As nanoparticles are being put to practical use as useful materials in the medical, pharmaceutical, and industrial fields, the importance of technologies that can evaluate not only nanoparticle populations of homogeneous size and density but also those of rich diversity is increasing. Nano-tracking analysis (NTA) has been commercialized and widely used as a method to measure individual nanoparticles in liquids and evaluate their size distribution by analyzing Brownian motion. We have combined deep learning (DL) for NTA to extract more property information and explored a methodology to achieve an evaluation for individual particles to understand their diversity. Practical NTA always assumes spherical shape when quantifying particle size using the Stokes–Einstein equation, but it is not possible to verify whether the measured particles are truly spherical. We developed a DL model that predicts the shape of nanoparticles using time series trajectory data of BM obtained from NTA measurements to address this problem. As a result, we were able to discriminate with ∼80% accuracy between spherical and rod-shaped gold nanoparticles of different shapes, which are evaluated to have nearly equal particle size without any discrimination by conventional NTA. Furthermore, we demonstrated that the mixing ratio of spherical and rod-shaped nanoparticles can be quantitatively estimated from measured data of mixed samples of nanoparticles. This result suggests that it is possible to evaluate particle shape by applying DL analysis to NTA measurements, which was previously considered impossible, and opens the way to further value-added NTA.

To evaluate the properties of heterogeneous nanoparticles with mixed particle types and sizes, it is necessary to measure and analyze multiple physical properties for each identified particle. For example, as new therapeutic and diagnostic technologies using extracellular vesicles and artificial nanoparticles attract attention, evaluation techniques for nanoparticles and nanocolloids that support safety and reliability are becoming increasingly important, but there are many challenges and limitations to measurement in the 10 to 100 nm range.1,2

Nano-tracking analysis (NTA),3,4 which captures the Brownian motion (BM) through scattered light imaging and obtains the particle size from each trajectory, is widely used as a technique for measuring micro- to nano-sized single-particles in liquids due to its simplicity and the time and cost advantages of the measurement. However, the properties that can be extracted with NTA are generally limited, such as the inability to directly see the shape of the particles. It is well known that nanoparticle shape strongly influences optical properties, such as plasmon resonance5–7 and Raman enhancement,8 but it is also an important factor indicating the degree of aggregation,9,10 chemical properties such as solubility,11 drug delivery performance,12 and material toxicity.13–15 We, therefore, tried to extract shape information from NTA data without significantly changing the experimental method.

It is also known that BM is affected by the shape of nanoparticles by previous studies on micrometer-sized particles16 and stochastic simulations17 in the long history of BM research from basic physics and analytical methods to the applied technology.16–28 Therefore, in principle, it is expected that particle shape information can be detected by analyzing the trajectory of BM (time-series coordinate data). However, it is practically difficult to detect shape anisotropy of nanoparticles using NTA measurement data because nanoparticles have relaxation times that are many orders of magnitude shorter than micrometer-sized particles (theoretical rotational relaxation time is nine orders of magnitude smaller from the following definition: τr = πηd3/2kBT), and the actual measured data contain experimental errors.

The analysis method also has points to be discussed. Although Stokes’ law is applied to the analysis of nanoparticles by assuming a rigid sphere, actual nanoparticles are often not spherical. In the case of non-spherical particles, a correction factor called the dynamic shape factor18 is introduced to consider translational BM, and the volume equivalent diameter was used as a solution to apply Stokes’ law. Both theoretical and experimental studies have been conducted to obtain the dynamic shape factor.19–22 For rotational BM, the similar analysis can be performed as for spherical particles by including a factor of shape in the rotational friction coefficient to represent the rotational diffusion coefficient.27–29 This means that if the optical measurement results for non-spherical particles are analyzed using the “spherical” Stokes–Einstein equation, the analytical values will include errors due to the shape factor not applied. Since the NTA method does not allow for direct observation of particle shape, it is impossible to determine the shape of each particle and then apply an appropriate analytical formula.

A promising solution is to apply deep learning (DL) to the analysis.30,31 It is superior in extracting hidden correlations in large datasets, even when the data acquired are averaged, and is used to analyze the diffusion characteristics of nanoparticles, such as the Hurst exponent of fractional BM and the diffusion coefficient of BM,32 classification of diffusion modes,33 and classification of particle trajectories in heterogeneous media from a number of diffusive states.34 Deep learning analysis may also be able to extract shape information from statistically similar BM trajectory patterns, as in the example of classifying particles by shape based on their scattered light patterns.35 When extracting shape information from the trajectory of BM, there is no explicit formula that directly links the two. In previous studies of BM of non-spherical particles using DL, shape is usually given as a known factor, and no previous studies have identified the particle shape from the BM trajectory measured by NTA. Therefore, in this study, we have developed a model that can estimate shape classification from time-series trajectory data of BM using DL.

