Artificial Intelligence Tools for Clinical Decision Support in Dermatology - PhD Thesis - Appendix Data
Description
Supplementary Resources for Artificial Intelligence Tools for Clinical Decision Support in Dermatology Description This repository contains the supplementary materials accompanying the PhD thesis "Artificial Intelligence Tools for Clinical Decision Support in Dermatology" submitted to Imperial College London. The resources support the development of the DermaSynth synthetic instruction dataset, the DermatoLlama multimodal vision-language model, and the DermaBench dermatology benchmark. They include instruction libraries, seed tasks, structured annotation templates, benchmark question sets, annotation guidelines, demonstration materials, and supplementary examples used throughout the research. Repository Contents Supplementary Movie 1: Running demonstration of DermatoLlama, illustrating example image-based interactions and generated responses. Supplementary Data 1: 190 seed tasks for instruction-following question-answer synthesis. Supplementary Data 2: 40 seed tasks for knowledge-based question-answer synthesis. Supplementary Data 3: 137 seed tasks for structured clinical case synthesis based on anamnesis. Supplementary Data 4: Disease list containing 137 dermatological diseases used for structured case synthesis. Supplementary Data 5: Anamnesis template consisting of 104 predefined clinical questions used during structured case generation. Supplementary Data 6: Expert-curated hierarchical annotation question set for the DERM12345 dataset used in DermaBench, comprising 93 root questions and 320 total questions. Supplementary Data 7: Annotation guidelines used during the construction of the DDI component of DermaBench, including annotation workflow, question definitions, response criteria, and annotator instructions. Supplementary Data 8: Expert-curated hierarchical annotation question set for the Diverse Dermatology Images (DDI) dataset used in DermaBench, provided in PDF format. Intended Use: These materials are provided to support the reproducibility of the research presented in the thesis. They may be used for developing, benchmarking, and evaluating dermatology-specific language models, vision-language models, instruction-tuning datasets, and medical question-answering systems. Related Publication: Abdurrahim Yılmaz. Artificial Intelligence Tools for Clinical Decision Support in Dermatology. PhD Thesis, Imperial College London. License: CC BY-NC 4.0