An Explainable Self-Supervised Learning Framework for Medical Image Enhancement and Multi-Label Chest X-Ray Disease Classification in Low-Resource Healthcare Settings
Research Digest on Engineering Management and Social Innovations · 26.05.2026 · 2 (5)
DOI: 10.46647/rdems0205078 ↗Kurz gefasst
Fachartikel in der Zeitschrift Research Digest on Engineering Management and Social Innovations, online erschienen am 26.05.2026. Verfasst von Anil Kishore Mugalapalli und 1 weiteren Autor:innen.
Abstract (Originalsprache)
Chest X-ray (CXR) is the most common imaging modality for the diagnosis of thoracic diseases. However, the accuracy of automated CXR interpretation is often limited by the lack of expert annotations, variable image quality, label ambiguity and the distribution shift across hospitals. In this paper, we introduce ESL-CXR, an explainable self-supervised learning framework for conservative CXR image enhancement and multi-label CXR disease classification in low-resource healthcare settings. ESL-CXR integrates image quality control, non-destructive contrast normalization, self-supervised visual encoder, uncertainty-aware supervised fine-tuning, probability calibration and Gradient-weighted Class Activation Mapping (Grad-CAM) to generate explainable model predictions. We employ patient-level data splitting for model training and evaluation. We evaluate the model on both internal test set and external dataset to assess generalizability and mitigate potential label leakage. We identify candidate public research datasets, such as MIMIC-CXR-JPG and CheXpert, based on access and label compatibility. We measure performance using area under the receiver operating characteristic curve (AUROC), area under the precision-recall curve (AUPRC), sensitivity at prespecified specificity, calibration error, and image-quality metrics. Instead of presenting experimental results, this paper introduces a reproducible research design and an integrated framework for future implementation and clinical evaluation.
Zitieren
APA
Mugalapalli, A. K., & Reddy, D. T. B. (2026). An Explainable Self-Supervised Learning Framework for Medical Image Enhancement and Multi-Label Chest X-Ray Disease Classification in Low-Resource Healthcare Settings. Research Digest on Engineering Management and Social Innovations, 2(5). https://doi.org/10.46647/rdems0205078
Vancouver
Mugalapalli AK, Reddy DTB. An Explainable Self-Supervised Learning Framework for Medical Image Enhancement and Multi-Label Chest X-Ray Disease Classification in Low-Resource Healthcare Settings. Research Digest on Engineering Management and Social Innovations. 2026;2(5). doi:10.46647/rdems0205078
BibTeX
@article{Mugalapalli2026,
author = {Mugalapalli, Anil Kishore and Reddy, Dr T Bhaskara},
title = {An Explainable Self-Supervised Learning Framework for Medical Image Enhancement and Multi-Label Chest X-Ray Disease Classification in Low-Resource Healthcare Settings},
journal = {Research Digest on Engineering Management and Social Innovations},
year = {2026},
volume = {2},
number = {5},
doi = {10.46647/rdems0205078}
}
EUVIMED