Development of the shape estimation model consisted of three stages: NTA measurements for raw data acquisition, creation of datasets and models for DL, and DL.

Brownian motion trajectories of spherical gold nanoparticles (AuNP, Citrate capping, 80 nm diameter, Sigma-Aldrich) and gold nanorods (AuNR, Citrate capping, 40 nm diameter × 180 nm length, Nanopartz) were measured by NTA. A microcapillary chip-based particle analysis system36 based on the NTA method developed for characterizing nanoparticles in liquid, which can simultaneously measure particle size and zeta potential, is shown in Fig. 1(a). The nanoparticles in the microfluidic channels made of cyclo-olefin polymers [Fig. 1(b)] are irradiated with a sheet laser (405 nm, 100 mW) from the side of the channel, and the scattered light is focused by an objective lens below the channel and formed by a high-sensitivity camera to obtain a high-contrast dark field image every 10 ms (100 fps) for 100 frames [Fig. 1(c)]. The center of gravity of the scattered light from each nanoparticle in consecutive 100-frame images was tracked [Fig. 2(a)] to obtain a trajectory [Fig. 2(b)].

FIG. 1.

(a) The configuration of the NTA equipment. (b) COP microfluidic chip. (c) Typical dark-field image.

FIG. 1.

(a) The configuration of the NTA equipment. (b) COP microfluidic chip. (c) Typical dark-field image.

Close modal
FIG. 2.

(a) Tracking of each AuNP (left) and AuNR (right) by NTA. The numbers next to particles are identification numbers. The colored lines extending from particles represent trajectories of Brownian motion. (b) Typical Brownian motion trajectory of an AuNP (left) and an AuNR (right). (c) An example dataset for deep learning (time-series trajectory + correct labels).

FIG. 2.

(a) Tracking of each AuNP (left) and AuNR (right) by NTA. The numbers next to particles are identification numbers. The colored lines extending from particles represent trajectories of Brownian motion. (b) Typical Brownian motion trajectory of an AuNP (left) and an AuNR (right). (c) An example dataset for deep learning (time-series trajectory + correct labels).

Close modal

The difference between adjacent frames of the time-series 2D coordinate data was converted into a displacement vector. Statistics for each BM trajectory in the displacement vector is shown in Table I. The time series trajectory data were given a correct answer label (AuNP: 0, AuNR: 1) indicating whether the trajectory related to spherical or rod-shaped particles and was used as the DL dataset [Fig. 2(c)]. In evaluating all DL models, the dataset was randomly divided into training data (80%), validation data (10%), and test data (10%). For shape prediction, the dataset was randomly divided into training data (90%) and validation data (10%). As the test data, another dataset was created and loaded.

TABLE I.

Statistics for each Brownian motion trajectory in the displacement vector.

Average (μm)SD (μm)SkewnessKurtosis
AuNP (d = 80 nm) x component 0.002 23 0.275 −0.0046 0.1067 
y component −0.005 74 0.269 −0.0020 0.1687 
AuNR (d = 40 nm, L = 180 nm) x component 0.002 06 0.264 −0.0038 0.1289 
y component −0.006 32 0.246 0.0024 0.1051 
Average (μm)SD (μm)SkewnessKurtosis
AuNP (d = 80 nm) x component 0.002 23 0.275 −0.0046 0.1067 
y component −0.005 74 0.269 −0.0020 0.1687 
AuNR (d = 40 nm, L = 180 nm) x component 0.002 06 0.264 −0.0038 0.1289 
y component −0.006 32 0.246 0.0024 0.1051 

Here, we designed and built DL models using Keras based on three orthodox techniques: multilayer perceptron (MLP), long short-term memory (LSTM),37 and one-dimensional convolutional neural network (1D CNN).38 MLP is a basic Fully Connected Neural Network (FCNN)39 and was used to compare the performance of the models. LSTM is a derivative model of Recurrent Neural Networks (RNNs)40,41 and can learn features of long-term time series data by introducing memory cells (LSTM block) into RNN and widely used for analysis and prediction of continuous data.42 Three gates (input, forgetting, and output) work together to learn how information is put in and out of the cell, control the balance between long-term memory and short-term memory, and retain information for the long term.43 1D CNN performs convolution processing in the time axis direction, extracts local features in the convolution layer, and summarizes them in the pooling layer.44 Compared to FCNN, it is possible to recognize spatial features with high accuracy with fewer parameters. Furthermore, in order to consider both the time series and the correlation with the surroundings, we designed and constructed a 1D CNN+Bi-LSTM composite model that can extract local features by one-dimensional convolution and accumulate trends in a bidirectional LSTM structure.

