy 
Haenssle et al. (2020)47  ISIC 2020 challenge dataset, private datasets Contains over 33 000 skin lesion images Deep learning Dermoscopy 
Menegola et al. (2017)48  ISIC Archive, PH2 dataset Contains various skin lesion images Transfer learning Dermoscopy 
Brinker et al. (2020)49  ISIC 2018 and 2019 challenge datasets Contains over 20 000 skin lesion images Deep learning Dermoscopy 
Esteva et al. (2019)50  Not applicable Review paper Review of techniques Not applicable 
Yap et al. (2019)51  PH2 dataset Contains dermoscopy images Deep learning Dermoscopy 
Codella et al. (2019)52  ISIC 2018 challenge dataset Contains skin lesion images Deep learning Dermoscopy 
Tschandl et al. (2019)53  Not specified Not specified Deep learning Not specified 
TABLE VIII.

Clinical implications and integration.

Paper title and authorsClinical integrationLimitationsRobustness
Esteva et al. (2017)11  High potential for clinical use and integration Limited diversity in data; overfitting risk Effective across various skin types 
Tschandl et al. (2018)36  Useful tool for dermatology practice Limited to dermatoscopic images Consistent performance on diverse datasets 
Brinker et al. (2019)37  Improved diagnostic accuracy; Clinically relevant Limited sample diversity; ethical considerations Consistent performance across cases 
Haenssle et al. (2018)38  Potential clinical use; valuable diagnostic tool Limited dataset size; Domain adaptation needed Comparable to dermatologists; Robust outcomes 
Codella et al. (2018)39  Contribution to diagnostic advancements Challenge-specific data; research-oriented Insight into algorithmic melanoma detection 
Haenssle et al. (2018)40  CNN-based architecture Preprocessing details Training methods 
Wiltgen et al. (2018)41  Real-time diagnosis; clinical application Limited dataset; real-world challenges Real-time performance; Robust diagnostic 
Kawahara et al. (2016)42  Enhanced lesion classification; diagnostic aid Limited to lesion classification; research Transferability; generalized lesion analysis 
Schmidhuber (2015)43  Deep learning overview; Informative Theoretical overview; not application-specific Overview of concepts; broad understanding 
Celebi et al. (2009)44  Improved lesion border detection; analysis Survey-based; focus on one aspect Broad overview; evaluation of techniques 
Bi et al. (2018)45  Enhanced lesion segmentation; clinical utility Patch-based; specific to segmentation Improved segmentation; patch-wise refinement 
Liu and Chen (2019)46  Mobile skin cancer detection; fast and efficient Mobile-focused; specific application Quick diagnosis; suitable for mobile apps 
Haenssle et al. (2020)47  Comparative analysis; performance evaluation Specific evaluation; market-approved model Performance comparison; diverse lesions 
Menegola et al. (2017)48  Transfer learning for skin lesion classification Transfer learning; Limited to classification Improved classification; shared knowledge 
Brinker et al. (2020)49  AI vs pathologists; classification comparison Comparative analysis; specific focus AI performance; comparison with experts 
Esteva et al. (2019)50  Deep learning in healthcare; comprehensive guide General overview; not focused on specific applications Guidance on implementation; healthcare-specific insights 
Yap et al. (2019)51  Keratinocyte detection; Automated analysis Patch-based; specific focus Automated detection; application to skin cancer images 
Codella et al. (2019)52  Melanoma detection; skin lesion analysis Challenge-specific; Limited to specific dataset Deep learning analysis; ISIC 2018 challenge 
Tschandl et al. (2019)53  Expert-level diagnosis; nonpigmented skin cancer Expert-based; limited to specific conditions Combined CNNs; enhanced diagnostic accuracy 
Paper title and authorsClinical integrationLimitationsRobustness
Esteva et al. (2017)11  High potential for clinical use and integration Limited diversity in data; overfitting risk Effective across various skin types 
Tschandl et al. (2018)36  Useful tool for dermatology practice Limited to dermatoscopic images Consistent performance on diverse datasets 
Brinker et al. (2019)37  Improved diagnostic accuracy; Clinically relevant Limited sample diversity; ethical considerations Consistent performance across cases 
Haenssle et al. (2018)38  Potential clinical use; valuable diagnostic tool Limited dataset size; Domain adaptation needed Comparable to dermatologists; Robust outcomes 
Codella et al. (2018)39  Contribution to diagnostic advancements Challenge-specific data; research-oriented Insight into algorithmic melanoma detection 
Haenssle et al. (2018)40  CNN-based architecture Preprocessing details Training methods 
Wiltgen et al. (2018)41  Real-time diagnosis; clinical application Limited dataset; real-world challenges Real-time performance; Robust diagnostic 
Kawahara et al. (2016)42  Enhanced lesion classification; diagnostic aid Limited to lesion classification; research Transferability; generalized lesion analysis 
Schmidhuber (2015)43  Deep learning overview; Informative Theoretical overview; not application-specific Overview of concepts; broad understanding 
Celebi et al. (2009)44  Improved lesion border detection; analysis Survey-based; focus on one aspect Broad overview; evaluation of techniques 
Bi et al. (2018)45  Enhanced lesion segmentation; clinical utility Patch-based; specific to segmentation Improved segmentation; patch-wise refinement 
Liu and Chen (2019)46  Mobile skin cancer detection; fast and efficient Mobile-focused; specific application Quick diagnosis; suitable for mobile apps 
Haenssle et al. (2020)47  Comparative analysis; performance evaluation Specific evaluation; market-approved model Performance comparison; diverse lesions 
Menegola et al. (2017)48  Transfer learning for skin lesion classification Transfer learning; Limited to classification Improved classification; shared knowledge 
Brinker et al. (2020)49  AI vs pathologists; classification comparison Comparative analysis; specific focus AI performance; comparison with experts 
Esteva et al. (2019)50  Deep learning in healthcare; comprehensive guide General overview; not focused on specific applications Guidance on implementation; healthcare-specific insights 
Yap et al. (2019)51  Keratinocyte detection; Automated analysis Patch-based; specific focus Automated detection; application to skin cancer images 
Codella et al. (2019)52  Melanoma detection; skin lesion analysis Challenge-specific; Limited to specific dataset Deep learning analysis; ISIC 2018 challenge 
Tschandl et al. (2019)53  Expert-level diagnosis; nonpigmented skin cancer Expert-based; limited to specific conditions Combined CNNs; enhanced diagnostic accuracy 

