Design and Evaluation of a Smartphone Application for Early Atopic Dermatitis Screening Using Large Language Model

Abstract

Atopic dermatitis is form of eczema. It is a chronic disease that results from pruritus, especially when the skin is dry. People who have atopic triad have skin barrier, upper respiratory, and lower respiratory defects. Most people, especially in rural areas, lack knowledge about this chronic inflammation. The objective of this study is to create a mobile app that utilizes LLMs in Expo Go for patient assessment to ascertain the level of damage that the skin barrier undergoes due to patients\' body parts. For contextually and conversationally responsive output based on users\' conditions and questions, LLMs integrate into the system using Lang Chain. The mobile application raises awareness among users through symptomatic guidance, preventive tips, and educational aspects.The end product will be an application that is user friendly and efficient for both the city and rural environment.. This will enable people to access information on this skin inflammation.

Introduction

Atopic Dermatitis (AD), a chronic inflammatory skin condition characterized by itching, redness, dryness, and irritation, which can significantly affect patients’ quality of life, including sleep and daily activities. It highlights the limitations of existing AI-based dermatology and mobile health apps, which mainly rely on image analysis and often ignore important contextual factors like itching severity, humidity, and temperature.

To address these gaps, the proposed system introduces an AI-powered mobile application for early detection of Atopic Dermatitis, developed using React Native (Expo) and Node.js. Unlike existing tools, it uses a multimodal approach, combining skin images with patient-reported data (such as itch level and condition duration) to improve diagnostic accuracy.

The literature review shows that while current healthcare systems using LLMs, deep learning, and mobile health apps improve accessibility and diagnostics, they often lack features such as multimodal integration, real-time mobile deployment, dermatology-specific focus, or lightweight usability for practical environments.

The methodology describes a client–server architecture consisting of three layers:

• A mobile interface layer for capturing images and patient symptoms
• A backend (Node.js) for secure data handling and processing
• A multimodal AI engine powered by an LLM, which analyzes both images and text to assess the likelihood and severity of AD

The AI pipeline includes image preprocessing, feature extraction, and integration of clinical inputs to generate predictions and preliminary recommendations, with a disclaimer that results are for screening only.

Conclusion

The system also ensures secure data transmission and provides a user-friendly interface for seamless interaction and report generation.The implementation shows that the system support early stage screening , severity estimation , and recommendation generation.

References

[1] V. Sakthivel, “Toward Inclusive Healthcare: An LLM-Based Multimodal Chatbot for Preliminary Diagnosis,” IEEE Access, vol. 13, 2025. [2] Rakeshkumar, “MedHub: LLM-Based Healthcare System,” IEEE Access, 2025. [3] F. R. Shafi, “A RAG-Based Multimodal Behavioral Therapist as Healthcare Assistant,” IEEE Access, 2025. [4] N. Tasnim, “AI-Powered Disease Detection using Multimodal Imaging and Explainable AI,” IEEE Access, 2025. [5] S. Yan, “Multimodal Vision Foundation Model for Clinical Dermatology,” 2025. [6] N. Vivek, “Deep Learning Multimodal Melanoma Detection Algorithm Development and Validation,” 2025. [7] Y. Hao, “Multimodal Integration in Healthcare: Development with Applications in Disease Management,” 2025. [8] Y. B. Sree, “Retrieval Augmented Generation Based Large Language Model for Improving Diagnosis for Physical and Mental Health,” IEEE Access, 2024. [9] Chengsu, “Hybrid RAG-Empowered Multimodal LLM for Secure Data Management in Internet of Medical Things,” IEEE Access, 2024. [10] K. Linuma, “Multiclass Classification of Pigmented Skin Lesions using Multimodal Large Language Model,” 2024. [11] Y. He, “Multimodal Bayesian Networks for Automatic Skin Disease Diagnosis,” 2024. [12] S. Mehta, “Multimodal Framework for Skin Lesion Classification using Deep Learning and Machine Learning Techniques,” 2024. [13] M. Effendie, “Multimodal AI Approach for Multiclass Skin Disease Classification,” 2024. [14] A. Bordia, “A Multimodal Approach for Skin Lesion Diagnosis,” 2024. [15] M. Narendra, “Advancing Skin Disease Diagnosis: A Multimodal Approach using Telegram API Chatbot for Text and Image Analysis,” 2024. [16] A. Dosovitskiy et al., “An Image is Worth 16×16 Words: Transformers for Image Recognition at Scale,” ICLR, 2021. [17] Y. Liu et al., “A Deep Learning System for Differential Diagnosis of Skin Diseases,” Nature Medicine, vol. 26, pp. 900–908, 2020. [18] M. Goyal et al., “AI-Based Image Classification Methods for Skin Cancer Diagnosis,” Computers in Biology and Medicine, vol. 127, 2020. [19] T. J. Brinker et al., “Deep Learning Outperformed 11 Pathologists in Melanoma Classification,” European Journal of Cancer, vol. 118, pp. 91–96, 2019. [20] P. Tschandl et al., “The HAM10000 Dataset: A Large Collection of Dermatoscopic Images,” Scientific Data, vol. 5, 2018. [21] A. Esteva et al., “Dermatologist-Level Classification of Skin Cancer with Deep Neural Networks,” Nature, vol. 542, pp. 115–118, 2017. [22] R. Rajpurkar et al., “CheXNet: Radiologist-Level Pneumonia Detection on Chest X-Rays,” arXiv, 2017. [23] R. R. Selvaraju et al., “Grad-CAM: Visual Explanations from Deep Networks,” ICCV, 2017. [24] K. He et al., “Deep Residual Learning for Image Recognition,” CVPR, 2016

Copyright

Copyright © 2026 Trishala AG, Dr. R. Ahila. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.