Abstract

The accurate diagnosis of skin conditions remains a significant challenge in dermatology, often constrained by variability in traditional diagnostic methods and limited access to specialist care. This study presents a novel AI-driven diagnostic system that integrates multiple dermatological datasets and employs a hybrid machine learning approach. Unlike prior works that rely on a single dataset, this research combines the SCIN dataset for metadata with medical-grade images from HAM10000 and ISIC to enhance diagnostic precision. Separate convolutional neural network (CNN) models were trained for image-based classification, while a metadata-driven machine learning model was developed for symptom-based predictions. These models were then fused into a unified ensemble, leveraging the strengths of both structured patient data and visual diagnostic cues. The resulting web-based application enables users to upload images and input metadata for real-time diagnostic feedback, bridging gaps in dermatological care, particularly for underserved populations. This study further explores critical challenges such as dataset bias, model interpretability, and ethical considerations in AI-driven diagnostics. By integrating multi-source data with an innovative ensemble approach, this research highlights the potential of AI to revolutionize dermatology, enhancing both diagnostic accuracy and accessibility in healthcare.

Advisor

Guarnera, Heather

Department

Computer Science

Disciplines

Artificial Intelligence and Robotics

Publication Date

2025

Degree Granted

Bachelor of Arts

Document Type

Senior Independent Study Thesis

Download

Share

COinS
 

© Copyright 2025 Habiba Hye