The fusion of AI and computer vision is transforming early skin cancer detection. By harnessing computer vision’s ability to scrutinize images, healthcare is moving toward quicker, more accurate, and timely identification of possible skin cancer, particularly melanoma. This approach holds the potential to make a significant impact on patient outcomes, ensuring that individuals receive the care they need promptly and effectively.

  • AI algorithms can analyze mammograms and spot subtle changes in breast tissue.

  • It helps find small tumors or abnormalities that might be missed in manual screenings.

  • Offers a higher sensitivity rate, leading to earlier detection and improved survival rates.

  • Can assist radiologists in making more accurate diagnoses.

  • AI examines chest CT scans and identifies potential lung nodules.

  • Detects nodules at an early stage, often before symptoms appear.

  • Increases chances of successful treatment and better patient outcomes.

  • Aids doctors in determining the malignancy risk of detected nodules.

  • AI analyzes dermatoscopic images of skin lesions to identify potential cancers.

  • Detects irregularities in skin patterns and textures that might indicate cancerous growth.

  • Provides quick and non-invasive assessment, reducing unnecessary biopsies.

  • Enhances early detection, leading to less invasive treatment options.

  • All three types benefit from AI’s ability to process vast amounts of data quickly.

  • Earlier detection increases treatment success rates and reduces healthcare costs.

  • AI helps reduce human error and enhances the efficiency of diagnosis.

  • Improves patient experience by offering more timely interventions.

