Application of smooth OWA operators to classification of retinitis pigmentosa

Retinitis Pigmentosa (RP) is a rare genetic retinal disorder characterized by the progressive degeneration of rod and cone photoreceptors, leading to vision impairment and eventual blindness. This study investigates the application of state-of-the-art convolutional neural networks (CNNs) and aggregation methods related to Ordered Weighted Averaging Operators (OWA) to classify RP with enhanced accuracy. Using pre-trained CNN architectures such as EfficientNet, ResNet, and DenseNet, individual classifiers were evaluated, among which EfficientNet achieved the highest performance. To improve these results, aggregation methods, including classic Ordered Weighted Averaging (OWA) operators and the novel Smooth OWA operators, were employed. The aggregation process significantly boosted classification accuracy, with the OWA operator variants achieving approximately 25 percentage point improvement over the best-performing individual classifier. The best results were obtained using Smooth OWA operators inspired by Newton-Cotes quadratures, achieving a consistent additional improvement over the base OWA operator. This study demonstrates the effectiveness of combining advanced CNN models with aggregation techniques for improving classification accuracy on small and imbalanced datasets. The results highlight the potential of Smooth OWA operators in enhancing the robustness and performance of machine learning models in medical diagnosis tasks.