Interpretable Tile-Based Classification of Paclitaxel Exposure

2025-10-27

AI Summary

Fine-grained classification of subtle paclitaxel (Taxol) dosage exposure in phase-contrast microscopy images of C6 glioma cells remains challenging due to low contrast and high inter-class similarity. Method: We propose a local patch-based aggregation analysis framework that abandons whole-image modeling in favor of lightweight patch encoding and feature aggregation, integrated with Grad-CAM, Score-CAM, and attention mechanisms for interpretable modeling. Contribution/Results: Our key insight is the critical role of local region robustness in drug response identification, validated biologically via visualization. On the benchmark dataset, our model achieves ~20 percentage points higher accuracy than whole-image baselines; five-fold cross-validation confirms its stability and generalizability. This work establishes an interpretable, reusable paradigm for fine-grained discrimination in low-contrast, high-similarity medical imagery.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Transparent, Interpretable, Explainable ML
• Natural Language Processing: Interpretability, Analysis, and Evaluation of NLP Models

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Large pretrained models with web data

Abstract

Medical image analysis is central to drug discovery and preclinical evaluation, where scalable, objective readouts can accelerate decision-making. We address classification of paclitaxel (Taxol) exposure from phase-contrast microscopy of C6 glioma cells -- a task with subtle dose differences that challenges full-image models. We propose a simple tiling-and-aggregation pipeline that operates on local patches and combines tile outputs into an image label, achieving state-of-the-art accuracy on the benchmark dataset and improving over the published baseline by around 20 percentage points, with trends confirmed by cross-validation. To understand why tiling is effective, we further apply Grad-CAM and Score-CAM and attention analyses, which enhance model interpretability and point toward robustness-oriented directions for future medical image research. Code is released to facilitate reproduction and extension.

Problem

• Classifying paclitaxel exposure from glioma cell microscopy images
• Addressing subtle dose differences challenging full-image models
• Developing interpretable tile-based analysis for medical image classification

Innovation

• Tile-based classification pipeline for drug exposure
• Grad-CAM and Score-CAM enhance model interpretability
• Local patch aggregation achieves state-of-the-art accuracy