  • AI systems need access to high-quality and diverse datasets for accurate training.

  • Ensuring AI’s reliability and robustness is crucial for real-world applications.

  • Ethical considerations, patient privacy, and data security are important concerns.

  • AI’s role in early cancer detection is expected to grow, making screenings more effective.

  • Continued research and collaboration will refine AI algorithms for even better accuracy.

  • Integration into healthcare systems will require close collaboration between AI experts and medical professionals.

In conclusion, AI’s application in early cancer detection for breast, lung, and skin cancers shows promising potential. It assists in catching cancers at their earliest stages, improving patient outcomes and healthcare practices. While challenges exist, the ongoing development of AI technology offers exciting opportunities to revolutionize cancer diagnosis and treatment.

In conclusion, the integration of AI-driven early cancer detection into healthcare marks an exciting new era. Through this review, we have seen how AI can work hand in hand with medical practices to catch cancers at their earliest stages. This has the potential to transform healthcare, making it more effective and accessible. AI is showing promising results in detecting breast, lung, and skin cancers, which are some of the most common and life-threatening cancers. As AI technology continues to advance and improve, we are on the brink of a future where diseases can be caught early, treatments can start sooner, and people can enjoy better well-being. This partnership between AI and healthcare holds great promise, paving the way for a healthier and brighter future for everyone.

The authors have no conflicts to disclose.

R. Deepa: Conceptualization (equal); Writing – original draft (equal). S. Arunkumar: Funding acquisition (equal); Resources (equal); Software (equal). V. Jayaraj: Methodology (equal). A. Sivasamy: Conceptualization (equal); Methodology (equal); Writing – original draft (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.

