Using deep neural networks to interpret quantitative image analysis and gene expression data in lung cancer (from Smedley et al)
Overview
In the current data-rich healthcare environment, our capacity to collect vast amounts of longitudinal observational data needs to be matched with a comparable ability to continuously learn from the data and tailor clinical decisions to an individual. The Hsu Lab develops and validates computational tools to extract clinically meaningful insights from multimodal datasets. We apply artificial intelligence (AI)/machine learning (ML) techniques to mine and analyze clinical, diagnostic imaging, and molecular data, harnessing the combination of genomic and environmental information to achieve evidence-based management of patients and precision health.
Lab activities revolve around three areas:
Establishment of computational pipelines for normalizing, spatially registering, and extracting quantitative information from medical images to facilitate multimodal radiology-pathology analysis;
Development and validation of algorithms to integrate and generate predictions using multimodal data; and
Deployment of algorithms into practice, evaluating their impact on clinical workflows and patient outcomes.
Focus Areas
Image harmonization to improve the reliability of quantitative image analysis
Multimodal characterization of indeterminate pulmonary nodules
Longitudinal, multimodal analysis
Implementation of AI to enable same-day diagnostic follow-up of abnormal screening mammograms
Utilization of AI to increase adherence to cancer screening recommendations
Active Funding
National Cancer Institute
Boston University-UCLA Biomarker Characterization Center (U2C)
PI: Marc Lenberg, MPIs: Steven Dubinett, Jennifer Beane, William Hsu
EFIRM Liquid Biopsy Research Laboratory: Early Lung Cancer Assessment (U01)
PI: David Wong, MPIs: Denise Aberle, William Hsu, Kostyantyn Krysan, Charles Strom, Fang Wei
National Institute for Biomedical Imaging and Bioengineering
Computational Toolkit for Normalizing the Impact of CT Acquisition and Reconstruction on Quantitative Image Features (R01)
PI: William Hsu, MPI: Michael McNitt-Gray
An AI/ML-ready Dataset for Investigating the Effect of Variations in CT Acquisition and Reconstruction (Supplement)
PI: William Hsu, MPI: Michael McNitt-Gray
Medical Imaging Informatics Training Grant (T32)
PI: Alex Bui, MPI: Ashley Prosper, William Hsu
Agency for Healthcare Research and Quality
Using artificial intelligence to support efficient same-day diagnostic imaging in breast cancer screening (R21/R33)
PI: William Hsu, MPI: Anne Hoyt
V Foundation
Integrating Cell-Free DNA Methylome and CT Imaging to Determine the Malignancy of Lung Nodules
PI: Xianghong Jasmine Zhou, Co-PI: William Hsu
Early Diagnostics Inc
Cloud-based Liquid-biopsy and Radiomics Platform for the Cancer Research Data Commons (subcontract to NCI SBIR Phase II)
Subcontract PI: William Hsu