Skip to Main Content

An official website of the United States government

Government Funding Lapse

Because of a lapse in government funding, the information on this website may not be up to date, transactions submitted via the website may not be processed, and the agency may not be able to respond to inquiries until appropriations are enacted. The NIH Clinical Center (the research hospital of NIH) is open. For more details about its operating status, please visit  cc.nih.gov. Updates regarding government operating status and resumption of normal operations can be found at OPM.gov.

AI-Driven Analyses of Morphological and Phenotypic Profiles Across Cancers

Principal Investigator
Name
JUNHAN ZHAO
Degrees
Ph.D
Institution
The University of Chicago
Position Title
Research Assistant Professor
Email
About this CDAS Project
Study
PLCO (Learn more about this study)
Project ID
PLCOI-1985
Initial CDAS Request Approval
Sep 22, 2025
Title
AI-Driven Analyses of Morphological and Phenotypic Profiles Across Cancers
Summary
This project will apply advanced multimodal artificial intelligence methods to uncover pan-cancer associations between morphology and clinical outcomes. We will develop AI algorithms to extract high-dimensional morphological features from histopathology images and integrate these with clinical data for profiling. Approaches such as vision-only and vision–language models will be used to link microscopic patterns with annotations, patient phenotypes, and survival outcomes. By harmonizing multimodal representations across the four cancer types in the PLCO study, we aim to identify both shared and cancer-specific morphological signatures. The resulting predictive models are expected to yield clinically interpretable insights into cancer diagnosis and patient survival outcomes.
Aims

Aim 1: Develop AI-based pipelines to extract high-dimensional morphological features from histopathology images.
We will build quantitative image analysis workflows that leverage advanced AI techniques, including convolutional neural networks and vision-based foundation models, to process whole-slide pathology images. These approaches will generate comprehensive morphological representations that capture subtle tissue and cellular patterns not easily discernible by traditional methods.

Aim 2: Integrate histopathology images with clinical information using multimodal approaches (e.g., vision–language models) to capture clinically relevant context.
We will align histopathology image features with textual data from pathology reports and structured clinical variables. By applying multimodal AI frameworks such as vision–language models, we will connect visual patterns with semantic descriptions of tumor grade, morphology, and disease state, thereby enhancing interpretability and linking image-derived features directly to clinical context.

Aim 3: Correlate image- and text-derived features with clinical phenotypes, including tumor stage, demographics, treatment response, and survival outcomes.
We will systematically test associations between multimodal features and key patient characteristics across the PLCO cohort. These analyses will enable the identification of signatures predictive of disease progression and survival. The integration of multimodal features with survival analyses will highlight clinically actionable biomarkers of prognosis.

Aim 4: Construct pan-cancer predictive models that harmonize morphology, reports, and phenotype data across the four cancer types in the PLCO study.
We will build integrative models that unify histopathology, pathology reports, and patient-level phenotype information across lung, colorectal, ovarian, and prostate cancers in the PLCO dataset. These models will identify shared pan-cancer signatures while preserving subtype-specific associations, producing robust predictive tools that generalize across cancer types and advance the interpretability of morphology–phenotype relationships.

Collaborators

JUNHAN ZHAO University of Chicago