Dermoscopy Diagnosis Using Dual Deep Learning via Sonification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for diagnosing skin cancer, particularly malignant melanoma, face limitations in diagnostic accuracy due to the complexity of visual inputs in dermoscopy images and reliance on physician skill, leading to high false positive rates and unnecessary biopsies.
Innovation Solution
A system and method that acquire visual data from skin lesions, convert it into audio signals through sonification, and utilize deep learning algorithms to enhance diagnostic precision by combining visual and audio analysis for more accurate classification of cancerous and non-cancerous tissue.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If visual data from dermoscopy images is used for skin cancer diagnosis, then diagnostic information can be obtained, but diagnostic accuracy is limited due to complexity of visual inputs and dependency on physician skills
Solution Approach 1:
The patent replaces the mechanical/visual inspection system with an acoustic analysis system. Visual dermoscopy images are converted into audio signals through sonification, allowing diagnostic information to be processed through acoustic patterns rather than complex visual analysis. This substitution transforms the diagnostic approach from visual pattern recognition to acoustic signal analysis, thereby improving measurement precision while reducing the impact of visual input complexity.
Solution Approach 2:
The patent changes the parameter domain from visual spatial parameters to acoustic temporal parameters. By converting static visual image data into dynamic audio signals with varying frequency, amplitude, and temporal characteristics, the system transforms the nature of diagnostic parameters. This parameter transformation allows deep learning algorithms to process diagnostic information in a different domain, improving accuracy by capturing patterns not visible in the original visual data.
2Measurement precision
If deep learning algorithms are used to analyze visual dermoscopy images, then diagnostic performance can be improved, but interpretability of results remains limited
Solution Approach 1:
The patent introduces audio signals as an intermediary between the deep learning algorithm and the final diagnosis. The visual input is first converted to audio through sonification, then processed by deep learning algorithms. This intermediary transformation maintains the diagnostic performance improvement from deep learning while enhancing interpretability, as acoustic patterns are more intuitively understandable and analyzable than raw visual pixel data or abstract algorithmic outputs.
3Measurement precision
If sonification techniques are applied to convert visual data to audio signals, then diagnostic accuracy is increased, but system complexity increases
Solution Approach 1:
The patent replaces the need for complex visual analysis systems with a sonification-based acoustic analysis system. By converting visual dermoscopy data into audio signals, the system simplifies the diagnostic process while improving accuracy. The acoustic representation consolidates complex visual information into interpretable sound patterns, reducing the overall system complexity despite adding the sonification step.
Data Source
AI summary
The present invention provides for a system and method for the diagnosis of skin cancer, wherein visual data is acquired from a skin lesion by a passive or active optical, electronical, thermal or mechanical method, processing the acquired visual field image into a classifier, applying a dermoscopic classification analysis to the image and converting, by sonification techniques, the data to raw audio signals, wherein the raw audio signals are analyzed with a second machine learning algorithm to increase interpretability and the precision of diagnosis.