Tumor Detection via Image Segmentation and Merging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
High-resolution images pose a challenge for convolutional neural networks in tumor detection, as direct input leads to prolonged training times and reduced efficiency, resulting in loss of detail when resized, making it difficult for healthcare professionals to accurately locate and analyze tumors.
Innovation Solution
A tumor detecting device and method that segments high-resolution images into overlapping blocks matching the input size of the convolutional neural network, applies image augmentation, and merges the blocks into a single image for accurate tumor detection, using modules like image obtaining, segmentation, deletion, preprocessing, training, detection, and merging to enhance detection speed and accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution images are directly input to convolutional neural network for tumor detection, then detection accuracy is improved, but training time becomes longer and training efficiency is reduced
Solution Approach 1:
The patent divides high-resolution images into multiple smaller patches or blocks before feeding them to the convolutional neural network. This segmentation approach allows the network to process manageable image sizes while collectively covering the entire high-resolution image, thus reducing training time while maintaining detection accuracy through comprehensive coverage of all image regions.
2Productivity
If high-resolution images are resized to match input size requirements, then training efficiency is improved, but image details are lost
Solution Approach 1:
Instead of resizing the entire high-resolution image which would lose details, the patent segments the image into multiple smaller patches that retain the original high-resolution details. Each patch is then processed independently by the neural network, preserving fine details while achieving efficient training through smaller input sizes.
3Productivity
If high-resolution images are segmented into blocks for processing, then training efficiency is improved, but tumor location precision may be reduced
Solution Approach 1:
The patent introduces coordinate transformation and position mapping mechanisms as intermediaries between the segmented patches and the final detection results. These intermediaries track and map the spatial relationships between patches, enabling precise reconstruction of tumor locations in the original high-resolution image coordinate system, thus maintaining location precision while benefiting from efficient segmented processing.
Data Source
AI summary
A method for detecting a tumor from images which are required to be shrunken in resolution obtains one or more first images. Then, the method segments or divides the detection images into a number of detection image blocks according to an input size of training data of a convolutional neural network architecture, before segmenting, each of the plurality of detection image blocks comprising coordinate values. The detection image blocks are input into a preset tumor detection model to generate image blocks of a result of the detection images. The method merges the image blocks into a single image according to the coordinate values of each detection image block. Colors of normal areas, abnormal areas, and overlapping areas of the abnormal areas are all different. The method generates a final detection according to color depths in the image. A tumor detection device and a non-transitory storage medium are provided.