The structure of our 1D CNN+Bi-LSTM model and a typical flowchart when performing shape prediction is shown in Figs. 3(a) and 3(b), respectively. Input data are the array with 100 × 1 elements. A filter with 7 × 1 elements is employed in first two convolution layers and that with 5 × 1 elements for the third convolution layer. The number of channels of all filters is set to be 16. Bidirectional LSTM blocks with five layers are installed instead of pooling layers. As the activation functions, we used ReLU for 1D conversion, hyperbolic tangent for the LSTM block, and softmax for the output layer. Binary cross-entropy loss is employed for the loss function, and adaptive moment estimation (ADAM)45 is employed for the optimization of the loss function. The dropout46 with the rate of 0.5 is employed in the affine layer to avoid the overfitting. The architecture and flowchart of MLP, LSTM, and 1D CNN models we used in this study are shown in  Appendix A. To implement the deep learning models, we used Python 3.6.9 as the programming language and implemented the shape estimation model using the deep learning libraries TensorFlow and Keras 2.2.0.

FIG. 3.

(a) Structure of the 1D CNN+Bi-LSTM deep learning model. (b) A typical flowchart of the 1D CNN+Bi-LSTM model performed in this study. Accuracy corresponded to the flowchart was 0.833 at epoch 50 with a batch size 32 and a threshold of 0.5.

FIG. 3.

(a) Structure of the 1D CNN+Bi-LSTM deep learning model. (b) A typical flowchart of the 1D CNN+Bi-LSTM model performed in this study. Accuracy corresponded to the flowchart was 0.833 at epoch 50 with a batch size 32 and a threshold of 0.5.

Close modal
In the process of developing the DL models, experimental data that were considered to contain errors were cut out from the DL dataset using particle diameter as an indicator to improve the accuracy of DL classification. The particle size distribution obtained by analyzing the NTA method with the general Stokes–Einstein law for rigid spheres is wider than the true distribution due to methodological error (uncertainty in determining the diffusion coefficient from the two-dimensional mean-square displacement47,48), digital error (spatial and time resolution issues), and errors due to improper parameterization, such as the non-use of dynamic shape factor in the analysis.4 The theoretical value of hydrodynamic diameter dh for the rod-shaped particles we used was calculated to be 93 nm using the following equations:29 
Dt=kBT3πηLlnLd+v,
v=0.312+0.565dL0.1dL2,
dh=kBT3πηDt(Stokes  Einstein formula),

where kB is the Boltzmann constant (J/K), T is the temperature (K), η is the viscosity (Pa s), L is the rod length (m), and d is the rod diameter (m).

The median particle size determined by NTA measurements was 113.3 nm for AuNP and 123.5 nm for AuNR in this experiment [Figs. 4(a) and 4(b)]. The hydrodynamic diameter dh of particles less than 100 nm in diameter has been measured to be about 5%–10% larger,49 but a larger difference was found in our case. Considering above factors, trajectories of 5000 AuNP and 5000 AuNR particles of 95–125 nm and 110–140 nm in diameter, respectively, were used in datasets for model development.

FIG. 4.

Particle size distribution obtained from Brownian motion measurements by NTA using the microcapillary chip-based particle analysis system. (a) AuNP (n = 9544, dave = 123.4 ± 46.6 nm, median = 113.3 nm). (b) AuNR (n = 9558, dave = 124.1 ± 29.7 nm, median = 123.5 nm).

FIG. 4.

Particle size distribution obtained from Brownian motion measurements by NTA using the microcapillary chip-based particle analysis system. (a) AuNP (n = 9544, dave = 123.4 ± 46.6 nm, median = 113.3 nm). (b) AuNR (n = 9558, dave = 124.1 ± 29.7 nm, median = 123.5 nm).

Close modal

Using time series data with varying trajectory lengths (20, 40, 60, 80, and 100 frames), we verified the convergence of learning by changing the hyperparameters for each number of frames for the four models (MLP, LSTM, 1D CNN, and 1D CNN+Bi-LSTM). Typical training/validation loss and accuracy curves at 100 frames for each model are shown in  Appendix B. The conditions showing good evaluation indices stably were used as the optimal hyperparameters for the following calculations. The shape classification accuracy of the four models was evaluated at their respective optimal hyperparameters [Fig. 5(a)]. The accuracy increased with trajectory length for all models, indicating that data length is an important factor in revealing features depending on the shape. The area under the curves (AUCs) of receiver operating characteristic (ROC) curves for four models [Fig. 5(b)] and the corresponding ROC curves [Fig. 5(c)] showed similar trends as accuracy. Here, the true positive rate represents the proportion of AuNR (rod shaped particles) correctly determined as AuNR (non-spherical), and false positive rate represents the proportion of AuNP (sphere particles) falsely determined as AuNR (non-spherical). The high accuracy above 80% at 100 frames for the LSTM and 1D CNN model [Fig. 5(a)] indicates that both the extraction of local features by convolution and accumulation of temporal dynamics are effective for extracting shape characteristics. At the same time, the high accuracy indicates that shape classification of nanoparticles in liquids has reached a realistic level for single-particle analysis using NTA with DL.