1
A.
Esteva
,
B.
Kuprel
,
R. A.
Novoa
,
J.
Ko
,
S. M.
Swetter
,
H. M.
Blau
, and
S.
Thrun
, “
Dermatologist-level classification of skin cancer with deep neural networks
,”
Nature
542
(
7639
),
115
118
(
2017
).
2
D. S.
Kermany
,
M.
Goldbaum
,
W.
Cai
,
C. C.
Valentim
,
H.
Liang
,
S. L.
Baxter
,
A.
McKeown
,
G.
Yang
,
X.
Wu
,
F.
Yan
,
J.
Dong
,
M. K.
Prasadha
,
J.
Pei
,
M. Y.
Ting
,
J.
Zhu
,
C.
Li
,
S.
Hewett
,
J.
Dong
,
I.
Ziyar
,
A.
Shi
,
R.
Zhang
,
L.
Zheng
,
R.
Hou
,
W.
Shi
,
X.
Fu
,
Y.
Duan
,
V. A.
Huu
,
C.
Wen
,
E. D.
Zhang
,
C. L.
Zhang
,
O.
Li
,
X.
Wang
,
M. A.
Singer
,
X.
Sun
,
J.
Xu
,
A.
Tafreshi
,
M. A.
Lewis
,
H.
Xia
,
K.
Zhang
, and
P. A.
Keane
, “
Identifying medical diagnoses and treatable diseases by image-based deep learning
,”
Cell
172
(
5
),
1122
1131.e9
(
2018
).
3
D.
Shen
,
G.
Wu
, and
H. I.
Suk
, “
Deep learning in medical image analysis
,”
Annu. Rev. Biomed. Eng.
19
,
221
248
(
2017
).
4
E. J.
Hwang
,
S.
Park
,
K. N.
Jin
,
J. I.
Kim
,
S. Y.
Choi
,
J. H.
Lee
,
J. M.
Goo
,
J.
Aum
,
J. J.
Yim
, and
C. M.
Park
, “
Development and validation of a deep learning-based automatic detection algorithm for active pulmonary tuberculosis on chest radiograph
,”
Clin. Infect. Dis.
69
(
5
),
739
747
(
2019
).
5
I. J.
Goodfellow
,
D.
Erhan
,
P.
Luc Carrier
,
A.
Courville
,
M.
Mirza
,
B.
Hamner
,
W.
Cukierski
,
Y.
Tang
,
D.
Thaler
,
D. H.
Lee
,
Y.
Zhou
,
C.
Ramaiah
,
F.
Feng
,
R.
Li
,
X.
Wang
,
D.
Athanasakis
,
J.
Shawe-Taylor
,
M.
Milakov
,
J.
Park
,
R.
Ionescu
,
M.
Popescu
,
C.
Grozea
,
J.
Bergstra
,
J.
Xie
,
L.
Romaszko
,
B.
Xu
,
Z.
Chuang
,
Y.
Bengio
, and
R.
Fergus
, “
Challenges in representation learning: A report on three machine learning contests
,”
Neural Networks
64
,
59
63
(
2015
).
6
B.
Ehteshami Bejnordi
,
M.
Veta
,
P.
Johannes van Diest
,
B.
van Ginneken
,
N.
Karssemeijer
,
G.
Litjens
,
J. A. W. M.
van der Laak
,
M.
Hermsen
,
Q. F.
Manson
,
M.
Balkenhol
,
O.
Geessink
,
N.
Stathonikos
,
M. C.
van Dijk
,
P.
Bult
,
F.
Beca
,
A. H.
Beck
,
D.
Wang
,
A.
Khosla
,
R.
Gargeya
,
H.
Irshad
,
A.
Zhong
,
Q.
Dou
,
Q.
Li
,
H.
Chen
,
H. J.
Lin
,
P. A.
Heng
,
C.
Haß
,
E.
Bruni
,
Q.
Wong
,
U.
Halici
,
M. Ü.
Öner
,
R.
Cetin-Atalay
,
M.
Berseth
,
V.
Khvatkov
,
A.
Vylegzhanin
,
O.
Kraus
,
M.
Shaban
,
N.
Rajpoot
,
R.
Awan
,
K.
Sirinukunwattana
,
T.
Qaiser
,
Y. W.
Tsang
,
D.
Tellez
,
J.
Annuscheit
,
P.
Hufnagl
,
M.
Valkonen
,
K.
Kartasalo
,
L.
Latonen
,
P.
Ruusuvuori
,
K.
Liimatainen
,
S.
Albarqouni
,
B.
Mungal
,
A.
George
,
S.
Demirci
,
N.
Navab
,
S.
Watanabe
,
S.
Seno
,
Y.
Takenaka
,
H.
Matsuda
,
H.
Ahmady Phoulady
,
V.
Kovalev
,
A.
Kalinovsky
,
V.
Liauchuk
,
G.
Bueno
,
M. M.
Fernandez-Carrobles
,
I.
Serrano
,
O.
Deniz
,
D.
Racoceanu
,
R.
Venâncio
, and
H.
Wang
, “
Diagnostic assessment of deep learning algorithms for detection of lymph node metastases in women with breast cancer
,”
JAMA
318
(
22
),
2199
2210
(
2017
).
7
Y.
Liu
,
K.
