Systems And Methods For Assessment Based On Pathology Image Analysis

A machine-learning system that analyzes pathology images to assess cancer characteristics and prognosis by quantifying neuroendocrine differentiation and chromosomal instability.

There's currently a lack of accessible, cost-effective molecular biomarkers for aggressive cancer phenotypes that include chromosomal instability (CIN). This results in the need for expensive, time-consuming, and tissue-destructive genomic and cytogenetic tests for cancer grading and prognosis.

This technology involves computer-implemented methods and systems that process images of cancerous tissue to identify areas with high-grade cancer and quantify histopathologic features indicative of neuroendocrine differentiation. It applies machine-learning models to evaluate these features and generate scores—such as neuroendocrine differentiation scores and chromosomal instability-related prognosis metrics. The approach includes segmenting nuclei, analyzing nuclear morphology features across perinuclear regions, and leveraging multiplexed imaging channels to improve classification and prognosis predictions. This real-time analysis supports pathologists with diagnostics, prognosis, and treatment decision-making and is designed to work with routinely available pathology slides.

• May enable rapid assessment of cancer tissue, facilitating timely clinical decisions.
• Analyzes high-grade cancer tissue using machine-learning models trained on large feature sets for high accuracy.
• Utilizes routinely collected hematoxylin and eosin stained slides, avoiding the cost and complexity of genomic assays.
• Provides interpretable scores for neuroendocrine differentiation and chromosomal instability linked to prognosis and treatment resistance
• Supports integration into existing pathology workflows through digital slide analysis and report generation.

Steven Christiansen
801.587.0915
steven.christiansen@utah.edu