FIG. 5.

Relationship between evaluation indices of shape classification and number of frames for each deep learning model. (a) Accuracy (average of four runs), (b) AUC (average of four runs), and (c) typical ROC curve at each number of consecutive frames. The size-selected training/evaluation dataset was used for model evaluation. Hyper-parameters: (MLP) batch size 1024, eight steps, epoch 30; (LSTM) batch size 512, 16 steps, epoch 200; (1D CNN) batch size 1024, eight steps, epochs 50 (40–100 frames) and 30 (20 frames); and (1D CNN+Bi-LSTM) batch size 32, 250 steps, epochs 10 (20, 40 frames), 15 (60 frames), 30 (80 frames), and 50 (100 frames).

FIG. 5.

Relationship between evaluation indices of shape classification and number of frames for each deep learning model. (a) Accuracy (average of four runs), (b) AUC (average of four runs), and (c) typical ROC curve at each number of consecutive frames. The size-selected training/evaluation dataset was used for model evaluation. Hyper-parameters: (MLP) batch size 1024, eight steps, epoch 30; (LSTM) batch size 512, 16 steps, epoch 200; (1D CNN) batch size 1024, eight steps, epochs 50 (40–100 frames) and 30 (20 frames); and (1D CNN+Bi-LSTM) batch size 32, 250 steps, epochs 10 (20, 40 frames), 15 (60 frames), 30 (80 frames), and 50 (100 frames).

Close modal

As seen above, the 1D CNN+Bi-LSTM model achieved accuracy, AUC, and ROC comparable to those of the 1D CNN and LSTM models, suggesting that it is possible to combine the characteristics of both models in one model. It means that local features can be extracted while capturing features of temporal dynamics. In actual BM, multiple factors, such as the particle size, shape, and surface condition, are involved, and it is considered that a DL model specialized for a single feature value cannot fully analyze the actual phenomenon. Therefore, rather than selecting a model only based on the high accuracy rate, we adopted the 1D CNN+Bi-LSTM model, which can consider multiple physical effects with the same accuracy as the conventional architectures, because it is more suitable for shape prediction.

Mixed suspensions of AuNP and AuNR in different mixing ratios (7:3, 5:5, 3:7 by number) were prepared for NTA. Figures 6(a)6(c) show the particle size distribution of mixtures. From the trajectory data of each mixture, a test dataset (total 500 time-series displacement vectors) was made and labeled spheres. The training/validation dataset was created from measurements of 9000 AuNP and 9000 AuNR particles without removing data containing errors that appear as size distribution, unlike model evaluation described above. It is for including equivalent error in the training/validation data and test data, as it is difficult to get a correct error reading (i.e., determine the optimal size range) for the mixed samples that serve as the test samples. To check the effect of experimental errors, such as agglomeration, we also generated and verified a pseudosample from unmixed AuNP and AuNR data according to the experimental mixing ratio [Fig. 6(d)]. Using the data of the mixed sample and the pseudosample at each mixture ratio as the test sample, shape prediction was performed with a 1D CNN+Bi-LSTM model to obtain the content of particles predicted to be non-spherical [Fig. 6(e)]. The difference between the value of predicted content (average of eight runs) for the mixed sample and generated pseudosample was from 0.2% to 5.6%. The two regression lines are in good agreement, and the difference between the values of predicted content and that calculated from the regression line is within 2.6%. This indicates that the regression line is valid as a calibration curve and allows for the estimation of the mixing ratio from the predicted content. For example, in the present case, a mixed solution predicted to have 50% non-spherical content would be considered to contain ∼43% non-spherical particles. We have shown that the shape of nanoparticles, for which information cannot be obtained only by ordinary NTA, can be distinguished when DL is used for analysis.

FIG. 6.

Particle size distribution of mixtures obtained from Brownian motion measurements by the NTA method using the microcapillary chip-based particle analysis system. (a) AuNP/AuNR mixed sample containing 30% AuNR (n = 701, dave = 114.0 ± 24.4 nm). (b) AuNP/AuNR mixed sample containing 50% AuNR (n = 802, dave = 121.1 ± 27.1 nm). (c) AuNP/AuNR mixed sample containing 70% AuNR (n = 768, dave = 122.4 ± 30.6 nm). (d) Schematic of the mixed sample and pseudosample containing 70% AuNR. (e) Comparison of predicted content of non-spherical particles (average of eight runs) in the mixed sample and generated pseudosample at various mixing ratios.

FIG. 6.

