Research pick: Not every sperm is sacred - "Feed forwarded neural network with learning-based tuna swarm optimisation (FFNN-LBTSO) for semen quality prediction systems"

Artificial intelligence can now provide an accurate perspective on the sensitive issue of male fertility. An algorithmic model described in the International Journal of System of Systems Engineering offers new hope in understanding a puzzling trend in global public health, the decline of sperm quality.

The team introduces a computational framework that is accurate to over 90 percent and outperforms conventional approached used in reproductive medicine. Computer scientists worked with biomedical researchers to develop a feedforward neural network to do the job. The digital architecture of this system builds on the way in which human neurons work but also incorporates an enhanced version of the so-called tuna swarm optimisation, which seeks out solutions to problems in a way analogous to how the fish hunt their prey.

This seemingly odd inspiration from neurons and tuna allows the hybrid model to identify subtle, non-linear patterns linking semen characteristics, such as sperm count, motility, and morphology, with biological and lifestyle factors as well as environmental influences on male fertility.

To avoid the bias inherent in medical datasets, that healthy samples outnumber abnormal ones, the researchers used the Synthetic Minority Oversampling Technique. This method generates artificial examples of the rarer cases, ensuring the model learns to recognise fertility problems as effectively as it recognises normal samples. They then tested the system on publicly available semen data from the University of California, Irvine (UCI) repository. The system achieved higher sensitivity, specificity, and overall accuracy than established approaches.

For many years, research has pointed to a troubling global decline in male fertility, which has been variously attributed to a combination of environmental pollutants, endocrine-disrupting chemicals, lifestyle factors such as poor diet and smoking, and biological influences such as metabolic disorders. The interactions among these variables have not yet been determined. This new approach might help uncover insights into what is causing the problem, although, it will inevitably be a complex combination of factors.

Shanthini, C. and Silvia Priscila, S. (2025) ‘Feed forwarded neural network with learning-based tuna swarm optimisation (FFNN-LBTSO) for semen quality prediction systems’, Int. J. System of Systems Engineering, Vol. 15, No. 5, pp.471–487.