Pathological Cell Classification With Multi-Region Weighted Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for pathological diagnosis struggle to accurately classify specimen cells as benign or malignant in pathological images, necessitating improved methods to enhance classification accuracy.
Innovation Solution
A classification apparatus and method utilizing multiple generative models to emphasize different regions of interest in pathological images, combined with feature analysis models to generate and classify feature quantities, and a training apparatus to update model parameters based on classification results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single trained model is used for pathological image recognition, then the system is simple and fast, but the classification accuracy of specimen cells between benign and malignant is insufficient
Solution Approach 1:
The patent divides the single classification model into multiple specialized models, each responsible for analyzing specific regions of interest (e.g., nuclear region, cytoplasmic region). This segmentation allows each model to focus on particular features, thereby improving overall classification accuracy while maintaining manageable system complexity through modular architecture
Solution Approach 2:
The patent introduces a new dimension of analysis by generating multiple weighted input images with different emphasis weights for different regions. Instead of analyzing the entire image uniformly, the system creates multiple versions of the image where different regions are emphasized, allowing the classification models to capture diverse features from different spatial dimensions
2Measurement precision
If multiple generative models are used to emphasize different regions, then the classification accuracy improves, but the processing time and computational resources increase
Solution Approach 1:
The patent performs preliminary actions by pre-training multiple generative models and feature analysis models before actual classification tasks. The models are pre-trained on diverse pathological images with labeled regions, so that during actual classification, they can quickly generate weighted input images and extract features without requiring extensive real-time computation
Solution Approach 2:
The patent creates multiple copies of the input pathological image through the generative models, where each copy is a weighted version emphasizing different regions. These copies are then processed by specialized feature analysis models. This copying approach allows parallel processing of multiple regional features simultaneously, improving accuracy while managing processing time through efficient model architecture
Data Source
AI summary
In order to improve accuracy in classification of a specimen cell between benignancy and malignancy in pathological diagnosis, a classification apparatus (1) includes: an acquisition section (11) for acquiring a pathological image; and a classification section (12) for classifying a specimen cell as a benign cell or a malignant cell using a classification model that receives input of (i) a feature quantity of a first weighted input image which has been processed with first weighting information for emphasizing a first region of interest and (ii) a feature quantity of a second weighted input image which has been processed with second weighting information for emphasizing a second region of interest which differs from the first region of interest.