Comprehensive Summary
The following study introduces a large-scale, longitudinal dataset combining tile images, corresponding dermoscopic images, and detailed metadata for the early detection of skin cancers. The dataset includes paired clinical and dermoscopic photos collected over time, thus enabling researchers to assess both the spatial and temporal changes in skin lesions. The following methodology addresses limitations of previous static image datasets, which have lacked longitudinal follow-up and detailed clinical metadata. Through rigorous imaging protocols and standardized metadata capture, the authors created a dataset that supports the development and validation of machine learning models. The following study highlights the dataset's potential for improving the classification accuracy of skin cancer detection algorithms by incorporating patient-level temporal information.
Outcomes and Implications
The dataset represents a significant advancement in AI dermatology research, as longitudinal data can enhance early diagnosis and monitoring of melanoma and non-melanoma skin cancer. By integrating dermoscopic, clinical, and metadata features, the dataset enables the creation of algorithms that better reflect real-world diagnostic workflows. The inclusion of repeated imaging over time provides a way for AI models to learn lesion evolution patterns, potentially leading ot earlier and more accurate identification of malignancies. This dataset could accelerate the validation of AI systems for teledermatology and support healthcare providers in remote screening and follow-up care. With open access under Scientific Data, the dataset lays the foundation for reproducible research and AI-driven decision-support systems that may reach clinical use within the next few years.