Breast Slice Image Processing With Neural Networks for Lesion Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional mammography techniques produce 2D images that obscure breast tissue overlap, leading to false positives and negatives due to clustered signals from above and below pathology, making early detection of breast diseases challenging.
Innovation Solution
Utilizing digital breast tomosynthesis (DBT) data and artificial neural networks (ANNs) to process slice images, grouping them into representative images, and analyzing these through 2D or 3D CNNs to enhance detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If conventional 2D mammography is used, then imaging simplicity is maintained, but tissue overlap and structure noise increase making detection difficult
Solution Approach 1:
The patent transitions from 2D mammography to 3D digital breast tomosynthesis, reconstructing multiple slice images from projection data captured at different angles. This dimensional change allows separation of overlapping tissues along the depth axis, eliminating structure noise while maintaining imaging capability.
Solution Approach 2:
The patent divides the breast into multiple tissue slices through 3D reconstruction, separating overlapping structures that appear clustered in 2D images. This segmentation enables clear distinction between normal tissues and lesions, improving detection accuracy by eliminating tissue overlap.
2Measurement precision
If 3D slice images are processed individually, then detection accuracy improves, but processing time and computational load increase
Solution Approach 1:
The patent combines multiple 3D slice images into a single representative image by merging corresponding regions across slices. This consolidation maintains the diagnostic information from all slices while reducing the total data volume, enabling faster processing without sacrificing detection accuracy.
Solution Approach 2:
The representative image serves multiple functions: it encapsulates diagnostic information from all slice images, maintains lesion detectability, and enables efficient processing. This multi-functional approach eliminates the need to process each slice separately while preserving detection accuracy.
3Reliability
If all slice images are processed through the neural network, then comprehensive analysis is achieved, but computational resources are wasted on irrelevant images
Solution Approach 1:
The patent extracts and processes only the most relevant diagnostic information by creating a single representative image that captures essential features from all slices. This extraction eliminates redundant computational processing of individual slices while maintaining comprehensive analysis capability.
Solution Approach 2:
The patent discards redundant slice image data after extracting essential diagnostic information into the representative image. This approach recovers computational efficiency by avoiding repeated processing of the same information while preserving complete diagnostic analysis.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Improves diagnostic accuracy by clearly distinguishing normal breast tissues from lesions, reducing false positives and negatives, and efficiently processing large DBT datasets.
Implementation Method 1
The plurality of slice images may be reconstructed based on x-ray images of the breast captured from different angles
Data Source
AI summary
Digital breast tomosynthesis (DBT) may provide richer information than full-field digital mammography (FFDM). DBT data such as DBT slices may be processed based on deep learning techniques such as using a neural network, and the DBT slices may be divided into groups and a pre-determined number of representative images may be derived based on the grouping. The neural network may be configured to process the representative images to predict the presence or non-presence of a breast disease such as breast cancer.