Multi-type skin diseases classification using OP-DNN based feature extraction approach.

Area of Science:

• Dermatology and Medical Imaging
• Artificial Intelligence and Machine Learning
• Computational Biology

Background:

• Accurate identification of skin diseases is challenging due to variations in skin tone, color, and hair presence in dermatological images.
• Existing research often struggles with precise skin disease prediction, necessitating advanced diagnostic tools.
• The complexity of visual diagnosis in dermatology highlights the need for automated and reliable systems.

Purpose of the Study:

• To propose a novel Optimal Probability-Based Deep Neural Network (OP-DNN) for accurate multi-type skin disease prediction.
• To enhance the diagnostic capabilities available to medical professionals for skin conditions.
• To address the limitations of current methods in classifying skin diseases from complex visual data.

Main Methods:

• Image pre-processing techniques were applied to clean and prepare dermatological image datasets.
• Features were extracted from pre-processed images and fed into the proposed OP-DNN for training.
• Whale optimization algorithm was employed to determine optimal weight values for the OP-DNN, minimizing training error.

Main Results:

• The OP-DNN model achieved a high accuracy rate of 95% in classifying skin diseases.
• The model demonstrated strong performance with a specificity of 0.97 and a sensitivity of 0.91.
• The proposed model outperformed previous methods in predicting multiple types of skin diseases.

Conclusions:

• The novel OP-DNN approach significantly improves the accuracy and reliability of skin disease diagnosis.
• The model's ability to predict diverse skin conditions offers substantial benefits for clinical practice.
• This AI-driven tool can assist physicians in making more certain diagnoses, potentially improving patient outcomes.