illustration of DNA strand and abstract shapes on a computer screen on a green gradient background

Medical, health, and genomics

Digital pathology helps decode tumor microenvironments for precision immunotherapy. GigaPath is a novel vision transformer that can scale to gigapixel whole-slide images by adapting dilated attention for digital pathology. In joint work with Providence and UW, we’re sharing Prov-GigaPath, the first whole-slide pathology foundation model pretrained on large-scale real-world data, for advancing clinical research and discovery.

GigaPath: Whole-Slide Foundation Model for Digital Pathology

Research Forum Ep2 | Panel | Transforming the Natural Sciences with AI

Research Forum 2 | Panel Discussion: Transforming the Natural Sciences with AI

white icons of first aid kit, DNA strand, laptop monitor with overlapping eye, and microscope on a blue and green gradient background

Scaling early detection of esophageal cancer with AI

female doctor / nurse reviewing a tablet

Current selections

Sort by: Most recent

Clear selections

Search within these results

Content Types

Published Date

Publication

Evaluating the role of pre-training dataset size and diversity on single-cell foundation model performance

Alan DenAdel, Madeline Hughes, Akshaya Thoutam, Anay Gupta, Nicolo Fusi, Andrew W. Navia, Srivatsan Raghavan, Peter S. Winter, Ava P. Amini, Lorin Crawford

bioRxiv | December 2024

Project

Publication

MAIRA-Seg: Enhancing Radiology Report Generation with Segmentation-Aware Multimodal Large Language Models

Harshita Sharma, Valentina Salvatelli, Shaury Srivastav, Kenza Bouzid, Shruthi Bannur, Daniel Coelho de Castro, Maximilian Ilse, Sam Bond-Taylor, Mercy Ranjit, Fabian Falck, Fernando Pérez-García, Anton Schwaighofer, Hannah Richardson (nee Murfet), Maria Teodora Wetscherek, Stephanie Hyland, Javier Alvarez-Valle

Machine Learning for Health Symposium 2024 | December 2024

Project

Publication

Attribute Structuring Improves LLM-Based Evaluation of Clinical Text Summaries

Zelalem Gero, Chandan Singh, Yiqing Xie, Sheng Zhang, Tristan Naumann, Jianfeng Gao, Hoifung Poon

NeurIPS 2024 | December 2024

Publication

Deeper evaluation of a single-cell foundation model

Rebecca Boiarsky, Nalini M. Singh, Alejandro Buendia, Ava P. Amini, Gad Getz, David Sontag

Nature Machine Intelligence | December 2024

Publication

Efficacy of a conversational chatbot for cigarette smoking cessation: Protocol of the QuitBot full-scale randomized controlled trial

Jonathan B Bricker, Brianna M Sullivan, Kristin E Mull, Juan M. Lavista Ferres, Margarita Santiago-Torres

Contemporary Clinical Trials | December 2024

Project

Microsoft Research Blog

Advances in run-time strategies for next-generation foundation models

November 27, 2024 | Eric Horvitz, Harsha Nori, Naoto Usuyama

Discover the most effective run-time strategies on the OpenAI o1-preview model, improving accuracy in medical language tasks.

A visual illustration of Medprompt performance on the MedQA benchmark. Moving from left to right on a horizontal line, the illustration shows how different Medprompt components and additive contributions improve accuracy starting with zero-shot at 81.7 accuracy, to random few-shot at 83.9 accuracy, to random few-shot, chain-of-thought at 87.3 accuracy, to kNN, few-shot, chain-of-thought at 88.4 accuracy, to ensemble with choice shuffle at 90.2 accuracy.

Tool

RAD-DINO is a vision transformer model trained to encode chest X-rays using the self-supervised learning method DINOv2 (opens in new tab). RAD-DINO is described in detail in RAD-DINO: Exploring Scalable Medical Image Encoders Beyond Text Supervision (F.…

Access Publication

Tool

MAIRA-2 model

MAIRA-2 is a multimodal transformer designed for the generation of grounded or non-grounded radiology reports from chest X-rays. It is described in more detail in MAIRA-2: Grounded Radiology Report Generation (S. Bannur, K. Bouzid et al.,…

Access Publication

Tool

RadFact: An LLM-based Evaluation Metric for AI-generated Radiology Reporting

RadFact is a framework for the evaluation of model-generated radiology reports given a ground-truth report, with or without grounding. Leveraging the logical inference capabilities of large language models, RadFact is not a single number but a suite of…

GitHub Publication

Publication

Artificial Intelligence–Based Copilots to Generate Causal Evidence

Maya Petersen, Ahmed Alaa, Emre Kiciman, Chris Holmes, Mark van der Laan

NEJM AI | November 2024

Medical, health, and genomics

GigaPath: Whole-Slide Foundation Model for Digital Pathology

Research Forum 2 | Panel Discussion: Transforming the Natural Sciences with AI

Scaling early detection of esophageal cancer with AI

Microsoft Health Futures

Evaluating the role of pre-training dataset size and diversity on single-cell foundation model performance

MAIRA-Seg: Enhancing Radiology Report Generation with Segmentation-Aware Multimodal Large Language Models

Attribute Structuring Improves LLM-Based Evaluation of Clinical Text Summaries

Deeper evaluation of a single-cell foundation model

Efficacy of a conversational chatbot for cigarette smoking cessation: Protocol of the QuitBot full-scale randomized controlled trial

Advances in run-time strategies for next-generation foundation models

RAD-DINO model

MAIRA-2 model

RadFact: An LLM-based Evaluation Metric for AI-generated Radiology Reporting

Artificial Intelligence–Based Copilots to Generate Causal Evidence

Medical, health, and genomics

Highlights