Deep Learning-Based Classification of Canine Cataracts from Ocular B-Mode Ultrasound Images DOI Creative Commons
Sanghyeon Park, Seokmin Go, S. Kim

и другие.

Animals, Год журнала: 2025, Номер 15(9), С. 1327 - 1327

Опубликована: Май 4, 2025

Cataracts are a prevalent cause of vision loss in dogs, and timely diagnosis is essential for effective treatment. This study aimed to develop evaluate deep learning models automatically classify canine cataracts from ocular ultrasound images. A dataset 3155 images (comprising 1329 No cataract, 614 Cortical, 1033 Mature, 179 Hypermature cases) was used train validate four widely (AlexNet, EfficientNetB3, ResNet50, DenseNet161). Data augmentation normalization techniques were applied address category imbalance. DenseNet161 demonstrated the best performance, achieving test accuracy 92.03% an F1-score 0.8744. The confusion matrix revealed that model attained highest cataract (99.0%), followed by Cortical (90.3%) Mature (86.5%) cataracts, while classified with lower (78.6%). Receiver Operating Characteristic (ROC) curve analysis confirmed strong discriminative ability, area under (AUC) 0.99. Visual interpretation using Gradient-weighted Class Activation Mapping indicated effectively focused on clinically relevant regions. learning-based classification framework shows significant potential assisting veterinarians diagnosing thereby improving clinical decision-making veterinary ophthalmology.

Язык: Английский

Deep Learning-Based Classification of Canine Cataracts from Ocular B-Mode Ultrasound Images DOI Creative Commons
Sanghyeon Park, Seokmin Go, S. Kim

и другие.

Animals, Год журнала: 2025, Номер 15(9), С. 1327 - 1327

Опубликована: Май 4, 2025

Cataracts are a prevalent cause of vision loss in dogs, and timely diagnosis is essential for effective treatment. This study aimed to develop evaluate deep learning models automatically classify canine cataracts from ocular ultrasound images. A dataset 3155 images (comprising 1329 No cataract, 614 Cortical, 1033 Mature, 179 Hypermature cases) was used train validate four widely (AlexNet, EfficientNetB3, ResNet50, DenseNet161). Data augmentation normalization techniques were applied address category imbalance. DenseNet161 demonstrated the best performance, achieving test accuracy 92.03% an F1-score 0.8744. The confusion matrix revealed that model attained highest cataract (99.0%), followed by Cortical (90.3%) Mature (86.5%) cataracts, while classified with lower (78.6%). Receiver Operating Characteristic (ROC) curve analysis confirmed strong discriminative ability, area under (AUC) 0.99. Visual interpretation using Gradient-weighted Class Activation Mapping indicated effectively focused on clinically relevant regions. learning-based classification framework shows significant potential assisting veterinarians diagnosing thereby improving clinical decision-making veterinary ophthalmology.

Язык: Английский

Процитировано

0