DepViT-CAD: Deployable Vision Transformer-Based Cancer Diagnosis in Histopathology

2025-07-14

AI Summary

To address the clinical need for precise, real-time multi-class cancer diagnosis in histopathological whole-slide images (WSIs), we propose a novel Multi-Attention Vision Transformer (MAViT). MAViT jointly models local morphological details and global structural context via complementary attention mechanisms, thereby enhancing discriminative capability for subtle inter-tumor morphological variations across 11 cancer types. Trained on 1,008 expert-annotated WSIs, the model achieves 94.11% sensitivity on the TCGA cohort and 92.0% on independent real-world clinical routine cases—surpassing existing ViT-based baselines. Designed for efficiency and robust generalization, MAViT features a lightweight architecture with reduced computational overhead while maintaining high accuracy and resilience to domain shifts. It has been successfully integrated into a deployable AI-assisted diagnostic system, delivering clinically actionable, high-fidelity, and robust decision support for pathologists.

Technology Category

• Computer Vision: Multi-modal Vision
• Machine Learning: Multi-instance/Multi-view Learning
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Responsible Web: Machine-in-the-loop, human agency and autonomy

Abstract

Accurate and timely cancer diagnosis from histopathological slides is vital for effective clinical decision-making. This paper introduces DepViT-CAD, a deployable AI system for multi-class cancer diagnosis in histopathology. At its core is MAViT, a novel Multi-Attention Vision Transformer designed to capture fine-grained morphological patterns across diverse tumor types. MAViT was trained on expert-annotated patches from 1008 whole-slide images, covering 11 diagnostic categories, including 10 major cancers and non-tumor tissue. DepViT-CAD was validated on two independent cohorts: 275 WSIs from The Cancer Genome Atlas and 50 routine clinical cases from pathology labs, achieving diagnostic sensitivities of 94.11% and 92%, respectively. By combining state-of-the-art transformer architecture with large-scale real-world validation, DepViT-CAD offers a robust and scalable approach for AI-assisted cancer diagnostics. To support transparency and reproducibility, software and code will be made publicly available at GitHub.

Problem

• Develops AI system for multi-class cancer diagnosis
• Captures fine-grained patterns in histopathology slides
• Validates accuracy on real-world clinical datasets

Innovation

• Multi-Attention Vision Transformer for cancer diagnosis
• Trained on 1008 expert-annotated whole-slide images
• Validated on 325 clinical cases with high accuracy