Quantitative Analysis of H&E Histopathology and Cancer Outcomes in PLCO

Principal Investigator

Name
James Kelley

Degrees
M.D./Ph.D.

Institution
Pictor Labs. Inc.

Position Title
Chief Medical Officer

Email
jkelley@pictorlabs.ai

About this CDAS Project

Study
PLCO (Learn more about this study)

Project ID
PLCOI-2081

Initial CDAS Request Approval
Sep 28, 2026

Title
Quantitative Analysis of H&E Histopathology and Cancer Outcomes in PLCO

Summary
This project will investigate whether quantitative patterns of features present in routine H&E-stained pathology images are associated with clinical outcomes across multiple cancer types in the PLCO cohort. We will perform a retrospective computational pathology study using de-identified whole-slide images together with linked clinical, pathological, treatment, and longitudinal follow-up information.
Computational image-analysis methods will be used to characterize tissue morphology and organization in whole-slide images and to derive quantitative representations of tumor histology. We will evaluate associations between these image-derived measurements and available longitudinal outcomes, with a primary focus on overall survival and cancer-specific mortality. Where appropriate, analyses will account for established clinical and pathological factors such as age, stage, histologic characteristics, and treatment information.
The availability of pathology images across multiple PLCO cancer types will also allow us to evaluate the reproducibility and generalizability of quantitative histopathology methods across distinct tissues and disease settings. Analyses will be conducted at the participant level, with appropriate separation of development and evaluation cohorts to minimize overfitting. Results will be analyzed in aggregate and may be shared through scientific presentations or publications.

Aims

Aim 1: Quantify clinically relevant morphological information in H&E whole-slide images. Develop and evaluate computational methods for deriving quantitative representations of tumor morphology, tissue architecture, and spatial organization from digitized PLCO pathology specimens.
Aim 2: Determine whether image-derived measurements are associated with longitudinal cancer outcomes. Evaluate associations with overall survival, cancer-specific mortality, and other appropriate longitudinal endpoints available within PLCO. Where relevant, assess whether image-derived information provides information beyond established clinical and pathological variables such as age, stage, histologic characteristics, and initial treatment.
Aim 3: Evaluate robustness and generalizability across cancer types. Use the multiple pathology cohorts available within PLCO to determine whether quantitative histopathology approaches remain informative across different tumor types, patient populations, and tissue characteristics.
Aim 4: Characterize sources of variation and model reliability. Examine performance across clinically relevant subgroups and evaluate the stability of image-derived measurements across variations in tissue specimens and pathology-slide characteristics.

Collaborators

Yair Rivenson Pictor Labs. Inc.
Yair Rivenson Pictor Labs. Inc.
David Sriker Pictor Labs. Inc.
James Kelley Pictor Labs. Inc.