Automated Lesion Segmentation for RECIST Assessment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for analyzing CT scans to assess tumour progression using the RECIST protocol are inefficient and prone to human error, as they require manual identification and measurement of lesions across multiple scans, which can lead to inaccurate assessments and failure to detect changes in tumour burden.
Innovation Solution
An automated method using a Fully Convolutional Neural Network (FCN) and Conditional Random Field for segmenting and measuring lesions in medical images, creating 3D volumetric masks and models, and performing statistical change analyses across scans, allowing for accurate and consistent identification and measurement of tumour changes without human intervention.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual identification and measurement of lesions is performed by radiologists, then human judgment and flexibility are applied, but time consumption and human error increase significantly
Solution Approach 1:
The patent replaces the manual mechanical process of radiologist measurement with an automated computer-based system that uses image processing algorithms to automatically identify, segment, and measure lesions across multiple CT scan slices, eliminating human time consumption while maintaining measurement accuracy
Solution Approach 2:
The system enables self-service automation where the computer automatically performs lesion identification, boundary detection, and volumetric measurement without requiring radiologist intervention for each measurement, allowing the system to serve itself in completing the assessment workflow
2Measurement precision
If radiologists manually assess tumour progression by comparing scans at different time points, then treatment decisions can be made, but inaccuracies occur due to inability to consistently identify the same lesions
Solution Approach 1:
The system implements feedback mechanisms by automatically tracking and comparing lesion characteristics across multiple scans, providing consistent reference measurements that feedback into the assessment process to ensure the same lesions are identified and measured accurately at different time points
Solution Approach 2:
The system creates digital copies and models of lesions with precise spatial coordinates and volumetric data, allowing exact replication and comparison of lesion measurements across different scans to ensure consistent identification and tracking of the same lesions over time
3Ease of manufacture
If only 1D measurements of lesions in 2D slices are performed, then simple measurement procedures are used, but tumour burden assessment is inherently limited in accuracy
Solution Approach 1:
The patent transitions from 1D linear measurements in 2D slices to 3D volumetric measurements by processing multiple contiguous CT scan slices, calculating the volume of lesions through integration across the third dimension, thereby providing accurate tumour burden assessment while maintaining procedural simplicity through automation
4Quantity of substance
If CT scans are presented in grayscale with single colour channel, then file size is reduced, but radiologists cannot visualise all information simultaneously without switching between multiple windows
Solution Approach 1:
The system merges multiple grayscale images representing different tissue windows into a single composite colour image, assigning different grayscale images to different colour channels (RGB), allowing radiologists to visualize multiple tissue types and lesion characteristics simultaneously in one view without switching between separate windows
Data Source
AI summary
The present invention relates to a method and system that automatically finds, segments and measures lesions in medical images following the Response Evaluation Criteria In Solid Tumours (RECIST) protocol. More particularly, the present invention produces an augmented version of an input computed tomography (CT) scan with an added image mask for the segmentations, 3D volumetric masks and models, measurements in 2D and 3D and statistical change analyses across scans taken at different time points.According to a first aspect, there is provided a method for determining volumetric properties of one or more lesions in medical images comprising the following steps: receiving image data; determining one or more locations of one or more lesions in the image data; creating an image segmentation (i.e. mask or contour) comprising the determined one or more locations of the one or more lesions in the image data and using the image segmentation to determine a volumetric property of the lesion.