Article Per Year (5 Year)
p-Index From 2021 - 2026
0.23
P-Index
This Author published in this journals
All Journal Bulletin of Electrical Engineering and Informatics
Krupali Rupesh Dhawale
Yeshwantrao Chavan College of Engineering (YCCE)
Electrical & Electronics Engineering

Published : 1 Documents Claim Missing Document
Claim Missing Document
Check
Articles

Found 1 Documents
Search

 1 
Papulosquamous SkinSense: a hybrid artificial intelligence model with visual explanations and chatbot Krupali Rupesh Dhawale; Arvind R. Bhagat Patil
Bulletin of Electrical Engineering and Informatics Vol 15, No 1: February 2026
Publisher : Institute of Advanced Engineering and Science

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.11591/eei.v15i1.10656

Abstract

Accurate diagnosis of papulosquamous skin diseases such as Psoriasis (PS), Lichen Planus (LP), Pityriasis Rosea (PR), and Pityriasis Rubra Pilaris(PRP) is challenging due to their visually similar features, particularly during healing stages. This study presents an optimized deep learning framework to improve diagnostic accuracy and interpretability. A dataset of 3,120 DermsNetZ images was used, with preprocessing through contrast limited adaptive histogram equalization(CLAHE) to enhance lesion visibility. Pretrained convolutional neural networks (CNNs) including MobileNetV2, InceptionV3, NASNet, and hybrid models were evaluated using accuracy, precision, recall, and F1-score. Among these, MobileNetV2 combined with gradient-class activation mapping (Grad-CAM) achieved the best results, delivering 94% accuracy, 95% precision, and strong F1-scores, while offering explainable artificial intelligence (AI) through lesion localization. To translate these results into practice, the SkinSense Detection App was developed, integrating with transfer learning, class balancing, augmentation, and Grad-CAM visualization within a user-friendly interface. The app also incorporates a large language model (LLM-powered) chatbot for real-time, personalized feedback. With an overall success rate of 98.08% and user ratings between 4.6–4.8/5, the system demonstrates high reliability and accessibility. This study highlights the value of interpretable deep learning in dermatology, bridging technical accuracy with clinical usability and offering scope for expansion to larger datasets and diverse skin conditions.
Co-Authors Arvind R. Bhagat Patil