Gadepalli
,
M.
Norouzi
,
G. E.
Dahl
,
T.
Kohlberger
,
A.
Boyko
,
S.
Venugopalan
,
A.
Timofeev
,
P. Q.
Nelson
,
G. S.
Corrado
,
J. D.
Hipp
,
L. H.
Peng
, and
M. C.
Stumpe
, “
Detecting cancer metastases on gigapixel pathology images
,”
arXiv:1703.02442
(
2017
).
8
D.
Komura
and
S.
Ishikawa
, “
Machine learning methods for histopathological image analysis
,”
Comput. Struct. Biotechnol. J.
16
,
34
42
(
2018
).
9
S.
Wang
,
D. M. J.
Yang
,
R.
Rong
,
X.
Zeng
,
T.
Ji
,
Z.
Li
, and
G.
Sun
, “
Artificial intelligence in lung cancer pathology image analysis
,”
Cancers
12
(
4
),
823
(
2020
).
10
S.
Arvind
,
K.
Konstantinos
,
S.
Joel
,
K.
Rakshit
,
V. B.
Yuri
,
C. L.
Zachary
, and
S.
Martin
, “
Artificial intelligence-based breast cancer nodal metastasis detection: Insights into the black box for pathologists
,”
Arch. Pathol. Lab. Med.
144
(
12
),
1457
1467
(
2020
).
11
T.
Araújo
and
G.
Aresta
, “
Retinal vessel segmentation using a convolutional neural network
,”
IEEE Trans. Med. Imaging
37
(
11
),
2393
2403
(
2017
).
12
A.
Yala
,
C.
Lehman
,
T.
Schuster
et al
, “
A deep learning mammography-based model for improved breast cancer risk prediction
,”
Radiology
29
(
1
),
60
66
(
2019
).
13
W.
Lotter
,
A. R.
Diab
,
B.
Haslam
et al
, “
Semi-supervised learning for chest X-ray interpretation
,” in
Proceedings of the IEEE International Conference on Computer Vision (ICCV)
(
IEEE
,
2019
), pp.
7023
7032
.
14
J.
Zhang
,
S. K.
Zhou
,
C. X.
Li
et al
, “
Soft tissue sarcoma classification using convolutional neural network
,”
J. Digital Imaging
30
(
4
),
487
496
(
2017
).
15
C. M.
Lo
,
J.
Hipp
,
A. C.
Jirapatnakul
et al
, “
Artificial intelligence for breast cancer detection in mammography and digital breast tomosynthesis: State of the art
,”
J. Digital Imaging
32
(
1
),
38
49
(
2019
).
16
D.
Ribli
,
A.
Horváth
,
Z.
Unger
et al
, “
Detecting and classifying lesions in mammograms with deep learning
,”
PLoS One
13
(
10
),
e0197542
(
2018
).
17
A. A. A.
Setio
,
A.
Traverso
,
T.
de Bel
,
M. S.
Berens
,
C. v. d.
Bogaard
,
P.
Cerello
,
H.
Chen
,
Q.
Dou
,
M. E.
Fantacci
,
B.
Geurts
,
R. v. d.
Gugten
,
P. A.
Heng
,
B.
Jansen
,
M. M.
de Kaste
,
V.
Kotov
,
J. Y. H.
Lin
,
J. T.
Manders
,
A.
Sóñora-Mengana
,
J. C.
García-Naranjo
,
E.
Papavasileiou
,
M.
Prokop
,
M.
Saletta
,
C. M.
Schaefer-Prokop
,
E. T.
Scholten
,
L.
Scholten
,
M. M.
Snoeren
,
E. L.
Torres
,
J.
Vandemeulebroucke
,
N.
Walasek
,
G. C.
Zuidhof
,
B. v.
Ginneken
,
C.
Jacobs
, and
B.
van Ginneken
, “
Validation, comparison, and combination of algorithms for automatic detection of pulmonary nodules in computed tomography images: The LUNA16 challenge
,”
Med. Image Anal.
42
,
1
13
(
2017
).
18
H.
Jiang
,
H.
Ma
,
W.
Qian
et al
, “
An automatic detection system of lung nodule based on multigroup patch-based deep learning network
,”
IEEE J. Biomed. Health Inf.
22
(
4
),
1227
1237
(
2018
).
19
M. H.
Yap
,
S.