Particle size distribution of mixtures obtained from Brownian motion measurements by the NTA method using the microcapillary chip-based particle analysis system. (a) AuNP/AuNR mixed sample containing 30% AuNR (n = 701, dave = 114.0 ± 24.4 nm). (b) AuNP/AuNR mixed sample containing 50% AuNR (n = 802, dave = 121.1 ± 27.1 nm). (c) AuNP/AuNR mixed sample containing 70% AuNR (n = 768, dave = 122.4 ± 30.6 nm). (d) Schematic of the mixed sample and pseudosample containing 70% AuNR. (e) Comparison of predicted content of non-spherical particles (average of eight runs) in the mixed sample and generated pseudosample at various mixing ratios.

Close modal

With the DL method, particles with different shapes, even if they have the same hydrated diameter, can be distinguished from a mixture based on their trajectories. In comparison, conventional NTA methods cannot distinguish between particle shapes, and nanoparticles of different shapes in a mixture cannot be distinguished unless there is a large difference in hydrated particle size or scattered light intensity. This method has the potential to open a new approach to the fundamental study of BM of nanoparticles in liquids.

We attempted to determine the shapes of two types of particles in this paper, but considering the types of shapes of commercially available nanoparticles, we think that this method can be used in practical applications, such as the detection of foreign substances in homogeneous systems. It would be interesting to extend the measurement object to particles of various shapes and materials, and it is a future research topic to examine the applicability of the DL+NTA method.

Here, we demonstrated the effectiveness of a new nanoparticle characterization method that combines DL and NTA measurements. For applying DL to general NTA data, we constructed a 1D CNN+Bi-LSTM model that predicts the shape of single nanoparticles in liquid with an accuracy of more than 80%. Using this model, the mixing ratio of the AuNP and AuNR mixed sample can be determined with a very small error: the difference between the values of predicted content and that calculated from the regression line is within 2.6%. We demonstrated that the shape of nanoparticles could be distinguished in a homogeneously dispersed system with little difference in particle size. It was shown that the shape of nanoparticles, which cannot be obtained by conventional NTA alone, can be identified by applying DL to BM trajectory analysis.

The datasets created for this paper are available within the supplementary material.

The authors are grateful for the technical assistance given by Ms. H. Kishita. This research was financially supported by the Japan Science and Technology Agency (JST) through the open innovation platform for industry-academia co-creation (COI-NEXT) program (Grant No. JPMJPF2202).

The authors have no conflicts to disclose.

Hiroaki Fukuda and Hiromi Kuramochi contributed equally to this work.

Hiroaki Fukuda: Formal analysis (supporting); Investigation (equal); Methodology (equal); Software (lead); Writing – original draft (supporting); Writing – review & editing (equal). Hiromi Kuramochi: Conceptualization (equal); Data curation (lead); Formal analysis (lead); Investigation (equal); Methodology (equal); Visualization (lead); Writing – original draft (lead); Writing – review & editing (equal). Yasushi Shibuta: Methodology (equal); Supervision (supporting); Writing – review & editing (equal). Takanori Ichiki: Conceptualization (equal); Funding acquisition (lead); Supervision (lead); Writing – review & editing (equal).

The data that support the findings of this study are available within the article and its supplementary material.

The architecture of MLP, LSTM, and 1D CNN models is as follows: Typical flowcharts are shown in Fig. 7.

FIG. 7.

Typical flowcharts of (a) MLP, (b) LSTM, and (c) 1D CNN model performed in this study. Accuracies corresponded to the flowchart were 0.704, 0.816, and 0.807, respectively.

FIG. 7.

Typical flowcharts of (a) MLP, (b) LSTM, and (c) 1D CNN model performed in this study. Accuracies corresponded to the flowchart were 0.704, 0.816, and 0.807, respectively.

Close modal

MLP: Input data of MLP are the array with 200 × 1 elements. The 2D output was obtained from a quadruple stack of 1000 layers, 500 layers, and two 100 layers. We used ReLU for activation in all layers except output layer (softmax). The dropout with the rate of 0.5 is employed in all stages. Binary cross-entropy loss is employed for the loss function.

LSTM: Input data of LSTM are the array with 200 × 1 elements. The 2D output was obtained with a structure in which ten layers of LSTM are stacked in three stages. A dropout rate of 0.05 is adopted for all layers. We used hyperbolic tangent as an activation function, sigmoid for recurrent activation, and softmax for the output layer. Binary cross-entropy loss is employed for the loss function.

1D CNN: Input data of 1D CNN are the array with 100 × 1 elements. A filter with 7 × 1 elements is employed in first two convolution layers and that with 5 × 1 elements for the third convolution layer. The number of channels of all filters is set to be 16. Max pooling with a pool size of five is employed. As the activation functions, we used ReLU for 1D conversion and softmax for the output layer. Binary cross-entropy loss is employed for the loss function, and ADAM is employed for the optimization of loss function. The dropout with the rate of 0.5 is employed in the affine layer to avoid the overfitting.

