Dermato

22 pages, 1891 KB

Open AccessArticle

Systematic Failure of Vision Transformers in Imbalanced Skin Lesion Classification

by Serra Aksoy, Pinar Demircioglu and Ismail Bogrekci

Dermato 2026, 6(2), 22; https://doi.org/10.3390/dermato6020022 - 11 Jun 2026

Viewed by 148

Background/Objectives: Vision Transformers (ViTs) have demonstrated impressive performance in dealing with large-scale natural image datasets. They have started to be used in medical image classification problems as well. However, how they behave under real-world conditions, such as data scarcity and extreme class imbalance, has not been well investigated. In this study, we examine the feasibility of using a standard Vision Transformer Base model that learned from scratch how to classify skin lesion images into multiple classes using the ISIC 2019 dataset. Methods: The Vision Transformer architecture was trained from scratch using stratified splitting of the data, class-balanced cross-entropy loss, multi-seed initialization, and control of hyperparameters such as patch size and dropout rate. The evaluation of the Vision Transformer architecture was performed using a hold-out test set with metrics such as accuracy, macro-F1, weighted-F1, and analysis of the confusion matrix. Results: Across all configurations, the training exhibited substantial instability and consistent overfitting behavior, with an average accuracy gap between validation and test sets of 22.7%. Test accuracy ranged from 8.0% to 37.8%, showing high sensitivity to initialization. For minority classes, the F1-score remained very low (F1 < 0.05) even though the classes were balanced in the loss function. Conclusions: The results indicate that a standard ViT-Base model trained from scratch can exhibit pronounced instability and a tendency toward majority-class bias when applied to multi-class skin lesion classification under conditions of extreme class imbalance and data scarcity. The findings point to the limitations of using simple transformer models without pre-training or other forms of inductive bias in scarce data settings. Full article

(This article belongs to the Special Issue Melanoma: Updates and Path Forward)

► Show Figures

Figure 1

9 pages, 1056 KB

Open AccessReview

Melanoma: Updates and Future Perspectives in Surgical Management

by José Maria Zepeda Torres and Valeria Contreras Oceguera

Dermato 2026, 6(2), 14; https://doi.org/10.3390/dermato6020014 - 10 Apr 2026

Viewed by 648

Abstract

Surgery continues to represent the central curative modality for melanoma despite major advances in systemic immunotherapy and targeted treatments. Contemporary surgical strategies aim to maintain oncologic safety while minimizing functional and aesthetic morbidity through optimized excision margins, highly selective use of sentinel lymph node biopsy (SLNB), and the omission of routine completion lymph node dissection (CLND). Rapid integration of neoadjuvant and adjuvant immunotherapies has begun to redefine surgical indications, timing, and extent—particularly for intermediate-stage and locoregionally advanced disease. Parallel innovations in Mohs micrographic surgery, reconstructive flap design, lymphatic reconstruction, and minimally invasive techniques further broaden the possibilities for individualized intervention. This expanded review synthesizes current evidence, ongoing controversies, and emerging trends that are shaping the future of melanoma surgery, highlighting how precision oncology, immunologic profiling, and technological advances are transforming the surgeon’s role and enabling more tailored, less invasive, and outcome-focused management. Full article

(This article belongs to the Special Issue Melanoma: Updates and Path Forward)

► Show Figures

Figure 1

Journal Menu

Journal Browser

Melanoma: Updates and Path Forward

Share This Special Issue

Editors

Special Issue Information

Keywords

Benefits of Publishing in a Special Issue

Published Papers (2 papers)

Research

Review

Further Information

Guidelines

MDPI Initiatives

Follow MDPI