Hybrid Ensemble of Segmentation-Assisted Classification and GBDT for Skin Cancer Detection with Engineered Metadata and Synthetic Lesions from ISIC 2024 Non-Dermoscopic 3D-TBP Images
Muhammad Zubair Hasan, Fahmida Yasmin Rifat · 2025
Read Hybrid Ensemble of Segmentation-Assisted Classification and GBDT for Skin Cancer Detection with Engineered Metadata and Synthetic Lesions from ISIC 2024 Non-Dermoscopic 3D-TBP Images by Muhammad Zubair Hasan, Fahmida Yasmin Rifat (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-03420.txt.
Description
Muhammad Zubair Hasan and Fahmida Yasmin Rifat built a mixed AI system to spot skin cancer. Their 2025 paper uses a large set of lesion photos from full-body imaging taken under phone-like conditions. Early spotting of skin cancer saves lives, yet photo-based calls are hard. The dataset holds over four hundred thousand cropped lesion images.
Their pipeline first outlines each lesion, then classifies it with vision models. A boosted tree model joins the image scores with patient data and engineered features. Fake lesions made by a diffusion model balance the rare cancer cases. Extra relabeling aligns outside datasets into three classes. On the chosen partial accuracy metric, the mixed system scored best among all their setups.
Catalog record
| CREATOR | Muhammad Zubair Hasan, Fahmida Yasmin Rifat |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-03420