As shown in Fig. 8, for three models (LSTM, 1D CNN, and 1D CNN+Bi-LSTM), the gap between the training and validation curves of both accuracy and loss is small; the training was performed properly and converged within the implemented epochs. MLP has a large gap, and accuracy and loss are unstable even if epochs are increased.

FIG. 8.

Typical accuracy and loss curves for the training dataset and validation dataset for each deep learning model. The size-selected training/evaluation dataset was used for model evaluation (100 frames), of which 80% was training data and 10% was validation data. Hyperparameters: (MLP) batch size 1024, eight steps, epoch 30; (LSTM) batch size 512, 16 steps, epoch 200; (1D CNN) batch size 1024, eight steps, epoch 50; and (1D CNN+Bi-LSTM) batch size 32, 250 steps, epoch 50.

FIG. 8.

Typical accuracy and loss curves for the training dataset and validation dataset for each deep learning model. The size-selected training/evaluation dataset was used for model evaluation (100 frames), of which 80% was training data and 10% was validation data. Hyperparameters: (MLP) batch size 1024, eight steps, epoch 30; (LSTM) batch size 512, 16 steps, epoch 200; (1D CNN) batch size 1024, eight steps, epoch 50; and (1D CNN+Bi-LSTM) batch size 32, 250 steps, epoch 50.

