Artificial intelligence (AI) has been used to evaluate tissue samples in men who produce low levels of, or no, sperm.
Researchers at Toho University School of Medicine in Japan tested whether or not the Google Cloud Automated Machine Learning (AutoML) Vision platform could be used to carry out the traditional Johnson scoring method, in place of pathologists. The Johnson scoring method is used to classify the ability of a male patient to create viable sperm, based on examination of tissue samples taken from the testes, and is often the first stage in the treatment of azoospermia (a condition with no sperm in semen). The researchers' findings have been published in Scientific Reports in Nature Communication.
Dr Hideyuki Kobayashi, associate professor of the urology department at Toho University School of Medicine and author of the paper 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'.
Researchers chose to use the pre-existing Google AutoML Vision tool, to avoid programming a whole new AI tool themselves. They used slides of tissues taken from the testes of 264 patients to create two datasets of images in Adobe Photoshop Elements 2020, two-thirds of which were used to 'train' the tool for one dataset and 78 percent for the other, so that it was able to identify different parts of the tissue found in the testes.
They then tested and validated the algorithm that had been developed using the remaining images to see if it could identify different parts of the tissue, evaluate its appearance and assign one of four labels, related to a Johnson score to it. This was cross-referenced with two researchers' findings using traditional methods, and the machine's findings were found to have a high degree of accuracy. The researchers note however that the value of the findings is limited by potential observer bias in the selection of the images taken from the slides for training, testing and validating the tool.
Dr Kobayashi said evaluating slides of tissues traditionally is complicated and time-consuming for pathologists, because of the 'complexity' of the tissue as it undergoes the multi-stage process of spermatogenesis. The use of an AI model could speed up this process, allowing faster referrals and could even eventually be used in remote areas and developing countries.
'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 and it was very time-consuming. Two colleagues worked very hard to obtain all the images used in the study. I really appreciate their dedicated efforts,' he said.