Manuscript received December 27, 2025; revised January 30, 2026; accepted April 8, 2026; published October 20, 2026.

Abstract—Breast cancer remains one of the most dangerous and prevalent diseases in women worldwide, so early and precise diagnostic methods are very important. Mammogram analysis using deep learning has been very promising, but single-model solutions are prone to poor generalization and suffer from overfitting, particularly when class-imbalanced or small medical datasets are involved. We present here a framework by integrating deep transfer learning with selective ensemble learning for improved classification of breast cancer from mammograms. We trained four state-of-the-art convolutional neural networks, i.e., DenseNet121, EfficientNetV2B0, MobileNetV2, and XceptionNet, on two datasets, the Curated Breast Imaging Subset of Digital Database for Screening Mammography (CBIS-DDSDM) and Digital Mammography Dataset for Breast Cancer Diagnosis (DMID). The models that performed best were then combined to work together, boosting overall accuracy. Our results demonstrate that the ensemble model outperforms the networks in isolation, achieving higher accuracy, lower loss, and improved generalization. The research also compares the energy consumption and carbon footprint of each model and determines EfficientNetV2B0 to be the most environmentally friendly architecture. Statistical inferences through Analysis of Variance (ANOVA) and Quantile-Quantile (QQ) plots were also performed on the models. The proposed method offers a clinically viable, efficient, and environmentally friendly approach to automatic breast cancer diagnosis.

Keywords—breast cancer, mammograms, deep learning, transfer learning, ensemble learning, carbon footprint
 

Cite: Mayank Kapadia and Parita Oza, "Selective Ensemble and Transfer Learning for Breast Cancer Classification from Mammograms," Journal of Image and Graphics, Vol. 14, No. 5, pp. 891-901, 2026.

Copyright © 2026 by the authors. This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited (CC BY 4.0).