Close modal
1.
C.
Théry
,
M.
Ostrowski
, and
E.
Segura
, “
Membrane vesicles as conveyors of immune responses
,”
Nat. Rev. Immunol.
9
,
581
593
(
2009
).
2.
C.
Théry
,
K. W.
Witwer
,
E.
Aikawa
et al, “
Minimal information for studies of extracellular vesicles 2018 (MISEV2018): A position statement of the international society for extracellular vesicles and update of the MISEV2014 guidelines
,”
J. Extracell. Vesicles
7
,
1535750
(
2018
).
3.
R. A.
Dragovic
,
C.
Gardiner
,
A. S.
Brooks
,
D. S.
Tannetta
,
D. J. P.
Ferguson
,
P.
Hole
,
B.
Carr
,
C. W. G.
Redman
,
A. L.
Harris
,
P. J.
Dobson
,
P.
Harrison
, and
I. L.
Sargent
, “
Sizing and phenotyping of cellular vesicles using nanoparticle tracking analysis
,”
Nanomed. Nanotechnol. Biol. Med.
7
,
780
788
(
2011
).
4.
J.
Gross
,
S.
Sayle
,
A. R.
Karow
,
U.
Bakowsky
, and
P.
Garidel
, “
Nanoparticle tracking analysis of particle size and concentration detection in suspensions of polymer and protein samples: Influence of experimental and data evaluation parameters
,”
Eur. J. Pharm. Biopharm.
104
,
30
41
(
2016
).
5.
M. A.
El-Sayed
, “
Some interesting properties of metals confined in time and nanometer space of different shapes
,”
Acc. Chem. Res.
34
,
257
264
(
2001
).
6.
C. J.
Orendorff
,
T. K.
Sau
, and
C. J.
Murphy
, “
Shape-dependent plasmon-resonant gold nanoparticles
,”
Small
2
,
636
639
(
2006
).
7.
J. J.
Mock
,
M.
Barbic
,
D. R.
Smith
,
D. A.
Schultz
, and
S.
Schultz
, “
Shape effects in plasmon resonance of individual colloidal silver nanoparticles
,”
J. Chem. Phys.
116
,
6755
6759
(
2002
).
8.
C. J.
Orendorff
,
A.
Gole
,
T. K.
Sau
, and
C. J.
Murphy
, “
Surface-enhanced Raman spectroscopy of self-assembled monolayers: Sandwich architecture and nanoparticle shape dependence
,”
Anal. Chem.
77
,
3261
3266
(
2005
).
9.
E. V.
Timofeeva
,
J. L.
Routbort
, and
D.
Singh
, “
Particle shape effects on thermophysical properties of alumina nanofluids
,”
J. Appl. Phys.
106
,
014304
(
2009
).
10.
E. M.
Hotze
,
T.
Phenrat
, and
G. V.
Lowry
, “
Nanoparticle aggregation: Challenges to understanding transport and reactivity in the environment
,”
J. Environ. Qual.
39
,
1909
1924
(
2010
).
11.
P.
Khadka
,
J.
Ro
,
H.
Kim
,
I.
Kim
,
J. T.
Kim
,
H.
Kim
,
J. M.
Cho
,
G.
Yun
, and
J.
Lee
, “
Pharmaceutical particle technologies: An approach to improve drug solubility, dissolution and bioavailability
,”
Asian J. Pharm. Sci.
9
,
304
316
(
2014
).
12.
J. A.
Champion
,
Y. K.
Katare
, and
S.
Mitragotri
, “
Particle shape: A new design parameter for micro- and nanoscale drug delivery carriers
,”
J. Controlled Release
121
,
3
9
(
2007
).
13.
C.
Carnovale
,
G.
Bryant
,
R.
Shukla
, and
V.
Bansal
, “
Identifying trends in gold nanoparticle toxicity and uptake: Size, shape, capping ligand, and biological corona
,”
ACS Omega
4
,
242
256
(
2019
).
14.
J. A.
Champion
and
S.
Mitragotri
, “
Role of target geometry in phagocytosis
,”
Proc. Natl. Acad. Sci. U.S.A.
103
,
4930
4934
(
2006
).
15.
L.
Risom
,
P.
Møller
, and
S.
Loft
, “
Oxidative stress-induced DNA damage by particulate air pollution
,”
Mutat. Res., Fundam. Mol. Mech. Mutagen.
592
,
119
137
(
2005
).
16.
Y.
Han
,
A. M.
Alsayed
,
M.
Nobili
,
J.
Zhang
,
T. C.
Lubensky
, and
A. G.
Yodh
, “
Brownian motion of an ellipsoid
,”
Science
314
,
626
630
(
2006
).
17.
I.
Hanasaki
and
Y.
Isono
, “
Detection of diffusion anisotropy due to particle asymmetry from single-particle tracking of Brownian motion by the large-deviation principle
,”
Phys. Rev. E
85
,
051134
(
2012
).
18.
N. A.
Fuchs
,
The Mechanics of Aerosols
(
Pergamon
,
Oxford
,
1964
), pp.
34
42
.
19.
G. K.
Batchelor
, “
Slender-body theory for particles of arbitrary cross-section in Stokes flow
,”
J. Fluid Mech.
44
,
419
440
(
1970
).
20.
G. K.
Youngren
and
A.
Acrivos
, “
Stokes flow past a particle of arbitrary shape: A numerical method of solution
,”
J. fluid Mech.
69
,
377
403
(
1975
).
21.
D. L.
Johnson
,
D.
Leith
, and
P. C.
Reist
, “
Drag on non-spherical, orthotropic aerosol particles
,”
J. Aerosol Sci.
18
,
87
97
(
1987
).
22.
C.-T.
Lee
and
D.
Leith
, “
Drag force on agglomerated spheres in creeping flow
,”
J. Aerosol Sci.
20
,
503
513
(
1989
).
23.
T.
Franosch
,
M.
Grimm
,
M.
Belushkin
,
F. M.
Mor
,
G.
Foffi
,
L.
Forró
, and
S.
Jeney
, “
Resonances arising from hydrodynamic memory in Brownian motion
,”
Nature
478
,
85
88
(
2011
).
24.
J.
Mo
and
M. G.
Raizen
, “
Highly resolved Brownian motion in space and in time
,”
Annu. Rev. Fluid. Mech.
51
,
403
428
(
2019
).
25.
C.
Peng
,
T.
Turiv
,
Y.
Guo
,
S. V.
Shiyanovskii
,
Q.-H.
Wei
, and
O. D.
Lavrentovich
, “
Control of colloidal placement by modulated molecular orientation in nematic cells
,”
Sci. Adv.
2
,