Peker
, and
D. L.
Rubin
, “
Using convolutional neural networks to identify cancer in CT scans of lungs
,” arXiv:1803.04765 (
2018
).
20
W.
Shen
,
M.
Zhou
,
F.
Yang
,
C.
Yang
,
J.
Tian
, and
K.
Wang
, “
Multi-scale convolutional neural networks for lung nodule classification
,”
Inf. Process Med. Imaging
24
,
588
599
(
2015
).
21
F.
Ciompi
,
K.
Chung
,
S. J.
van Riel
,
A. A. A.
Setio
,
P. K.
Gerke
,
C.
Jacobs
,
E. T.
Scholten
,
C.
Schaefer-Prokop
,
M. M. W.
Wille
,
A.
Marchianò
,
U.
Pastorino
,
M.
Prokop
,
B.
van Ginneken
, and
E. T.
Scholten
, “
Towards automatic pulmonary nodule management in lung cancer screening with deep learning
,”
Sci. Rep.
7
(
1
),
46479
(
2017
).
22
S.
Christodoulidis
,
M.
Anthimopoulos
,
L.
Ebner
,
A.
Christe
, and
S.
Mougiakakou
, “
Multisource transfer learning with convolutional neural networks for lung pattern analysis
,”
IEEE J. Biomed. Health Inf.
21
(
1
),
76
84
(
2017
).
23
Q.
Dou
,
H.
Chen
,
L.
Yu
,
J.
Qin
, and
P. A.
Heng
, “
Multilevel contextual 3D CNNs for false positive reduction in pulmonary nodule detection
,”
IEEE Trans. Biomed. Eng.
63
(
12
),
2613
2623
(
2016
).
24
X.
Zhang
,
Y.
Xie
,
Y.
Xia
,
Q.
Li
, and
X.
Zhou
, “
Detecting peripherally located pulmonary nodules in CT scans by one deep convolutional neural network and one SVM classifier
,” in
2017 IEEE International Conference on Bioinformatics and Biomedicine (BIBM)
(
IEEE
,
2017
), pp.
38
42
.
25
C.
Ding
,
Q.
Li
, and
Y.
Mao
, “
Detection of pulmonary nodules in CT images based on template matching and deep learning
,”
Neurocomputing
260
,
98
104
(
2017
).
26
D.
Ardila
,
A. P.
Kiraly
,
S.
Bharadwaj
,
B.
Choi
,
J. J.
Reicher
,
L.
Peng
,
D.
Tse
,
M.
Etemadi
,
W.
Ye
,
G.
Corrado
,
D. P.
Naidich
, and
S.
Shetty
, “
End-to-end lung cancer screening with three-dimensional deep learning on low-dose chest computed tomography
,”
Nat. Med.
25
(
6
),
954
961
(
2019
).
27
M.
Liang
,
W.
Tang
,
D.
Xu
, and
J.
Huang
, “
Lung nodule detection using 3D deep convolutional neural networks
,”
IEEE Trans. Med. Imaging
38
(
2
),
469
479
(
2019
).
28
W.
Shen
,
M.
Zhou
,
F.
Yang
,
C.
Yang
,
J.
Tian
, and
K.
Wang
, “
Learning from experts: Developing transferable deep features for patient-level lung cancer prediction
,” in
International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI)
(
Springer
,
2016
), pp.
124
132
.
29
Y.
Zhang
,
Q.
Li
, and
S.
Wang
, “
Detection of small lung cancers in computed tomography scans using a deep convolutional neural network ensemble
,”
Neurocomputing
282
,
243
251
(
2018
).
30
M.
Anthimopoulos
,
S.
Christodoulidis
,
L.
Ebner
,
A.
Christe
, and
S.
Mougiakakou
, “
Lung pattern classification for interstitial lung diseases using a deep convolutional neural network
,”
IEEE Trans. Med. Imaging
35
(
5
),
1207
1216
(
2016
).
31
J.
De Fauw
,
J. R.
Ledsam
,
B.
Romera-Paredes
,
S.
Nikolov
,
N.
Tomasev
,
S.
Blackwell
,
H.
Askham
,
