Artificial Intelligence used to predict male infertility

Male infertility is on the rise globally and unexplained male infertility is present in up to 40% of men with abnormal semen parameters¹.

Testicular biopsy can sometimes be part of the diagnostic process in infertile men, and often confirms a disturbance of spermatogenesis. In particular, it is important in the diagnosis of obstructive azoospermia: in these patients, surgical repair of the genital tract may be possible and, if successful, can result in the presence of spermatozoa in semen².

In order to retrieve biopsy samples testicular sperm extraction (TESE) is undertaken under anaesthesia (local, spinal or general). This involves making an incision in the median raphe of the scrotum through the dartos fibres and the tunica vaginalis. From there incisions are made through the outer covering of the testis to retrieve biopsies of seminiferous tubules containing mature spermatozoa.

When examined, histological specimens are typically given a score, called the Johnsen score, on a scale of 1 to 10, based on the histopathological features of the retrieved tissue.

Johnsen Score table

Stained histograms showing the stages of the Johnsen Score

The Johnsen score has become a feature in urology since it was first postulated half a century ago. However, histopathological evaluation of the testis is not an easy task and is time consuming due to the complexity of testicular tissue arising from the multiple, highly specialized steps in spermatogenesis.

No programming to create an AI model for individual patient data sets

Now a team in the urology department at Toho University in Japan have simplified the time-consuming steps of diagnosis by taking advantage of Artificial Intelligence (AI) technology. To achieve this, they chose Google’s automated machine learning (AutoML Vision) protocol, which requires no programming to create an AI model for individual patient data sets. With AutoML Vision, clinicians with no programming skills can use deep learning in building their own models without help from data scientists.

Professor Hideyuki Kobayashi, Team Leader said, “The model we created can classify histological images of the testis without help from pathologists. I hope that our approach will enable clinicians in any field of medicine to build AI-based models which can be used in their daily clinical practice.”

To simplify the use of Johnsen scores in clinical practice, Professor Kobayashi defined four zones relating to Johnsen scores of 1-3, 4-5, 6-7, and 8-10. He and his co-researchers then uploaded a dataset of 7,155 TESE histograms to the AI platform.

The results were encouraging for the X400 magnification image dataset, the average precision (positive predictive value) of the algorithm was 82.6%.

Professor Kobayashi is enthusiastic about the system and is keen to emphasise its ease of use. He said, “The cloud-based machine learning framework we used is for everyone. It can become such a powerful tool in medicine that, in the near future, doctors in hospitals will be using AI-based medical image classifiers with ease, in the same way they use Microsoft PowerPoint or Excel now.”

The most difficult part was taking images of testis pathology

According to him, the most difficult part was taking images of testis pathology, which was very time consuming. He said, “Two colleagues worked very hard to obtain all the images used in the study. I really appreciate their dedicated efforts.”

AI has become popular in the diagnostic setting and is being applied in all fields of medicine. However, the routine use of AI by clinicians in hospitals is still hampered by the need for major input from data scientists in the use of the technology. This work is therefore significant because it is the first report of an AI algorithm that can be used for predicting Johnsen Scores with precision, without having to rely on specialist pathologists and AI technology experts.