Advertisement
Hong Kong healthcare and hospitals
Hong KongHealth & Environment

Hong Kong researchers develop AI model that detects fertile sperm with 96% accuracy

HKU researchers trained model by examining more than 40,000 sperm images from 117 men diagnosed with infertility

Reading Time:2 minutes
Why you can trust SCMP

10:38

Hong Kong researchers develop AI model that detects fertile sperm with 96% accuracy

Hong Kong researchers develop AI model that detects fertile sperm with 96% accuracy
Angel Woo
A team of Hong Kong researchers has developed an artificial intelligence model that can identify fertile sperm more accurately compared with traditional methods, paving the way for faster and more effective fertility treatments.
Researchers from the University of Hong Kong (HKU) said on Monday that the AI model had an accuracy rate of more than 96 per cent when it came to identifying fertile sperm after being trained with over 1,000 images of male sex cells.

The team further developed the model by examining more than 40,000 sperm images involving 117 men who were diagnosed with infertility or had the condition without a clear cause.

Advertisement

“Traditionally, technicians at the laboratories identify fertile sperm under a microscope in accordance with the guidelines of the World Health Organization. But these methods were not standardised, and results can vary in different laboratories,” said William Yeung Shu-biu, professor of the department of obstetrics and gynaecology at HKU’s clinical medicine school.

Yeung said that traditional methods relied heavily on laboratory technicians’ judgment, making the process subjective and prone to human error. He added that these methods often failed to identify subtle abnormalities in sperm.

Advertisement

According to the WHO, infertility will become the third most common disease after cancer and cardiovascular disease in the 21st century. Male factors contribute to 20 to 70 per cent of infertility cases.

Advertisement
Select Voice
Choose your listening speed
Get through articles 2x faster
1.25x
250 WPM
Slow
Average
Fast
1.25x