X.
Glorot
,
B.
O’Donoghue
,
D.
Visentin
,
G.
van den Driessche
,
B.
Lakshminarayanan
,
C.
Meyer
,
F.
Mackinder
,
S.
Bouton
,
K.
Ayoub
,
R.
Chopra
,
D.
King
,
A.
Karthikesalingam
,
C. O.
Hughes
,
R.
Raine
,
J.
Hughes
,
D. A.
Sim
,
C.
Egan
,
A.
Tufail
,
H.
Montgomery
,
D.
Hassabis
,
G.
Rees
,
T.
Back
,
P. T.
Khaw
,
M.
Suleyman
,
J.
Cornebise
,
P. A.
Keane
, and
O.
Ronneberger
, “
Clinically applicable deep learning for diagnosis and referral in retinal disease
,”
Nat. Med.
24
(
9
),
1342
1350
(
2018
).
32
I. M.
Baltruschat
,
N.
Wiedemann
,
N.
Heller
,
R.
Groh
, and
T.
Twellmann
,
Detection and classification of pulmonary nodules using advanced deep learning
,”
Eur. Radiol.
30
(
6
),
3663
3673
(
2020
).
33
W.
Shen
,
M.
Zhou
,
F.
Yang
,
D.
Yu
,
D.
Dong
et al
, “
Multi-crop convolutional neural networks for lung nodule malignancy suspiciousness classification
,”
Pattern Recognit.
61
,
663
673
(
2017
).
34
L.
Bi
,
J.
Kim
,
A.
Kumar
,
M.
Fulham
,
D.
Feng
, and
J.
Kim
, “
Cancer lesion detection and segmentation in PET/CT images using texture features, deep learning, and support vector machine
,”
Neurocomputing
330
,
14
22
(
2019
).
35
W.
Sun
,
B.
Zheng
,
W.
Qian
, and
H.
Zhang
, “
Lung nodule classification with multi-level context convolutional network
,”
J. Digital Imaging
31
(
6
),
851
860
(
2018
).
36
P.
Tschandl
,
C.
Rosendahl
, and
H.
Kittler
, “
The HAM10000 dataset, a large collection of multi-source dermatoscopic images of common pigmented skin lesions
,”
Sci. Data
5
,
180161
(
2018
).
37
T. J.
Brinker
,
A.
Hekler
,
A. H.
Enk
,
J.
Klode
,
A.
Hauschild
,
C.
Berking
, and
C.
von Kalle
, “
Deep learning outperformed 11 pathologists in the classification of histopathological melanoma images
,”
Eur. J. Cancer
115
,
79
86
(
2019
).
38
H. A.
Haenssle
,
C.
Fink
,
R.
Schneiderbauer
,
F.
Toberer
,
T.
Buhl
,
A.
Blum
,
A.
Kalloo
,
A. B. H.
Hassen
,
L.
Thomas
,
A.
Enk
,
L.
Uhlmann
,
C.
Alt
,
M.
Arenbergerova
,
R.
Bakos
,
A.
Baltzer
,
I.
Bertlich
,
A.
Blum
,
T.
Bokor-Billmann
,
J.
Bowling
,
N.
Braghiroli
,
R.
Braun
,
K.
Buder-Bakhaya
,
T.
Buhl
,
H.
Cabo
,
L.
Cabrijan
,
N.
Cevic
,
A.
Classen
,
D.
Deltgen
,
C.
Fink
,
I.
Georgieva
,
L. E.
Hakim-Meibodi
,
S.
Hanner
,
F.
Hartmann
,
J.
Hartmann
,
G.
Haus
,
E.
Hoxha
,
R.
Karls
,
H.
Koga
,
J.
Kreusch
,
A.
Lallas
,
P.
Majenka
,
A.
Marghoob
,
C.
Massone
,
L.
Mekokishvili
,
D.
Mestel
,
V.
Meyer
,
A.
Neuberger
,
K.
Nielsen
,
M.
Oliviero
,
R.
Pampena
,
J.
Paoli
,
E.
Pawlik
,
B.
Rao
,
A.
Rendon
,
T.
Russo
,
A.
Sadek
,
K.
Samhaber
,
R.
Schneiderbauer
,
A.
Schweizer
,
F.
Toberer
,
L.
Trennheuser
,
L.
Vlahova
,
A.
Wald
,
J.
Winkler
,
P.
Wölbing
,
I.
Zalaudek
, and
R.
Hofmann-Wellenhof
, “
Man against machine: Diagnostic performance of a deep learning convolutional neural network for dermoscopic melanoma recognition in comparison to 58 dermatologists
,”
Ann. Oncol.
29
(
8
),
1836
1842
(
2018
).
39
N. C.
Codella
