Neural Network Object Detection in Medical Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Medical imaging technologies face challenges in accurately detecting and diagnosing lesions due to the quality of scanned images and human error, leading to potential missed or false diagnoses.
Innovation Solution
A method utilizing a deep-learning neural network (DNN) to partition medical images into sub-regions, detect objects within these regions, and determine a region of interest (ROI) with a confidence score, which is then labeled with annotations related to tissue types, enhancing the accuracy of lesion detection and providing insights to medical practitioners.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep-learning neural networks are used to detect objects in medical images, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The medical image is divided into multiple sub-regions, and the neural network processes each sub-region independently to detect parts of the object. This segmentation approach improves measurement precision by focusing computational resources on specific areas while managing device complexity through modular processing.
Solution Approach 2:
Bounding boxes are introduced as an intermediary element to represent detected object parts within sub-regions. These bounding boxes serve as mediators between the neural network's detection output and the final region of interest determination, improving precision while maintaining manageable system complexity.
2Measurement precision
If multiple sub-regions are analyzed to determine region of interest, then measurement precision is improved, but loss of time increases
Solution Approach 1:
The system analyzes only a selection of sub-regions rather than processing every possible region, applying partial action to achieve sufficient precision while reducing processing time. The bounding box confidence scores enable the system to focus on the most promising regions.
Solution Approach 2:
Bounding boxes with confidence scores above a threshold are identified in advance before determining the final region of interest. This preliminary action filters out low-probability regions, improving precision for the final determination while reducing the time required to evaluate all possible regions.
3Measurement precision
If bounding boxes with confidence scores are used to determine region of interest, then measurement precision is improved, but device complexity increases
Solution Approach 1:
A confidence score parameter is introduced to quantify the likelihood of object parts within bounding boxes. By changing the evaluation criterion from binary detection to probabilistic scoring, the system improves measurement precision while managing complexity through a straightforward threshold-based filtering mechanism.
4Productivity
If neural network partitioning is applied to medical images, then productivity is improved, but device complexity increases
Solution Approach 1:
The medical image is partitioned into multiple sub-regions that can be processed independently and in parallel by the neural network. This segmentation improves productivity by enabling concurrent processing while managing device complexity through a modular architectural approach where each sub-region is handled by the same processing pipeline.
Data Source
AI summary
A device to detect an object in a medical image is described. An image analysis application, executed by the device, receives the medical image as an input. The medical image is next partitioned to sub-regions. Parts of the object are detected in a selection of the sub-regions using a deep-learning neural network (DNN) model. Bounding boxes for the selection are also determined. The bounding boxes are evaluated based on a confidence score detected as above a threshold level. The confidence score designates the parts as contained within the selection. Next, a region of interest (ROI) is determined as a group including the selection. Similar orientations associated with the bounding boxes are comparable to similar orientations of a positive training model of the DNN model. Furthermore, the selection is designated as the ROI within the medical image. The medical image is provided with the ROI to a user.