A Combination of DNN and BN for Automatic Skin Disease Diagnosis

Artificial intelligence aims to mimic human intelligence by giving machines the ability to perceive and think. However, current AI is on the horns of a dilemma. Tools like deep learning are good at finding similarities but cannot make reasonable inferences like humans. Their inability to integrate w...

Full description

Saved in:
Bibliographic Details
Main Authors: He, Yunqi, Cai, Linglong, Cui, Taimei, Li, You, Zhou, Hai
Format: Conference Proceeding
Language:English
Subjects:
Artificial neural networks
Bayes methods
Deep learning
Knowledge engineering
Skin
Training
Training data
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Artificial intelligence aims to mimic human intelligence by giving machines the ability to perceive and think. However, current AI is on the horns of a dilemma. Tools like deep learning are good at finding similarities but cannot make reasonable inferences like humans. Their inability to integrate with human knowledge also requires large amounts of input data for tuning. In this paper, we propose an architecture that combines deep neural networks as nodes into Bayesian networks, which combines human knowledge with the perceptual results of deep learning tools. Using Bayesian networks for inference provides good interpretability and acceptable training data requirements. This architecture can correlate symptoms, demographic information, and deep neural networks in dermatological diagnosis. We conducted experiments on the ISIC 2019: Training dataset. With the help of dermatologists' expert knowledge, the architecture achieves an overall diagnostic accuracy of 88.1%, which is 23.6% higher than the pure deep learning approach.
ISSN:1945-8452
DOI:10.1109/ISBI53787.2023.10230768