Automated Kaplan-Meier Curve Digitization for Accurate IPD Extraction

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Conventional methods for extracting independent patient data (IPD) from Kaplan-Meier (KM) curves are labor-intensive, prone to errors, and lack precision, especially when dealing with complex or intertwined curves, affecting the reliability and reproducibility of the extracted data.

Innovation Solution

A seven-step automated KM digitizer process that converts a graphical KM plot into a 3D array, identifies axes and tick marks, segments colored pixels using K-means clustering, and applies regression to generate a digitized KM plot, ensuring accurate extraction of IPD.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual methods are used to extract independent patient data from Kaplan-Meier curves, then the process is simple to implement, but the extraction is labor-intensive and prone to errors

Engineering Contradiction:

Improveextraction precisionVSAvoidsystem complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent replaces manual mechanical extraction methods with an automated computer vision system that uses image processing, color segmentation, and regression analysis to extract data from Kaplan-Meier curves, eliminating human labor and associated errors while maintaining simplicity through algorithmic automation

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system enables self-service extraction by automatically processing KM curve images through a seven-step pipeline including color mask generation, axis identification, curve segmentation, and digitization, allowing the system to extract independent patient data without requiring manual intervention or specialized expertise

Inventive Principle:

Principle #25Self-service

2Productivity

If automated image processing is applied to extract data from KM curves, then extraction speed increases, but accuracy decreases for complex or intertwined curves

Engineering Contradiction:

Improveextraction speedVSAvoidextraction accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent applies local quality by generating specific color masks tailored to different curve segments and using localized regression analysis for each identified curve, allowing the system to maintain high accuracy for complex intertwined curves by treating each local region with specialized processing rather than applying a uniform approach

Inventive Principle:

Principle #3Local quality

Solution Approach 2:

The system segments the KM curve image into distinct colored pixel groups representing different curves, then processes each segment separately through identification and digitization steps, enabling accurate extraction from complex intertwined curves by isolating and analyzing each curve individually rather than processing the entire image as a whole

Inventive Principle:

Principle #1Segmentation

3Reliability

If conventional extraction methods are used, then the process is quick to implement, but the reliability and reproducibility of extracted data are compromised

Engineering Contradiction:

Improvedata reliabilityVSAvoidprocessing time

Core Design Contradiction:

ReliabilityVSLoss of time

Solution Approach 1:

The patent replaces unreliable manual extraction with a standardized automated system that consistently applies the same image processing and regression algorithms to all KM curves, ensuring high reliability and reproducibility while reducing processing time through efficient automated pipelines

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system incorporates feedback mechanisms by using identified axis positions and tick marks to guide subsequent curve identification and digitization steps, continuously refining its analysis based on previously extracted information to maintain high reliability across varying curve complexities

Inventive Principle:

Principle #23Feedback

Data Source

PatentUS20250308107A1Kaplan-Meier Digitizer for Automating Kaplan-Meier Curve Analysis

Publication Date: 2025.10.02 BRISTOL MYERS SQUIBB CO

AI summary

A method includes receiving an input image containing a graphical Kaplan-Meier (KM) plot and processing the input image to convert the graphical KM plot into a three-dimensional (3D) array. The method also includes processing the 3D array to generate a black pixel matrix mask and a colored pixel matrix mask, processing the black pixel matrix mask to identify pixel coordinates that define x- and y-axis of the graphical KM plot, cropping the colored pixel matrix mask based on the identified pixel coordinates, and processing the cropped colored pixel matrix to segment the colored pixels from the cropped colored pixel matrix mask into respective groups of clustered pixels. The method also includes processing each respective group of clustered pixels to generate a respective digitized representation of a corresponding KM curve and generating a digitized KM plot based on the respective digitized representation generated for each corresponding KM curve.