Osedax-GAN: The revolutionary Artificial Intelligence enhancing images for early disease detection in fish

Key Takeaways:

• The Osedax-GAN system achieves 92.7% accuracy in disease classification.
• The technology enables pathogen identification 1.6 days earlier than other methods.
• It is the first system of its kind viable for real-time monitoring in commercial fish farms.

One of the greatest threats to the profitability and sustainability of aquaculture is disease, causing annual global losses exceeding $6 billion. Early detection is the best defense, but how can we act in time when we cannot clearly see what is happening underwater?

Computer vision systems show promise, reaching accuracy levels above 95% with high-quality laboratory images. However, their effectiveness plummets to 65-75% when faced with the real-world conditions of an aquaculture farm. Turbid water, variable lighting, reflections, and fish movement corrupt the images, creating “lost pixels” that obscure the subtle, initial signs of disease.

To solve this critical challenge, a research team from Kafrelsheikh University, King Khalid University, Princess Nourah Bint Abdulrahman University, and Delta University for Science and Technology has developed Osedax-GAN. This innovative artificial intelligence framework not only repairs damaged underwater images but does so with unprecedented accuracy and efficiency, heralding a new era for health monitoring in aquaculture.

The problem: When seeing isn’t enough

Traditional visual inspection is laborious, subjective, and often too late. By the time an operator detects a disease by sight, mortality rates can already be between 60% and 80%. Artificial intelligence (AI) promised to automate and accelerate this process, but ran into a fundamental obstacle: the poor quality of real-world images.

Images captured in commercial tanks and cages often suffer from 40% to 60% pixel corruption. Existing AI methods, such as Generative Adversarial Networks (GANs), attempt to “fill in” or “impute” these missing pixels but often fail in two ways:

1. They are not robust enough for dynamic underwater conditions.
2. They prioritize making the image “look good” rather than preserving the crucial pathological details required for an accurate diagnosis.

This creates a dangerous gap: the technology fails precisely when it is needed most—during the early stages of a disease.

An AI inspired by the deep-sea “Zombie Worm”

Osedax-GAN sets itself apart through a unique combination of innovations. Its name derives from its most novel component: a metaheuristic inspired by the Osedax worm, also known as the “zombie” or “bone-eating” worm.

This deep-sea organism has evolved an incredibly efficient strategy for locating and exploiting whale carcasses in the vast, dark ocean. Researchers adapted this biological search strategy to optimize the AI’s training. Instead of searching for nutrients, the algorithm seeks the best way to reconstruct the lost pixels, achieving a perfect balance between exploring new solutions and exploiting the most promising ones.

This approach is integrated into an AI architecture with four key innovations:

1. Osedax Metaheuristic: It uses the “zombie worm’s” logic to guide the AI, achieving superior global search capabilities and greater stability in the complex underwater environment.
2. Identity Block: This acts as a safeguard, protecting original, undamaged pixels to ensure that existing critical visual information is not degraded during the repair process.
3. Adaptive 8-Connected Neighborhood Strategy: Unlike fixed methods, this system analyzes the surroundings of each lost pixel and dynamically adjusts its reconstruction strategy based on turbidity, light, and other water effects.
4. Pathological Feature Loss Function: This is perhaps the most critical innovation for aquaculture. It is a specific instruction for the AI, directing it to prioritize the preservation and reconstruction of visual disease markers (e.g., lesions, spots) over general aesthetic quality.

A quantitative leap in accuracy and speed

Researchers tested Osedax-GAN against nine of the most advanced existing AI models using a dataset of 1,750 images of freshwater fish, covering six common diseases and healthy subjects. The results were compelling, outperforming the competition across all evaluated metrics.

• Superior Image Quality: It achieved the best image restoration quality, with a PSNR of 29.86 dB (a measure of reconstruction quality) and an SSIM of 0.918 (a measure of structural similarity), significantly surpassing the next-best method.
• Higher Diagnostic Accuracy: After repairing the images, classification models achieved an accuracy of 92.7% and an F1-score of 91.7%, an improvement of over 3 percentage points compared to the nearest competitor.
• The Power of Early Detection: The most significant breakthrough was in its early detection capability. Osedax-GAN reached 83.9% accuracy in the initial stages, enabling a diagnosis 1.6 days earlier than existing methods. This time advantage is crucial, as it falls within the window where treatments are over 85% successful.
• Real-World Efficiency: The system is not only more accurate but also more practical. It is 10.9% faster at processing images and consumes 9.9% less energy than the next-best model, making it the first truly viable solution for deployment on real-time monitoring devices directly on farms.

Implications for a more sustainable and Intelligent Aquaculture

The results from Osedax-GAN represent more than a mere technical improvement; they carry direct and profound implications for the sector:

• Reduced Economic Losses: By detecting outbreaks nearly two days earlier, producers can act swiftly to isolate sick fish and apply treatments, drastically reducing mortality rates and preventing multi-million dollar losses.
• Reduced Use of Antibiotics: Early and precise intervention allows for more targeted, less widespread treatments, contributing to more sustainable farming practices and combating antimicrobial resistance.
• Democratization of Technology: Its computational efficiency makes this advanced technology accessible not only to large corporations but also to facilities with limited resources, including farms in developing regions.

Conclusion

According to the study published in the journal Aquacultural Engineering, Osedax-GAN has proven to be a robust and effective solution to one of modern aquaculture’s greatest obstacles: health management under real-world operational conditions. By intelligently repairing underwater images and prioritizing signs of disease, this technology closes the gap between AI’s potential and its practical application in the field.

The ability to reliably detect diseases 1.6 days earlier is a paradigm shift that can strengthen global food security, improve the industry’s sustainability, and equip producers with a powerful tool to protect their most valuable asset: the health of their fish.

Contact
Mostafa Elbaz
Department of Computer Science, Faculty of Computers and Informatics, Kafrelsheikh University
Kafrelsheikh, Egypt
Email: mostafa.albaz@fci.kfs.edu.eg

Reference
Elbaz, M., Alhag, S. K., Al-Shuraym, L. A., Moghanm, F. S., & Marie, H. S. (2025). Osedax-GAN: A novel metaheuristic approach for missing pixel imputation imagery for enhanced detection accuracy of freshwater fish diseases in aquaculture. Aquacultural Engineering, 111, 102606. https://doi.org/10.1016/j.aquaeng.2025.102606

Milthon Lujan

Editor at the digital magazine AquaHoy. He holds a degree in Aquaculture Biology from the National University of Santa (UNS) and a Master’s degree in Science and Innovation Management from the Polytechnic University of Valencia, with postgraduate diplomas in Business Innovation and Innovation Management. He possesses extensive experience in the aquaculture and fisheries sector, having led the Fisheries Innovation Unit of the National Program for Innovation in Fisheries and Aquaculture (PNIPA). He has served as a senior consultant in technology watch, an innovation project formulator and advisor, and a lecturer at UNS. He is a member of the Peruvian College of Biologists and was recognized by the World Aquaculture Society (WAS) in 2016 for his contribution to aquaculture.