cs.LGAug 17, 2026

Data-Efficient and Interpretable Classification of Circulating Tumor Cell Phenotypes in Microfluidic Devices via Deep Learning

Authors: Serena Su, Yifan Wang, Senwei Liang

Organizations: Department of Mathematics & Statistic, Texas Tech University, Lubbock, 79407, Texas, USA

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

Accurate classification of circulating tumor cell (CTC) phenotypes can provide valuable information for assessing metastatic potential. Label free microfluidic devices provide a hydrodynamic obstacle course that transforms subtle biophysical characteristics of CTCs, including size and deformability, into distinct kinematic trajectories. However, the highly nonlinear fluid structure interactions governing these trajectories make the inverse problem of inferring cellular phenotype from trajectory data analytically intractable. While deep neural networks (DNNs) have emerged as a powerful approach for addressing this inverse problem, their effectiveness is constrained by the limited availability of trajectory data and the lack of physical interpretability. To address these challenges, we propose an interpretable and data efficient DNN framework for trajectory based CTC classification. To mitigate the scarcity of data, we develop Subsequence (SubSeq), a targeted augmentation strategy that randomly extracts informative local trajectory segments during training to promote learning from localized patterns. We further apply Gradient Weighted Class Activation Mapping to identify the trajectory features and physical regions of the microfluidic device that drive model predictions. Experimental results demonstrate that SubSeq improves classification accuracy over the evaluated baseline and augmentation methods. Furthermore, interpretability analysis suggests that localized trajectory segments contain substantial biophysical information relevant to accurate classification. This provides justification for SubSeq and also highlights the redundancy of full-length trajectories. More broadly, the proposed framework views microfluidic geometries as physical encoders of cellular mechanical properties, providing mechanistic insights that may inform the future design of diagnostic devices.

Figures & tables

Figure 1: Subfigure I. The hyperuniform microfluidic device is shown in (a), with a schematic illustration of the micropost arrangement presented in (b). Subfigure II. Experimentally observed trajectories of PC3 and SKBR3 cancer cells are shown in an overall view in (a) and in two enlarged views in (b) and (c), corresponding to the regions indicated by the labeled black boxes (Images courtesy of the Wei Li laboratory). PC3 cells are shown in green, whereas SKBR3 cells are shown in blue. Subfigure III. The numerical simulation demonstrates distinct trajectories for two simulated circulating tumor cell types, shown in red and yellow, with different elastic properties. Both cells are released from the same initial location ( Rejuan et al., 2025 ) .
Figure 2: Convolutional neural network for CTC classification ( Kumar et al., 2025 ) . The model processes the position and velocity inputs through parallel branches, concatenates the extracted features, and finally performs classification using a multilayer perceptron to produce the final phenotype classification.
Figure 3: Comparison of original and SubSeq model performance across varying train-test splits. Boxplots show the distribution of testing accuracy (left) and ROC-AUC (right), with paired boxes indicating models trained on identical data splits. The central line denotes the median, boxes indicate the interquartile range, whiskers represent the data range, and hollow circles indicate outliers. Results show that SubSeq consistently achieves higher median testing accuracy and ROC-AUC than the baseline.
Figure 4: Performance comparison of data augmentation strategies under an 80%/20% training/testing split. Boxplots display the distribution of testing accuracy (left) and ROC-AUC (right) across the SubSeq, Cutout, and Mixup methods. Results show that SubSeq offers the most stable performance (lower variability) in both testing accuracy and ROC-AUC across the evaluated augmentation methods.
Figure 5: Effect of input representation on model performance. Boxplots display the distributions of testing accuracy (left) and ROC-AUC (right) for models trained using SubSeq with velocity-only/position-only inputs or combined position and velocity inputs across varying train–test splits. Results show that velocity provides most of the discriminative information, while combining position and velocity yields the best performance.
Figure 6: ROC-AUC under an 80/20 train–test split across different input and augmentation configurations. Panels illustrate models trained with: (Top Left) combined position and velocity inputs without data augmentation, (Top Right) Position-only inputs with SubSeq, (Bottom Left) Velocity-only inputs with SubSeq, and (Bottom Right) combined position and velocity inputs with SubSeq. Results show that SubSeq improves the ROC-AUC from 0.8612 to 0.9347 when both position and velocity are used. We can see that velocity provides more discriminative information than position, while combining both features achieves the best classification performance.
Figure 7: Grad-CAM importance maps for several representative test trajectories under the soft- and hard-cell prediction targets, for each input modality (position or velocity). Lighter colors indicate larger Grad-CAM importance scores. The labels on the right denote the specific conditions and outcomes for each trajectory. For example, label (a) indicates a Soft Cell trajectory analyzed via the Position Branch, which the model correctly classified with a probability of 1.0.
Figure 8: Grad-CAM importance maps aggregated over test trajectories under the soft- and hard-cell prediction targets, for each input modality (position and velocity). Lighter colors indicate larger Grad-CAM importance scores. The labels on the right indicate the conditions for the aggregated Grad-CAM maps. For instance, label (a) corresponds to the aggregated importance map for Soft Cells evaluated through the Position Branch.
More in this topic

Interpretable ML

Interpretable machine learning develops models and explanations that make predictions, evidence, and decision processes understandable to people.

