VMI-ATN-RCNN: A hybrid deep learning model for fish disease segmentation and classification in aquaculture

Aquaculture is a fast-growing industry in global food production. However, fish disease management is a significant challenge that can severely reduce productivity, yield substantial economic losses, and create major threats to food safety. Accurate disease classification and early diagnosis are cru...

Full description

Saved in:
Bibliographic Details
Published in:Aquaculture Vol. 611; p. 743047
Main Authors: Kabitha, P., Nandini, D. Usha
Format: Journal Article
Language:English
Published: Elsevier B.V 01.01.2026
Subjects:
Aquaculture
Deep learning
Fish disease
Multi-scale convolution layer
Segmentation and classification
Self-adaptive crayfish optimization algorithm
Deep learning
Multi-scale convolution layer
Self-adaptive crayfish optimization algorithm
Segmentation and classification
Fish disease
Aquaculture
ISSN:0044-8486
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Aquaculture is a fast-growing industry in global food production. However, fish disease management is a significant challenge that can severely reduce productivity, yield substantial economic losses, and create major threats to food safety. Accurate disease classification and early diagnosis are crucial for managing illnesses and promoting the adoption of sustainable aquaculture practices. This study introduces a novel deep learning model, VMI-ATN-RCNN, for the segmentation and classification of fish diseases. The proposed approach utilizes a vast dataset of freshwater fish diseases from South Asia, thereby outsmarting the problems faced with small or imbalanced datasets. The proposed model introduces a VMINet backbone network to improve feature extraction for segmentation and classification. Moreover, the Attention-Based Multi-Scale Convolution Layer is introduced in the RPN to enhance the model by focusing more on disease-affected regions and improving its performance over various infection scales. Additionally, the study integrates a grading technique with the proposed model to assess the severity of fish disease. Furthermore, the Self-Adaptive Crayfish Optimization Algorithm is used for automatic hyperparameter tuning to achieve efficient model performance. The proposed VMI-ATN-RCNN model demonstrated superiority with a segmentation accuracy of 99.72 %, a dice coefficient of 0.94, and an IoU of 0.92. The results showed a classification accuracy of 99.72 %, precision of 99.50 %, recall of 99.00 %, and an F1-score of 99.24 %. The experimental results confirm that the VMI-ATN-RCNN model is an effective approach for handling fish diseases in sustainable aquaculture and enhancing food security. •Hybrid VMI-ATN-RCNN enables precise segmentation and classification of fish disease.•Attention-based multi-scale Region Proposal Network boosts lesion detection accuracy.•SA-COA algorithm automates hyperparameter tuning for robust model performance.
ISSN:0044-8486
DOI:10.1016/j.aquaculture.2025.743047