Self-supervised pretraining for dermoscopic lesion segmentation with limited labels: A controlled comparison of initialisation strategies
- 1University of Cambridge
Abstract
BackgroundSelf-supervised learning may improve segmentation when labelled data are limited, but its benefit may depend on model and initialisation strategy. This study evaluated in-domain masked-reconstruction pretraining across limited annotation budgets.
MethodsThe International Skin Imaging Collaboration (ISIC) 2016 development set was partitioned into 765 images for self-supervised pretraining, 94 for fine-tuning and 41 for validation. Five nested annotation budgets (12-94 images) and five seeds were evaluated on the 379-image test set. A custom U-Net compared random and in-domain self-supervised initialisation; a ResNet-34 U-Net compared random, ImageNet, in-domain self-supervised and sequential ImageNet-to-self-supervised initialisation. Models trained with 94 labelled images were evaluated without adaptation on ISIC 2017. A separate experiment examined restoration-based pretraining.
ResultsIn the custom U-Net, self-supervised pretraining achieved higher mean Dice than random initialisation at all five budgets, including 0.664 versus 0.615 with 12 labels and 0.811 versus 0.768 with 94 labels. Seed-level 95% confidence intervals for paired differences excluded zero at 47 and 94 labels but not at the other budgets. With 47 labels, the self-supervised model achieved mean Dice of 0.789, compared with 0.768 for random initialisation with 94 labels. In ResNet-34, ImageNet initialisation achieved higher mean Dice than in-domain self-supervised pretraining at all five budgets, and adding self-supervised pretraining after ImageNet reduced mean Dice at four of five budgets. All six 94-label configurations achieved lower mean Dice on ISIC 2017 than ISIC 2016. In the restoration experiment, mean Dice was 0.686 with random initialisation and 0.638-0.642 following pretraining.
ConclusionsIn-domain masked-reconstruction pretraining improved mean Dice relative to random initialisation in the custom U-Net, but not consistently in the ResNet-34 U-Net, where ImageNet initialisation performed better at every budget. Adding self-supervised pretraining after ImageNet provided no consistent gain. These findings are specific to the dermoscopic lesion-segmentation experiments and do not establish melanoma-specific diagnostic or clinical performance.
Related
- A novel morphological assessment approach for abdominal aortic aneurysms: automated segmentation and measurement using a semi-supervised deep learning algorithm
- HSV-Net: hybrid state-space and self-supervised vision fusion with progressive contour learning for CT pulmonary nodule segmentation.
- Comparative effects of supervised aerobic and combined training with dietary counseling on metabolic biomarkers in sedentary adult male bank employees: a randomized controlled trial
- Effects of time-restricted eating and supervised exercise on intrahepatic fat and cardiometabolic health in adults with obesity: the TEMPUS randomized controlled trial
- Six-month supervised telerehabilitation combining pilates-based therapeutic exercise and self-myofascial release in adults with fibromyalgia: a randomized controlled trial.
BibTeX
@misc{hamda2026selfsupervised,
title = {{Self-supervised pretraining for dermoscopic lesion segmentation with limited labels: A controlled comparison of initialisation strategies}},
author = {H. K. H. Hamda and S. Mustafa and G. Khalili Moghaddam},
howpublished = {medRxiv},
year = {2026},
month = oct,
doi = {10.64898/2026.10.14.26365443},
url = {https://www.medrxiv.org/content/10.64898/2026.10.14.26365443v1},
note = {CC BY},
}
RIS
TY - UNPB
TI - Self-supervised pretraining for dermoscopic lesion segmentation with limited labels: A controlled comparison of initialisation strategies
AU - H. K. H. Hamda
AU - S. Mustafa
AU - G. Khalili Moghaddam
PY - 2026
DA - 2026/10/14
PB - medRxiv
DO - 10.64898/2026.10.14.26365443
UR - https://www.medrxiv.org/content/10.64898/2026.10.14.26365443v1
AB - BackgroundSelf-supervised learning may improve segmentation when labelled data are limited, but its benefit may depend on model and initialisation strategy. This study evaluated in-domain masked-reconstruction pretraining across limited annotation budgets. MethodsThe International Skin Imaging Collaboration (ISIC) 2016 development set was partitioned into 765 images for self-supervised pretraining, 94 for fine-tuning and 41 for validation. Five nested annotation budgets (12-94 images) and five seeds were evaluated on the 379-image test set. A custom U-Net compared random and in-domain self-supervised initialisation; a ResNet-34 U-Net compared random, ImageNet, in-domain self-supervised and sequential ImageNet-to-self-supervised initialisation. Models trained with 94 labelled images were evaluated without adaptation on ISIC 2017. A separate experiment examined restoration-based pretraining. ResultsIn the custom U-Net, self-supervised pretraining achieved higher mean Dice than random initialisation at all five budgets, including 0.664 versus 0.615 with 12 labels and 0.811 versus 0.768 with 94 labels. Seed-level 95% confidence intervals for paired differences excluded zero at 47 and 94 labels but not at the other budgets. With 47 labels, the self-supervised model achieved mean Dice of 0.789, compared with 0.768 for random initialisation with 94 labels. In ResNet-34, ImageNet initialisation achieved higher mean Dice than in-domain self-supervised pretraining at all five budgets, and adding self-supervised pretraining after ImageNet reduced mean Dice at four of five budgets. All six 94-label configurations achieved lower mean Dice on ISIC 2017 than ISIC 2016. In the restoration experiment, mean Dice was 0.686 with random initialisation and 0.638-0.642 following pretraining. ConclusionsIn-domain masked-reconstruction pretraining improved mean Dice relative to random initialisation in the custom U-Net, but not consistently in the ResNet-34 U-Net, where ImageNet initialisation performed better at every budget. Adding self-supervised pretraining after ImageNet provided no consistent gain. These findings are specific to the dermoscopic lesion-segmentation experiments and do not establish melanoma-specific diagnostic or clinical performance.
ER -