Copyright © 2026 Authors retain the copyright of this article. This article 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.

@article{210479,
        author = {Yuvraj Aute and Vaibhav Bobade and Ramnath Gadhe and Raviraj Avhad},
        title = {Agentic AI-Based Intelligent Dermatological Assessment and Skin Care Decision System},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {13},
        number = {6},
        pages = {778-795},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=210479},
        abstract = {Early and accurate assessment of skin lesions supports timely clinical care, yet access to dermatological expertise is limited and visual examination is subject to inter-observer variability. Most deep learning systems for skin lesion analysis stop at image classification and do not combine prediction reliability, risk-oriented triage and patient-facing guidance in one transparent workflow. This paper proposes an Agentic AI-based dermatological assessment and skin-care decision-support system built around a seven-agent architecture: image preprocessing, deep learning classification, confidence evaluation, risk assessment, skin-care recommendation, decision coordination and report generation. Skin lesion images are classified into the seven diagnostic categories of the HAM10000 dataset (10,015 dermatoscopic images) with an EfficientNet-B0 network initialised by transfer learning and followed by a softmax layer. The maximum softmax probability is treated as a model-confidence indicator, not as a calibrated measure of diagnostic certainty, and is combined with transparent predefined rules to recommend professional evaluation for uncertain or potentially concerning results. A coordination agent merges the agent outputs into a structured report containing the predicted category, confidence, risk level and general skin-care guidance. Only the classification agent is a learned component; the remaining agents are deterministic, rule-based software modules, and no large language model is assumed. The paper presents the design, algorithms, mathematical formulation and an evaluation framework with placeholders for measured results; no experimental performance is claimed. The system is intended for preliminary decision support and teledermatology triage under professional dermatological supervision.},
        keywords = {Agentic AI, Clinical decision support, Deep learning, EfficientNet-B0, HAM10000, multi-agent system, Skin lesion classification, Transfer learning},
        month = {November},
        }