See all 306 papers

Explore similar work

CardsList
May 14, 2026

Towards Label-Free Single-Cell Phenotyping Using Multi-Task Learning

Label-free single-cell imaging offers a scalable, non-invasive alternative to fluorescence-based cytometry, yet inferring molecular phenotypes directly from bright-field morphology remains challenging. We present a unified Deep Learning (DL) framework that jointly performs White Blood Cell (WBC) classification and continuous protein-expression regression from label-free Differential Phase Contrast (DPC) images. Our model employs a Hybrid architecture that fuses convolutional fine-grained texture features with transformer-based global representations through a learnable cross-branch gating module, enabling robust morpho-molecular inference from DPC images. To support downstream interpretability, we further incorporate a Large Language Model (LLM) that generates concise, biologically grounded summaries of the predicted cell states. Experiments on the Berkeley Single Cell Computational Microscopy (BSCCM) and Blood Cells Image benchmarks demonstrate strong performance, achieving a 91.3% WBC classification accuracy and a 0.72 Pearson correlation for CD16 expression regression on BSCCM. These results underscore the promise of label-free single-cell imaging for cost-effective hematological profiling, enabling simultaneous phenotype identification and quantitative biomarker estimation without fluorescent staining. The source code is available at https://github.com/saqibnaziir/Single-Cell-Phenotyping.

Saqib Nazir, Ardhendu Behera

Department of Computer Science, Edge Hill University, UK

Multi-Task LearningWhite Blood Cell ClassificationMicroscopy Image Analysis
AskPDF
Jun 15, 2026

Robust Image-Driven Phenotyping of Ovarian Tumor Cells using Optimized Dynamic Features in Hyperbolic Channels

Label-free, image-based cellular mechanophenotyping in microfluidic devices provides a high-throughput method for single-cell profiling. However, while complex microchannels (e.g., hyperbolic geometries) reveal transient deformation dynamics under continuous extensional stress, the resulting high-dimensional feature spaces are highly susceptible to hydrodynamic artifacts. Flow rate variations often distort discriminative boundaries, linking feature distributions to fluid conditions rather than intrinsic biology. To overcome this, we introduce a stability-guided analytical framework that decouples flow-induced noise from authentic mechanobiological signatures. We tracked the morphodynamic, kinematic, and intracellular optical-density trajectories of healthy and malignant ovarian cells to build a 93-dimensional feature space. Using a cross-flow screening strategy based on structural consistency and statistical persistence, we isolated robust descriptors, creating task-adapted subsets (20 features for binary classification; 25 for cancer subtyping). Variance-attribution analysis confirmed the neutralization of flow-conditioned artifacts; notably, flow-associated variance in the primary principal component fell from 69.9% to 9.3% in the subtyping task. We also found that macroscopic binary discrimination depends on bulk kinematic transitions, while clonal subtyping requires localized intracellular optical heterogeneity. These optimized subsets maintained diagnostic fidelity across multiple machine learning architectures and restricted sampling conditions. This framework establishes a robust, flow-independent foundation for continuous dynamic phenotyping.

Hong-Fei Li, Xi-Lin Gao, Yi-Juan Xiang +6

Cancer Hospital of Dalian University of Technology, Dalian University of Technology, Shenyang 110000, China · Faculty of Medicine, Dalian University of Technology, Dalian 116024, China · Hunan Provincial Maternal and Child Health Care Hospital, Changsha 410000, China +2

Feature Selection
AskPDF
Sep 7, 2026

Multi-label versus multi-class classification of blood cells and their aggregates in microfluidic channels

Deformability cytometry (DC) is a type of imaging flow cytometry, which uses a camera-equipped device to measure cellular stiffness in addition to other cellular properties at high throughput. Cellular properties such as area and elongation can identify cell types, but this requires prior knowledge of distinguishing properties and cannot be applied to clinically important cell aggregates. Using DC data, we evaluated conventional multi-class (MC) classification and introduced a multi-label (ML) approach for identifying blood cells and their aggregates. In particular, an ML classifier can simultaneously assign multiple cell-type labels to a single imaged event. We show that, unlike MC classification, ML classification can identify cell aggregates not represented in the training data. It also avoids the need for exhaustive, strictly defined aggregate labels, thereby simplifying and speeding up annotation. Since automated blood analyzers do not reliably analyze cell aggregates, our approach may help address this clinical gap.

Igor Zingman, Shada Abuhattum, Sara Kaliman +8

Max Planck Institute for the Science of Light, Erlangen, Germany · Max-Planck-Zentrum für Physik und Medizin · Friedrich-Alexander Universiät Erlangen-Nürnberg, Erlangen, Germany

Multi-Label ClassificationMulticlass ClassificationWhite Blood Cell Classification
AskPDF