AI Toxicant Classification With Heatmaps for Early Poisoning Diagnosis

Solution Overview

Problem

The diagnostic efficiency of toxicant classes in poisoning cases is inadequate due to unclear toxic substances in early stages, necessitating improved preliminary differential diagnosis.

Innovation Solution

A classification method and device using a classification model to generate heatmaps and probabilities for toxicant classes, incorporating data pre-processing and fine-tuning, and employing deep learning models like BERT and GPT for accurate toxicant classification.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If preliminary differential diagnosis is made based on poisoning symptoms when toxic substance is unknown, then diagnostic process can proceed, but diagnostic efficiency is insufficient

Engineering Contradiction:

Improvediagnostic efficiencyVSAvoiddiagnostic accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent introduces an AI classification model as an intermediary tool between the symptoms and the final diagnosis. The model takes poisoning symptoms as input and outputs the probability of different toxicant classes, serving as a mediator that enhances the efficiency of preliminary differential diagnosis while maintaining diagnostic accuracy through visualized interpretation.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent replaces the traditional manual differential diagnosis process with an AI-based classification system. The machine learning model automatically processes symptoms and generates diagnostic probabilities, substituting the mechanical manual analysis process with an automated intelligent system that improves diagnostic efficiency.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If AI classification model is used to improve diagnostic efficiency, then productivity increases, but device complexity increases

Engineering Contradiction:

Improvediagnostic efficiencyVSAvoidsystem complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The patent introduces an AI classification model as an intermediary tool between the symptoms and the final diagnosis. The model takes poisoning symptoms as input and outputs the probability of different toxicant classes, serving as a mediator that enhances the efficiency of preliminary differential diagnosis while maintaining diagnostic accuracy through visualized interpretation.

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent uses heatmap visualization to create a simplified representation or 'copy' of the complex internal decision-making process of the AI model. The heatmap shows the influence of each token on the classification result, providing an interpretable copy of the model's reasoning that reduces the perceived complexity for users.

Inventive Principle:

Principle #26Copying

Data Source

PatentUS12423387B2Classification method and classification device thereof

Publication Date: 2025.09.23 WISTRON CORP

AI summary

A classification device and a classification method for the classification device are disclosed to improve efficiency to find out a toxicant class. The classification method includes obtaining first data, and generating at least one heatmap and at least one probability of at least one toxicant class according to a classification model by using the first data. Any of the at least one toxicant class corresponds to one of the at least one probability. Any of the at least one heatmap is used to visualize influence of each of a plurality of tokens of the first data on attributing cause of poisoning to one or more of the at least one toxicant class.