SkinViT-EfficientX: A Hybrid Vision Transformer Model with Token Pruning and Explainable AI for Multiclass Skin Cancer Diagnosis
Bibliographic record
Abstract
Skin cancer is a common and serious health issue, making early diagnosis crucial for better outcomes. Traditional manual dermoscopy can be slow and inconsistent, demonstrating a need for automated diagnostic tools. This study introduces SkinViT-EfficientX, a hybrid deep learning model specifically designed for classifying skin lesions. It utilizes an EfficientNetV2-S encoder and a lightweight Vision Transformer connected by a residual cross-attention mechanism for effective local-global feature extraction. To enhance performance, a confidence-guided token pruning strategy is employed, and Grad-CAM is used for class-specific visual explanations. The model underwent thorough preprocessing and augmentation on two benchmark datasets: HAM10000 and the combined ISIC 2019 + DermNet dataset. SkinViT-EfficientX achieved a 97.36% F1-Score, 95.64% MCC, and 97.93% Specificity on HAM10000, while scoring 98.42% F1-Score, 96.51% MCC, and 98.86% Specificity on the combined dataset. It outperformed top models like MaxViT, Swin V2-T, DeiT III-S, and MobileViT V2-S in all metrics. The model's robustness and stability for rare lesion classes were validated through confusion matrix and learning curve analyses. Further, it is integrated into a web application for dermoscopic image uploads, class predictions, and heatmap visualizations. SkinViT-EfficientX provides an efficient, accurate, and interpretable AI-driven solution for skin cancer screening.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".