LUCAID: Agentic Multimodal AI for Lung Cancer Precision Pathology
为解决肺癌病理诊断复杂且现有AI工具性能有限的问题，开发了LUCAID系统，集成多种功能模块，实现精准诊断，并在临床验证中表现出高一致性。
0 citationsRead paper
Selected work
Representative Papers
Atlas H&E-TME: Scalable AI-Based Tissue Profiling at Expert Pathologist-Level Accuracy
Current H&E whole-slide images lack scalable, high-precision quantitative analysis methods capable of robustly characterizing the tumor microenvironment. This study leverages the Atlas foundation model in pathology to develop an AI system that accurately predicts tissue quality, regional annotations, and cell types, generating over 4,500 cell-level quantitative metrics per slide. To enhance annotation consistency, the authors introduce an innovative immunohistochemistry (IHC)-guided multi-pathologist consensus protocol and validate performance through a dual-verification framework combining IHC and large-scale H&E annotations. Evaluated on real-world data encompassing over 1,500 cases across eight cancer types with more than 200,000 high-confidence annotations, the system demonstrates robust performance across cancer types and scanning platforms, achieving diagnostic accuracy comparable to or exceeding that of human experts—marking the first scalable, high-precision quantitative analysis of H&E slides.
0 citationsRead paper
In-Context Multiple Instance Learning
This work addresses the challenge of overfitting and poor generalization in multiple instance learning (MIL) under label-scarce conditions by proposing a context-based, fine-tuning-free approach. The method leverages a Perceiver architecture pretrained on diverse synthetic bag-structured datasets, integrating complementary inductive biases from varied generation strategies. This enables the model to perform accurate classification on new MIL tasks through a single forward pass with only a few labeled bags, without requiring any gradient-based adaptation. Evaluated across twelve established MIL benchmarks, the proposed approach consistently outperforms supervised baselines that rely on task-specific training, demonstrating substantially improved generalization and practical utility in few-shot MIL scenarios.
0 citationsRead paper
Context Sensitivity Improves Human-Machine Visual Alignment
This work addresses the limited contextual sensitivity of existing visual representation models, which hinders alignment between human and machine visual judgments. To bridge this gap, the authors propose a context-aware similarity computation method that introduces contextual mechanisms into visual embedding learning for the first time. Specifically, they formulate a triplet-based “odd-one-out” detection task, where the anchor image serves as a shared context to dynamically modulate object representations. Evaluated on both original and human-aligned foundation vision models, the approach significantly improves consistency with human perceptual judgments, achieving up to a 15% increase in odd-one-out detection accuracy.
0 citationsRead paper
OpenTME: An Open Dataset of AI-powered H&E Tumor Microenvironment Profiles from TCGA
This study addresses the lack of large-scale, quantitative, and consistent characterization of the tumor microenvironment (TME) from routine hematoxylin and eosin (H&E)-stained whole-slide images. Leveraging the Atlas foundation model in computational pathology, we developed an AI-driven H&E-TME analysis pipeline to process 3,634 whole slides across five cancer types from The Cancer Genome Atlas (TCGA). The pipeline performs tissue quality control, semantic segmentation, cell detection and classification, and spatial neighborhood analysis, yielding over 4,500 cell-level quantitative features per slide. We constructed and publicly released the OpenTME dataset—the first large-scale, high-resolution, AI-generated quantitative atlas of the TME derived solely from H&E stains. OpenTME is now available on Hugging Face for non-commercial academic research, aiming to advance spatial biology and computational methodology development.
0 citationsRead paper
Recent publications
Latest Papers
LUCAID: Agentic Multimodal AI for Lung Cancer Precision Pathology
为解决肺癌病理诊断复杂且现有AI工具性能有限的问题，开发了LUCAID系统，集成多种功能模块，实现精准诊断，并在临床验证中表现出高一致性。
0 citationsRead paper
Atlas H&E-TME: Scalable AI-Based Tissue Profiling at Expert Pathologist-Level Accuracy
Current H&E whole-slide images lack scalable, high-precision quantitative analysis methods capable of robustly characterizing the tumor microenvironment. This study leverages the Atlas foundation model in pathology to develop an AI system that accurately predicts tissue quality, regional annotations, and cell types, generating over 4,500 cell-level quantitative metrics per slide. To enhance annotation consistency, the authors introduce an innovative immunohistochemistry (IHC)-guided multi-pathologist consensus protocol and validate performance through a dual-verification framework combining IHC and large-scale H&E annotations. Evaluated on real-world data encompassing over 1,500 cases across eight cancer types with more than 200,000 high-confidence annotations, the system demonstrates robust performance across cancer types and scanning platforms, achieving diagnostic accuracy comparable to or exceeding that of human experts—marking the first scalable, high-precision quantitative analysis of H&E slides.
0 citationsRead paper
In-Context Multiple Instance Learning
This work addresses the challenge of overfitting and poor generalization in multiple instance learning (MIL) under label-scarce conditions by proposing a context-based, fine-tuning-free approach. The method leverages a Perceiver architecture pretrained on diverse synthetic bag-structured datasets, integrating complementary inductive biases from varied generation strategies. This enables the model to perform accurate classification on new MIL tasks through a single forward pass with only a few labeled bags, without requiring any gradient-based adaptation. Evaluated across twelve established MIL benchmarks, the proposed approach consistently outperforms supervised baselines that rely on task-specific training, demonstrating substantially improved generalization and practical utility in few-shot MIL scenarios.
0 citationsRead paper
Context Sensitivity Improves Human-Machine Visual Alignment
This work addresses the limited contextual sensitivity of existing visual representation models, which hinders alignment between human and machine visual judgments. To bridge this gap, the authors propose a context-aware similarity computation method that introduces contextual mechanisms into visual embedding learning for the first time. Specifically, they formulate a triplet-based “odd-one-out” detection task, where the anchor image serves as a shared context to dynamically modulate object representations. Evaluated on both original and human-aligned foundation vision models, the approach significantly improves consistency with human perceptual judgments, achieving up to a 15% increase in odd-one-out detection accuracy.
0 citationsRead paper
OpenTME: An Open Dataset of AI-powered H&E Tumor Microenvironment Profiles from TCGA
This study addresses the lack of large-scale, quantitative, and consistent characterization of the tumor microenvironment (TME) from routine hematoxylin and eosin (H&E)-stained whole-slide images. Leveraging the Atlas foundation model in computational pathology, we developed an AI-driven H&E-TME analysis pipeline to process 3,634 whole slides across five cancer types from The Cancer Genome Atlas (TCGA). The pipeline performs tissue quality control, semantic segmentation, cell detection and classification, and spatial neighborhood analysis, yielding over 4,500 cell-level quantitative features per slide. We constructed and publicly released the OpenTME dataset—the first large-scale, high-resolution, AI-generated quantitative atlas of the TME derived solely from H&E stains. OpenTME is now available on Hugging Face for non-commercial academic research, aiming to advance spatial biology and computational methodology development.
0 citationsRead paper