Digital Histology Grading for Bladder Cancer Recurrence Risk
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The current NMIBC grading system is highly subjective and inadequate for accurately categorizing patients by risk of cancer recurrence or progression, leading to overtreatment and excessive monitoring, and lacks a feedback loop for pathologists to improve grading accuracy.
Innovation Solution
A computer-implemented method using image analysis software to quantify nuclear features such as size, shape, and texture, combined with prognostic classifiers like Cox Proportional Hazards and Random Survival Forest models, to generate a continuous prognostic score for predicting recurrence-free survival.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If subjective histologic grading is used, then the grading process is simple and quick, but the prognostic accuracy and ability to predict recurrence/progression is poor
Solution Approach 1:
The patent replaces the subjective mechanical grading process with an automated computational system that uses image analysis software to objectively measure nuclear features. This substitution eliminates human bias and subjectivity while providing precise, reproducible quantitative measurements of nuclear morphology, size, and other features that predict cancer recurrence and progression.
Solution Approach 2:
The patent transforms qualitative histologic grading parameters into quantitative measurable parameters. Instead of relying on pathologist subjective assessment, the system measures specific nuclear features (size, shape, texture, chromatin patterns) and converts them into numerical data that can be statistically analyzed to predict patient outcomes with high accuracy.
2Productivity
If current NMIBC grading system is used, then the system is simple to implement, but it leads to overtreatment and excessive monitoring
Solution Approach 1:
The patent implements a feedback mechanism where quantitative nuclear feature measurements are used to generate prognostic scores that directly inform treatment decisions. The system provides feedback to clinicians about a patient's specific recurrence and progression risk, enabling personalized monitoring intervals and treatment strategies that avoid both overtreatment and undertreatment.
Solution Approach 2:
The system changes the monitoring parameter from binary high/low grade classification to continuous quantitative risk assessment. By measuring multiple nuclear features and generating continuous prognostic scores, the system enables graduated monitoring strategies that match actual patient risk, reducing excessive monitoring for low-risk patients while ensuring adequate surveillance for high-risk patients.
3Reliability
If pathologists manually grade tumors, then the workflow is straightforward, but there is no feedback loop to improve grading accuracy
Solution Approach 1:
The patent replaces manual pathologic grading with an automated computational system that consistently applies the same measurement criteria to all samples. This eliminates inter-pathologist variability and ensures grading consistency through standardized algorithmic assessment of nuclear features, while the system automatically integrates with clinical outcome data to continuously improve accuracy.
Solution Approach 2:
The system establishes a feedback loop where quantitative nuclear feature measurements are correlated with actual patient outcomes (recurrence and progression). This feedback enables the system to learn from real-world data and continuously refine its grading algorithms, improving reliability over time while maintaining automated workflow.
4Measurement precision
If binary classification models are used for prognosis, then the model is simple, but it causes harmful data loss and does not account for right-censoring
Solution Approach 1:
The patent changes from binary classification to continuous quantitative prediction. By measuring multiple nuclear features continuously and generating proportional hazard ratios, the system preserves all information about tumor aggressiveness without forcing arbitrary categorizations. This continuous approach naturally handles right-censoring by providing risk estimates for patients who have not yet experienced the outcome event.
Data Source
AI summary
A computer-implemented method for classifying cancer uses image analysis software to analyze a digital histology image of tumour cells of a patient. The image analysis software is trained to segment tissue regions, identify nuclei, and measure nuclear features of a plurality of nuclei. Summary statistics for nuclear feature values are obtained and one or multiple prognostic classifiers are applied to the patient's nuclear feature values to produce a prognostic score for the patient from the prognostic classifiers. The prognostic score may be for recurrence-free survival of the patient or may discriminate between high-grade and low-grade tumours.