Pixel-Wise Cancer Certainty Heatmap via Overlapping Subimage Scoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for analyzing medical images for cancer detection often struggle to provide detailed insights into how individual pixels contribute to the overall score for cancer presence, as they typically compute scores for larger regions, masking the contributions of specific cells and structures.
Innovation Solution
The method generates a local scoring system by dividing images into overlapping subimages, allowing for pixel-level contribution analysis through color coding, where red indicates cancer evidence, blue indicates against cancer, and green denotes low evidence, and uses kernel matrices and bag of words features for classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If scores are computed for larger regions containing multiple cells, then robustness of the score is improved, but the contributions of individual cells and structures to the score cannot be seen
Solution Approach 1:
The image is divided into multiple overlapping subimages, each processed to generate local scores. This segmentation allows individual pixel contributions to be identified while maintaining robustness through the overlapping regions that provide multiple measurements for the same area.
Solution Approach 2:
The patent transitions from computing a single score for large regions to computing scores across multiple subimages with overlapping regions. This dimensional approach allows the system to track pixel contributions across different subimage contexts, preserving information while maintaining robustness.
2Measurement precision
If the image is divided into overlapping subimages for pixel-level analysis, then detailed visualization of pixel contributions is achieved, but the computational complexity increases
Solution Approach 1:
Dividing the large image into smaller overlapping subimages reduces the computational burden on each processing unit while maintaining overall accuracy. The segmentation allows parallel processing of multiple subimages, distributing computational complexity across multiple independent tasks.
Solution Approach 2:
The patent uses overlapping subimages where regions are processed multiple times in different subimages. This excessive action ensures that boundary pixels and transition zones are adequately captured, improving measurement precision at the cost of increased computational effort, which is justified by the enhanced pixel-level analysis capability.
3Productivity
If a single score is computed for each region, then the scoring process is simple and fast, but detailed insights into cancer detection are lost
Solution Approach 1:
By segmenting the image into overlapping subimages and computing scores for each, the system maintains relatively simple and fast scoring processes for individual subimages while accumulating detailed information across all subimages. The overlapping ensures comprehensive coverage without requiring complex post-processing.
Solution Approach 2:
The results from multiple subimage scores are merged to create a comprehensive view of pixel contributions. This combining approach preserves the simplicity of individual scoring operations while aggregating detailed insights across the entire image, achieving both speed and information retention.
Data Source
Figure 1
Figure 1A
Figure 1A
AI summary
Method for the automatic analysis of an image (1, 11, 12, 13) of a biological sample with respect to a pathological relevance, wherein fj local features of the image (1, 1.1, 12, 13) are aggregated to a global feature of the image (1, 11, 12, 13) using a bag of visual word approach, g) step a) is repeated at least two times using different methods resulting in at least two bag of word feature datasets,, h) computation of at least two similarity measures using the bag of word features obtained from a training image dataset and bag of word features from the image (1, 1 1, 1 2, 13) i) the image training dataset comprising a set of visual words, classifier parameters, including kernel weights and bag of word features from the training images, j) the computation of the at least two similarity measures is subject: to an adaptive computation of kernel normalization parameters and / or kernel width parameters, f) for each image (1, 11, 12, 13) one score is computed depending on the classifier parameters and kernel weights and the at least two similarity measures, the at least one score being a measure of the certainty of one pathological category compared to the image training dataset, g) for each pixel of the image (1, 11, 12, 13) a pixel-wise score is computed using the classifier parameters, the kernel weights, the at least two similarity measures, the bag of word features of the image (1, 11, 12, 13), all the local features used in the computation of the bag of word features of the image (1, 11, 12, 13) and the pixels used in the computations of the local features, h) the pixel-wise score is stored as a heatmap dataset linking the pixels of the image (1, 11, 12, 13) to the pixel-wise scores.