,
D.
Gutman
,
M. E.
Celebi
,
B.
Helba
,
M. A.
Marchetti
,
S. W.
Dusza
, and
A.
Halpern
, “
Skin lesion analysis toward melanoma detection: A challenge at the 2017 international symposium on biomedical imaging (ISBI), hosted by the international skin imaging collaboration (ISIC)
,” arXiv:1710.05006 (
2018
).
40
K.
Matsunaga
and
A.
Hamada
, “
How to suppress saturation of melanin in dermoscopy images? A proposal based on three-color signals
,”
Skin Res. Technol.
22
(
3
),
380
387
(
2016
).
41
M.
Wiltgen
,
A.
Nobile
,
F.
Ambrogi
and
C.
Fink
, “
Real-time in-vivo skin cancer diagnosis with convolutional neural networks
,” in
Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops
(
IEEE
,
2018
), pp.
57
64
.
42
J.
Kawahara
,
A.
BenTaieb
,
G.
Hamarneh
, and
R.
Dror
, “
Deep features to classify skin lesions
,” in
International Workshop on Machine Learning in Medical Imaging
(
Springer
,
Cham
,
2016
), pp.
133
141
.
43
J.
Schmidhuber
, “
Deep learning in neural networks: An overview
,”
Neural Networks
61
,
85
117
(
2015
).
44
M. E.
Celebi
,
Q.
Wen
,
H.
Iyatomi
,
K.
Shimizu
,
H.
Zhou
,
G.
Schaefer
, and
W. V.
Stoecker
, “
A state-of-the-art survey on lesion border detection in dermoscopy images
,”
Skin Res. Technol.
15
(
3
),
307
317
(
2009
).
45
L.
Bi
,
J.
Kim
,
E.
Ahn
,
D.
Feng
, and
M.
Fulham
, “
Skin lesion segmentation using deep convolutional neural network and patch-wise contour refinement
,” in
2018 IEEE 15th International Symposium on Biomedical Imaging
(
ISBI
,
2018
), pp.
912
916
.
46
F.
Liu
and
P. H. C.
Chen
, “
Fast deep learning-based skin cancer detection for mobile applications
,”
Comput. Biol. Med.
110
,
48
60
(
2019
).
47
H. A.
Haenssle
et al
, “
Man against machine reloaded: Performance of a market-approved convolutional neural network in classifying a broad spectrum of skin lesions in comparison with 96 dermatologists
,”
Ann. Oncol.
31
(
7
),
1007
1014
(
2020
).
48
A.
Menegola
et al
, “
Transfer learning for the classification of skin lesions
,”
J. Dermatol. Sci.
88
(
2
),
210
215
(
2017
).
49
T. J.
Brinker
et al
, “
Comparison between deep learning–based artificial intelligence algorithms and 11 pathologists in the classification of melanocytic neoplasms
,”
JAMA Dermatol.
156
(
3
),
292
300
(
2020
).
50
A.
Esteva
et al
, “
A guide to deep learning in healthcare
,”
Nat. Med.
25
(
1
),
24
29
(
2019
).
51
J.
Yap
et al
, “
Automated keratinocyte detection for skin cancer images using patch-based deep learning
,”
Comput. Biol. Med.
113
,
103396
(
2019
).
52
N. C.
Codella
et al
, “
Skin lesion analysis towards melanoma detection using deep learning: ISIC 2018 challenge
,” in
International Workshop on Skin Image Analysis
(
Springer
,
Cham
,
2019
), pp.
98
110
.
53
P.
Tschandl
et al
, “
Expert-level diagnosis of nonpigmented skin cancer by combined convolutional neural networks
,”
JAMA Dermatol.
155
(
1
),
58
65
(
2019
).