Deep Learning Mammography Analysis for Real-Time Test Recommendations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current mammography screening methods rely heavily on human radiologists, leading to inefficiencies and delays in detecting abnormalities due to superimposed tissues in 2D images, resulting in unnecessary further tests and psychological stress for patients.
Innovation Solution
A computer-aided method using deep learning, specifically convolutional neural networks, to analyze medical images in real-time, determining the need for additional tests such as CT scans, ultrasounds, or biopsies, without human input, by analyzing characteristics like tissue density and lesion detection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If human radiologists manually examine mammograms, then diagnostic accuracy can be achieved, but the process is slow and leads to delays in detecting abnormalities
Solution Approach 1:
The patent introduces an automated image analysis system as an intermediary between the mammogram and the radiologist. This system pre-processes and analyzes the mammogram images, highlighting suspicious areas and providing preliminary assessments, thereby reducing the time radiologists need to spend on initial screening while maintaining diagnostic accuracy through human oversight of the automated results.
Solution Approach 2:
The system performs preliminary analysis of mammograms automatically before radiologist review. By conducting initial detection, characterization, and prioritization of abnormalities through automated algorithms, the system prepares the images and findings in advance, allowing radiologists to focus their expertise on confirming and refining diagnoses, thus reducing overall detection time without sacrificing accuracy.
2Ease of operation
If radiologists examine superimposed breast tissues in 2D mammograms, then comprehensive analysis is attempted, but visibility of malignant abnormalities is reduced leading to errors
Solution Approach 1:
The patent employs three-dimensional breast tomosynthesis (3D mammography) which captures multiple X-ray images from different angles and reconstructs them into volumetric data. This dimensional transformation allows the system to virtually section through breast tissue, eliminating the superimposition problem inherent in 2D mammograms. Radiologists can then examine specific tissue layers independently, significantly improving the visibility and detection accuracy of malignant abnormalities while maintaining comprehensive analysis capability.
3Reliability
If further medical tests are requested after mammography screening, then diagnostic certainty is improved, but unnecessary procedures and patient stress increase
Solution Approach 1:
The automated analysis system applies partial action by selectively identifying and prioritizing only the most suspicious areas in mammograms for further investigation. Rather than recommending comprehensive follow-up for all cases, the system uses risk stratification to determine which lesions require immediate attention and which can be monitored, thereby reducing unnecessary procedures and patient stress while maintaining high diagnostic certainty for true positives through targeted follow-up.
Data Source
AI summary
The present invention relates to deep learning implementations for medical imaging. More particularly, the present invention relates to a method and system for indicating whether additional medical tests are required after analysing an initial medical screening, in substantially real-time.Aspects and/or embodiments seek to provide a method and system for recommending additional medical tests, in substantially real-time, based on analysing an initial medical scan, with the use of deep learning.