Hierarchical Neural Network for Real-Time ROI Detection and Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing medical image processing methods face challenges in achieving both high accuracy and real-time processing for region-of-interest detection and classification, particularly in models designed for real-time applications where reducing image resolution compromises detailed structure evaluation.
Innovation Solution
A hierarchical neural network approach that includes a feature extraction network, a detection subnetwork, and a classification subnetwork, where intermediate feature amounts are used for detection and higher resolution feature amounts are utilized for classification, allowing separate training of these components and avoiding redundant feature extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If image resolution is reduced for high-speed processing, then processing speed is improved, but classification accuracy deteriorates
Solution Approach 1:
The neural network is divided into multiple processing stages: a first processing stage performs detection on low-resolution feature amounts for speed, while a second processing stage performs classification on high-resolution feature amounts for accuracy. This segmentation allows each stage to operate at its optimal resolution level.
Solution Approach 2:
The system processes feature amounts at multiple resolution dimensions simultaneously. Instead of choosing a single resolution, the network maintains and processes both low-resolution and high-resolution feature representations, enabling fast detection and accurate classification to occur in parallel across different resolution dimensions.
2Measurement precision
If feature extraction is performed again from high-resolution input image, then classification accuracy is improved, but processing time increases
Solution Approach 1:
The feature extraction network pre-processes the input image to generate feature amounts at multiple resolution levels before the detection and classification stages. This preliminary action eliminates the need for repeated feature extraction later, as all required feature representations are already prepared.
Solution Approach 2:
The system merges multiple feature extraction operations into a single unified feature extraction network that simultaneously generates feature amounts at different resolution levels. This combining of operations prevents redundant processing and reduces overall computation time.
Data Source
AI summary
There are provided a medical image processing device, a hierarchical neural network, a medical image processing method, and a program capable of achieving highly accurate and real-time processable region-of-interest detection and class classification. A medical image processing device acquires a medical image, extracts a first feature amount and a second feature amount having a resolution relatively higher than a resolution of the first feature amount from the medical image by processing the medical image in a feature extraction network of a hierarchical neural network, detects a region of interest included in the medical image by processing the first feature amount in a first subnetwork of the hierarchical neural network, and classifies the region of interest by processing the second feature amount in a second subnetwork of the hierarchical neural network.