What question did this study set out to answer?

To improve melanoma diagnosis accuracy using the HEVA architecture integrating various models and clinical features.

March 14, 20260 citationsOpen Access

HEVA: Integrating ViT and ResNet50 with ABCD Clinical Features for Skin Cancer Diagnosis

ACAntônio Fernando CrepaldiUniversidade Estadual Paulista (Unesp)BABianca Aparecida AndradePontifícia Universidade Católica de Campinas AAAdemar Takeo AkabanePontifícia Universidade Católica de Campinas

Key Points

To improve melanoma diagnosis accuracy using the HEVA architecture integrating various models and clinical features.
Developed the HEVA architecture combining ViT, ResNet50, and SegFormer-B5 with ABCD descriptors.
Evaluated the model on HAM-10000 and ISIC 2018 datasets.
Calculated performance metrics including accuracy, F1-score, and mean IoU.
HEVA achieved 93.2% accuracy and 84.2% F1-score on HAM-10000.
On ISIC 2018, HEVA reached 87.7% accuracy with 82.01% macro F1.
SegFormer achieved 82.6% mean IoU and 91.2% accuracy in segmentation.

Abstract

Melanoma is the deadliest skin cancer, causing over 90% of skin tumor deaths. Early detection ensures survival above 95%, but below 15% in late stages. We propose the HEVA architecture, which integrates ensemble deep learning with ABCD descriptors and a SegFormer-B5 model, for accurate lesion masks. HEVA achieved 93.2% accuracy and 84.2% F1-score on HAM-10000, and 87.7% accuracy with 82.01% macro F1 on ISIC 2018 (1,512 images). SegFormer achieved 82.6% mean IoU and 91.2% accuracy in segmentation.

AI से पूछें

Bookmark

View Full Paper