e1600932
(
2016
).
26.
H.
Brenner
, “
Nonisothermal Brownian motion: Thermophoresis as the macroscopic manifestation of thermally biased molecular motion
,”
Phys. Rev. E
72
,
061201
(
2005
).
27.
W. A.
Steele
, “
The rotation of molecules in dense phases
,” in
Advances in Chemical Physics
,
34
(
John Wiley & Sons
,
1976
), pp.
1
104
.
28.
F.
Perrin
, “
Mouvement Brownien d'un ellipsoide - I. Dispersion diélectrique pour des molécules ellipsoidales
,”
J. Phys. Radium
5
,
497
511
(
1934
).
29.
M. M.
Tirado
,
C. L.
Martínez
, and
J. G.
de la Torre
, “
Comparison of theories for the translational and rotational diffusion coefficients of rod-like macromolecules. Application to short DNA fragments
,”
J. Chem. Phys.
81
,
2047
2052
(
1984
).
30.
Y.
LeCun
,
Y.
Bengio
, and
G.
Hinton
, “
Deep learning
,”
Nature
521
,
436
444
(
2015
).
31.
Y.
Hu
and
M. J.
Buehler
, “
Deep language models for interpretative and predictive materials science
,”
APL Mach. Learn.
1
,
010901
(
2023
).
32.
N.
Granik
,
L. E.
Weiss
,
E.
Nehme
,
M.
Levin
,
M.
Chein
,
E.
Perlson
,
Y.
Roichman
, and
Y.
Shechtman
, “
Single-particle diffusion characterization by deep learning
,”
Biophys. J.
117
,
185
192
(
2019
).
33.
P.
Kowalek
,
H.
Loch-Olszewska
, and
J.
Szwabiński
, “
Classification of diffusion modes in single-particle tracking data: Feature-based versus deep-learning approach
,”
Phys. Rev. E
100
,
032410
(
2019
).
34.
Y.
Matsuda
,
I.
Hanasaki
,
R.
Iwao
,
H.
Yamaguchi
, and
T.
Niimi
, “
Estimation of diffusive states from single-particle trajectory in heterogeneous medium using machine-learning methods
,”
Phys. Chem. Chem. Phys.
20
,
24099
(
2018
).
35.
P.
Piedra
,
A.
Kalume
,
E.
Zubko
,
D.
Mackowski
,
Y.-L.
Pan
, and
G.
Videen
, “
Particle-shape classification using light scattering: An exercise in deep learning
,”
J. Quant. Spectrosc. Radiat. Transfer
231
,
140
156
(
2019
).
36.
T.
Akagi
and
T.
Ichiki
, “
Microcapillary chip-based extracellular vesicle profiling system
,” in
Extracellular Vesicles: Methods and Protocols
, edited by
W. P.
Kuo
,
S.
Jia
(
Springer
,
New York
,
2017
), pp.
209
217
.
37.
F. A.
Gers
,
J.
Schmidhuber
, and
F.
Cummins
, “
Learning to forget: Continual prediction with LSTM
,”
Neural Comput.
12
,
2451
2471
(
2000
).
38.
N.
Hatami
,
G.
Yann
, and
J.
Debayle
, “
Classification of time-series images using deep convolutional neural networks
,” in
Proceedings of the 10th International Conference on Machine Vision (ICMV,
2017
) (SPIE,
2018
), p.
10696
.
39.
J.
Schmidhuber
, “
Deep learning in neural networks: An overview
,”
Neural Networks
61
,
85
117
(
2015
).
40.
J. T.
Connor
,
R. D.
Martin
, and
L. E.
Atlas
, “
Recurrent neural networks and robust time series prediction
,”
IEEE Trans. Neural Networks
5
,
240
254
(
1994
).
41.
T.
Mikolov
,
M.
Karafiát
,
L.
Burget
,
J.
Cernocký
, and
S.
Khudanpur
, “
Recurrent neural network based language model
,” in
Proceedings of Interspeech 2010 (International Speech Communication Association, 2010),
pp.
1045
1058
.
42.
K.
Sase
and
Y.
Shibuta
, “
Prediction of microstructure evolution at the atomic scale by deep generative model in combination with recurrent neural networks
,”
Acta Mater.
259
,
119295
(
2023
).
43.
S.
Hochreiter
and
J.
Schmidhuber
, “
Long short-term memory
,”
Neural Comput.
9
,
1735
1780
(
1997
).
44.
Y.
Le Cun
,
B.
Boser
,
J. S.
Denker
,
R. E.
Howard
,
W.
Habbard
, and
L. D.
Jackel
, “
Handwritten digit recognition with a back-propagation network
,” in
Advances in Neural Information Processing Systems
,
2
(
MIT Press
,
1990
), pp.
396
404
.
45.
K.
Diederik
and
J.
Ba
, “
Adam: A method for stochastic optimization
,” arXiv:1412.6980 (
2014
), ICLR2015.
46.
N.
Srivastava
,
G.
Hinton
,
A.
Krizhevsky
,
I.
Sutskever
, and
R.
Salakhutdinov
, “
Dropout: A simple way to prevent neural networks from overfitting
,”
J. Machine Learn. Res.
15
,
1929
1958
(
2014
).
47.
Y.
Matsuura
,
N.
Ouchi
,
A.
Nakamura
, and
H.
Kato
, “
Determination of an accurate size distribution of nanoparticles using particle tracking analysis corrected for the adverse effect of random Brownian motion
,”
Phys. Chem. Chem. Phys.
20
,
17839
17846
(
2018
).
48.
K.
Braeckmans
,
K.
Buyens
,
W.
Bouquet
,
C.
Vervaet
,
P.
Joye
,
F. D.
Vos
,
L.
Plawinski
,
L.
Doeuvre
,
E.
Angles-Cano
,
N. N.
Sanders
,
J.
Demeester
, and
S. C. D.
Smedt
, “
Sizing nanomatter in biological fluids by fluorescence single particle tracking
,”
Nano Lett.
10
,
4435
4442
(
2010
).
49.
V.
Filipe
,
A.
Hawe
, and
W.
Jiskoot
, “
Critical evaluation of nanoparticle tracking analysis (NTA) by NanoSight for the measurement of nanoparticles and protein aggregates
,”
Pharm. Res.
27
,
796
810
(
2010
).

Supplementary Material