Comparative Analysis of RNN and Deep Learning in Psoriasis and Vitiligo Detection with LLM for Text Generation

This research discusses various deep learning models for detecting skin diseases, such as vitiligo and psoriasis, using skin images. For Vitiligo Detection, we use ResNet50, ResNetRS50, ResNet101, EfficientNetB0, EfficientNetB6 and EfficientNetV2. We found that EfficientNet models provide high accuracy and less computational power, and the ResNet models are reliable and perform well across diverse skin images. For the psoriasis skin detection, we use CNN-based models such as MobileNet, which are effective in identifying fine patterns and textures, while LSTM is used in hybrid models for sequential analysis, and it gives us accurate results, and these models are lightweight and efficient. This paper also presents an image-based detection system that uses a large language model, DistilGPT-2. This model automatically generates medical reports, summaries, and predictions. This means our system not only detects the disease but also gives the result in text. By combining image detection and natural language processing, we build a single system that gives the result faster with accuracy, and it also assists clinicians in interpreting results more effectively. The proposed system demonstrates improved accuracy, efficiency, and robustness for automated skin disease detection in real-world scenarios. Keywords—Vitiligo detection, Psoriasis detection, Deep Learning, Convolutional Neural Networks (CNN), ResNet, EfficientNet, RNN, LSTM, MobileNet, DistilGPT2, LLM, Artificial Intelligence.