Abstract

Abstract

Abstract

Abstract
Reject option is a technique used to improve classifier’s reliability in decision support systems. It consists in withholding the automatic classification of an item, if the decision is considered not sufficiently reliable. The rejected item is then handled by a different classifier or by a human expert. The vast majority of the works on this issue has been concerned with the development of reject option mechanisms to be used by supervised learning architectures (e.g., MLP, LVQ or SVM). In this paper, however, we aim at proposing alternatives to this view, which are based on the self-organizing map (SOM), originally an unsupervised learning scheme, but that has also been successfully used in the design of prototype-based classifiers. The basic hypothesis we defend is that it is possible to design SOM-based classifiers endowed with reject option mechanisms whose performances are comparable to or better than those achieved by standard supervised classifiers. For this purpose, we carried out a comprehensively evaluation of the proposed SOM-based classifiers on two synthetic and three real-world datasets. The obtained results suggest that the proposed SOM-based classifiers consistently outperform standard supervised classifiers. © 2015 The Natural Computing Applications Forum

Abstract

Abstract
Accurate retrieval of post-surgical images plays a critical role in surgical planning for breast cancer patients. However, current content-based image retrieval methods face challenges related to limited interpretability, poor robustness to image noise, and reduced generalization across clinical settings. To address these limitations, we propose a multistage retrieval pipeline integrating saliency-based explainability, noise-reducing image pre-processing, and ensemble learning. Evaluated on a dataset of post-operative breast cancer patient images, our approach achieves contrastive accuracy of 77.67% for Excellent/Good and 84.98% for Fair/Poor outcomes, surpassing prior studies by 8.37% and 11.80%, respectively. Explainability analysis provided essential insight by showing that feature extractors often attend to irrelevant regions, thereby motivating targeted input refinement. Ablations show that expanded bounding box inputs improve performance over original images, with gains of 0.78% and 0.65% contrastive accuracy for Excellent/Good and Fair/Poor, respectively. In contrast, the use of segmented images leads to a performance drop (1.33% and 1.65%) due to the loss of contextual cues. Furthermore, ensemble learning yielded additional gains of 0.89% and 3.60% over the best-performing single-model baselines. These findings underscore the importance of targeted input refinement and ensemble integration for robust and generalizable image retrieval systems. © 2025 Elsevier B.V., All rights reserved.