Scribble-Based Medical Image Segmentation Using Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical image segmentation technologies face challenges in accurately identifying and outlining target elements in medical imaging, often resulting in improper or blurry shapes, which can hinder diagnosis.
Innovation Solution
A system and method utilizing a machine learning model that receives medical images, allows user scribbles to correct segmentation errors, computes a loss function to adjust weights, and generates optimized images with precise outlines of target elements.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If existing medical image segmentation techniques are used, then automated segmentation can be achieved, but the segmentation accuracy and outline precision deteriorate resulting in improper or blurry shapes
Solution Approach 1:
The system implements feedback by allowing radiologists to provide corrections on automatically segmented images. These corrections are fed back into the training dataset, enabling the machine learning model to iteratively improve its segmentation accuracy while maintaining automated operation.
Solution Approach 2:
The system performs preliminary automated segmentation before radiologist review. This preliminary action provides a starting point that reduces the radiologist's workload while ensuring initial automated processing, with subsequent refinements made based on expert feedback.
2Manufacturing precision
If manual segmentation by radiologists is performed, then segmentation accuracy improves, but time consumption and human effort increase
Solution Approach 1:
The system applies partial manual action by having radiologists review and correct only the automatically segmented images rather than performing complete manual segmentation. This partial intervention significantly reduces time consumption while maintaining high accuracy through targeted expert corrections.
Solution Approach 2:
The machine learning model acts as an intermediary between fully automated segmentation and complete manual segmentation. It handles the initial segmentation task, reducing radiologist workload, while allowing expert intervention when needed, thus optimizing the balance between speed and accuracy.
3Manufacturing precision
If deep learning models are trained with extensive data, then segmentation precision improves, but training time and computational resources increase
Solution Approach 1:
The system performs preliminary data collection by gathering radiologist corrections as they naturally occur during clinical workflow. This preliminary accumulation of labeled data continues over time without requiring dedicated training periods, enabling progressive model improvement without interrupting clinical operations.
Solution Approach 2:
The system enables continuous model training by continuously collecting labeled data from radiologist corrections in the clinical environment. This continuous action allows the model to progressively improve precision without requiring periodic shutdowns for batch training, optimizing both precision and time efficiency.
Data Source
AI summary
A system and method generating an optimized medical image using a machine learning model are provided. The method includes (i) receiving one or more medical images, (ii) segmenting to generate a transformed medical image for detecting a plurality of target elements, (iii) displaying the transformed medical image, (iv) receiving markings and scribblings associated with scribble locations from a user, (v) identifying errors associated with an outline of a target element, (vi) computing a loss function for a location of pixels where the target element is located on the transformed medical image, (vii) modifying the pre-defined weights (w) to match the segmentation output and the determined target element, (viii) determining whether the segmentation output is matched with the target element and (ix) generating the optimized medical image if the segmentation output is matched with the determined target element.