Deep Learning Liver Pathology Scoring for NASH Trial Variability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current ordinal scoring systems for nonalcoholic steatohepatitis (NASH), such as the NAS and CRN fibrosis scoring systems, are unsuitable for accurately assessing the continuum of liver disease severity due to high inter- and intra-observer variability and inadequate capture of disease heterogeneity, leading to challenges in powering clinical trials and determining treatment response.
Innovation Solution
A machine learning-based approach using deep learning models, specifically deep convolutional neural networks, is developed to assess liver pathology by predicting tissue characteristic categories like steatosis, lobular inflammation, and fibrosis stage, generating continuous scores that capture disease heterogeneity and treatment effects, improving reproducibility and accuracy over manual scoring.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual ordinal scoring systems (NAS, CRN) are used to assess liver pathology, then the assessment process is simple and familiar to pathologists, but the measurement precision and reproducibility are poor due to high inter- and intra-observer variability
Solution Approach 1:
The patent replaces the manual mechanical scoring process with an automated machine learning system. Deep learning models process liver biopsy images to generate continuous pathology scores, eliminating human observer variability while maintaining assessment capability. The system substitutes human pathologist evaluation with algorithm-based analysis, achieving superior reproducibility without requiring complex manual procedures.
Solution Approach 2:
The patent transforms the assessment output from discrete ordinal categories (0-4 scale) to continuous numerical scores. This parameter change allows for finer granularity in disease severity measurement, capturing subtle variations that ordinal systems miss. The continuous scores enable more precise tracking of disease progression and treatment response while maintaining compatibility with existing clinical frameworks.
2Adaptability or versatility
If ordinal classification systems are used, then the scoring framework is simple and established, but the ability to capture disease heterogeneity and continuum of severity is inadequate
Solution Approach 1:
The patent adds a dimensional transformation by converting categorical ordinal data into continuous numerical space. This allows the system to represent disease severity along a continuous spectrum rather than discrete steps, capturing the true continuum of liver disease progression. The continuous scores provide an additional dimension of information that preserves subtle gradations in pathology severity.
Solution Approach 2:
The patent segments the liver biopsy image analysis into multiple independent continuous scores for different pathology features (steatosis, inflammation, ballooning, fibrosis). Each feature is assessed separately on a continuous scale, allowing the system to capture heterogeneity across different pathological dimensions while maintaining overall disease severity assessment capability.
3Reliability
If manual histological scoring is used, then the current regulatory framework is satisfied, but the high rate of histological response in placebo patients creates challenges for powering clinical trials
Solution Approach 1:
The patent creates a digital copy of the histological assessment process through machine learning models trained on annotated pathology images. This digital replica reproduces and enhances human pathologist evaluation capabilities while adding consistency and precision. The ML-based scores serve as a reliable alternative endpoint that reduces placebo response variability, enabling more efficient clinical trial design and patient enrollment calculations.
Data Source
AI summary
In some aspects, the described systems and methods provide for a method for training a deep learning model to assess liver pathology, including accessing annotated liver pathology images associated with a group of patients in one or more randomized controlled clinical trials of nonalcoholic steatohepatitis therapy, each of the annotated liver pathology images including at least one annotation describing one or more tissue characteristic categories for a portion of the image, and training the deep learning model based on the annotated liver pathology images to predict the tissue characteristic categories, selected from a group comprising steatosis, lobular inflammation, hepatocyte ballooning, and fibrosis stage.