Interpretable Cell Cycle Phase Learning With CNN and Feature Contributions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing deep learning models for determining cell cycle phases in images, such as those using CNNs, provide accurate results but lack explainability, making it difficult for users to understand the basis for their decisions.
Innovation Solution
A learning apparatus that combines a CNN for image processing with a neural network for feature values, using LIME for linear regression in a readable feature space to visualize the contributions of both image and feature values, enhancing explainability while maintaining accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a deep learning model (CNN) is used for cell cycle phase determination, then determination accuracy is improved, but explainability deteriorates
Solution Approach 1:
The patent introduces an explanation model as an intermediary component that bridges the deep learning model and the user. This explanation model receives inputs (image and feature values) and outputs explanations in terms of contribution degrees, making the black-box deep learning model interpretable without sacrificing its accuracy
Solution Approach 2:
The patent segments the determination process into two independent parts: the deep learning model for accurate determination and the explanation model for interpretability. By separating these functions, the system can maintain high accuracy while providing clear explanations through the explanation model's analysis of feature contributions
2Measurement precision
If only image data is used for determination, then determination accuracy is improved, but ease of operation deteriorates
Solution Approach 1:
The patent makes the determination system universal by accepting multiple types of inputs (image data and feature values) and processing them through a unified deep learning model. This multi-functional approach allows the system to maintain high accuracy while improving ease of operation, as users can input data in various formats and the system handles them consistently
Data Source
AI summary
A learning apparatus includes: a storage unit which stores a learning model trained by setting, as an input, a training image set and a training feature value set related to a subject of the training image set and obtained by quantifying a predetermined interpretable feature, and by setting, as an output, results of a determination on the training image set and the training feature value set; a determination unit which outputs, by using the learning model stored in the storage unit, results of a determination on a target image and a first feature value related to a subject of the target image and obtained by quantifying the predetermined interpretable feature; and an explanation output unit which outputs degrees of contribution of the target image and the first feature value, for the result of the determination on the target image and the first feature value by the learning model.