Automated Tumor Detection Using 3D Segmentation and Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Training machine-learning models for tumor detection in medical images is challenging due to the time-consuming manual labeling process and the confounding effect of large unknown masses, which can obscure relevant features, leading to increased analysis time or error rates.
Innovation Solution
A computer-implemented method using a series of image processing operations, including segmentation, registration, and classification with a trained machine-learning model, such as a support vector machine, to identify tumors in three-dimensional images, utilizing techniques like watershed transformation and U-Net models for preprocessing to isolate and classify image objects based on structural characteristics.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If manual labeling is used to train machine-learning models, then the models can be trained to identify tumors, but the process becomes time-consuming and requires many experts
Solution Approach 1:
The patent applies preliminary action by performing automated preprocessing of medical images before manual labeling. This includes generating three-dimensional representations from two-dimensional images, segmenting anatomical structures, and pre-labeling potential tumor regions. These preliminary steps reduce the time required for expert manual labeling while maintaining training accuracy.
2Measurement precision
If large unknown masses are present in training images, then they can be detected, but they obscure other relevant features and increase error rates
Solution Approach 1:
The patent applies segmentation by dividing the medical images into multiple three-dimensional representations corresponding to different anatomical structures or regions of interest. This segmentation isolates large unknown masses from other relevant features, allowing the machine-learning model to analyze each segment independently and reduce the obscuring effect while maintaining overall detection precision.
3Adaptability or versatility
If conventional machine-learning models are used to segment images, then tissue types can be identified, but the training process becomes difficult and time-consuming
Solution Approach 1:
The patent introduces three-dimensional representations as an intermediary between the original two-dimensional medical images and the machine-learning model. This intermediary layer simplifies the training process by providing pre-processed, structured data that highlights relevant anatomical features and reduces noise, making the model more adaptable to different tissue types while reducing training complexity.
Data Source
AI summary
Methods and systems disclosed herein relate generally to processing images to estimate whether at least part of a tumor is represented in the images. A computer-implemented method includes accessing an image of at least part of a biological structure of a particular subject, processing the image using a segmentation algorithm to extract a plurality of image objects depicted in the image, determining one or more structural characteristics associated with an image object of the plurality of image objects, processing the one or more structural characteristics using a trained machine-learning model to generate estimation data corresponding to an estimation of whether the image object corresponds to a lesion or tumor associated with the biological structure, and outputting the estimation data for the particular subject.