EnsembleEdgeFusion: advancing semantic segmentation in microvascular decompression imaging with innovative ensemble techniques DOI Creative Commons

B. Dhiyanesh,

M. Vijayalakshmi,

P. Saranya

и другие.

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Май 23, 2025

Semantic segmentation involves an imminent part in the investigation of medical images, particularly domain microvascular decompression, where publicly available datasets are scarce, and expert annotation is demanding. In response to this challenge, study presents a meticulously curated dataset comprising 2003 RGB decompression each intricately paired with annotated masks. Extensive data preprocessing augmentation strategies were employed fortify training dataset, enhancing robustness proposed deep learning model. Numerous up-to-date semantic approaches, including DeepLabv3+, U-Net, DilatedFastFCN JPU, DANet, custom Vanilla architecture, trained evaluated using diverse performance metrics. Among these models, DeepLabv3 + emerged as strong contender, notably excelling F1 score. Innovatively, ensemble techniques, such stacking bagging, introduced further elevate performance. Bagging, Naïve Bayes approach, exhibited significant improvements, underscoring potential methods image segmentation. The EnsembleEdgeFusion technique superior loss reduction during compared achieved maximum Mean Intersection over Union (MIoU) scores 77.73%, surpassing other models. Category-wise analysis affirmed its superiority accurately delineating various categories within test dataset.

Язык: Английский

EnsembleEdgeFusion: advancing semantic segmentation in microvascular decompression imaging with innovative ensemble techniques DOI Creative Commons

B. Dhiyanesh,

M. Vijayalakshmi,

P. Saranya

и другие.

Scientific Reports, Год журнала: 2025, Номер 15(1)

Опубликована: Май 23, 2025

Semantic segmentation involves an imminent part in the investigation of medical images, particularly domain microvascular decompression, where publicly available datasets are scarce, and expert annotation is demanding. In response to this challenge, study presents a meticulously curated dataset comprising 2003 RGB decompression each intricately paired with annotated masks. Extensive data preprocessing augmentation strategies were employed fortify training dataset, enhancing robustness proposed deep learning model. Numerous up-to-date semantic approaches, including DeepLabv3+, U-Net, DilatedFastFCN JPU, DANet, custom Vanilla architecture, trained evaluated using diverse performance metrics. Among these models, DeepLabv3 + emerged as strong contender, notably excelling F1 score. Innovatively, ensemble techniques, such stacking bagging, introduced further elevate performance. Bagging, Naïve Bayes approach, exhibited significant improvements, underscoring potential methods image segmentation. The EnsembleEdgeFusion technique superior loss reduction during compared achieved maximum Mean Intersection over Union (MIoU) scores 77.73%, surpassing other models. Category-wise analysis affirmed its superiority accurately delineating various categories within test dataset.

Язык: Английский

Процитировано

0