The Global Problem of Azoospermia and Methods for Its Clinical Diagnosis
One of the most significant obstacles to couples achieving their reproductive goals is azoospermia. This condition is characterized by the complete absence of sperm in the ejaculate, making natural conception virtually impossible. In modern medicine, a testicular biopsy—the surgical removal of a tissue sample to search for viable sperm—is used to assist such patients. These sperm are subsequently used in ART protocols, where IVF (in vitro fertilization) becomes the primary method for achieving pregnancy. As part of these procedures, artificial intelligence in IVF helps doctors select the best sperm samples and significantly increase the chances of success.
A key tool in this process is the histological evaluation of the obtained samples. For decades, the Johnson scale has remained the gold standard here. This method involves a thorough examination of the tissue under a microscope and assigning it a rating from 1 to 10 points, where the highest score indicates full spermatogenesis, and the lowest indicates its complete absence.
Integration of Artificial Intelligence Technologies into Histological Analysis
Despite its effectiveness, the traditional Jones scale assessment has a significant drawback—it is extremely labor-intensive and requires the pathologist or urologist to maintain the utmost concentration and devote a considerable amount of time. “This scale has been used in urological practice for over half a century,” notes Hideyuki Kobayashi, associate professor in the Department of Urology at Toho University in Japan. “However, manual tissue analysis is a complex and exhausting process.”
To modernize this stage of diagnosis, a research team led by Dr. Kobayashi implemented artificial intelligence (AI) technologies. Google Automated Machine Learning (AutoML) Vision software was chosen as the foundation. The main advantage of this tool is its accessibility: AutoML Vision technology is designed so that medical professionals without deep programming or coding skills can use it effectively.
Model Training Methodology and Achieved Results
To improve the efficiency of automation, Dr. Kobayashi optimized the structure of the Jonsen scale by combining the ten levels into four broader groups:
- 1–3 points
- 4–5 points
- 6–7 points
- 8–10 points
This approach simplified the classification task for machine vision algorithms while retaining all the clinical significance necessary for diagnosis. An impressive amount of data was used to train the neural network: 7,155 high-quality images of histological specimens. Final testing of the model yielded impressive results: the AI’s accuracy in automated slide assessment reached 80%.
Prospects and the Future of AI in Clinical Medicine
The introduction of AI into urology is part of a global trend. Previously, similar technologies have already proven their effectiveness in diagnosing breast cancer and detecting lung pathologies in CT scans. The use of smart classifiers is becoming particularly relevant in complex fertility programs. For example, during IVF with egg donation or IVF with double donation, the accuracy of assessing male infertility factors using AI allows for the most effective preparation of biomaterial and prediction of treatment outcomes.
Nevertheless, the widespread adoption of algorithms in everyday hospital practice has long been hindered by the need for specialized data experts. The use of tools like AutoML Vision is a game changer. Dr. Kobayashi is convinced that in the near future, AI-based medical image classifiers will become as familiar and easy to use for doctors as standard office software is today. This will significantly speed up diagnosis and make precision medicine more accessible.
FAQ: Frequently Asked Questions
1. What is the Johnson Scale and why is it needed? The Johnson Scale is a system for assessing the condition of testicular tissue based on biopsy results. It helps doctors determine the degree of preservation of sperm production function. The score ranges from 1 (complete absence of spermatogenic cells) to 10 (active, normal spermatogenesis).
2. How does artificial intelligence help with IVF? AI is used to analyze medical images (sperm, embryos, histological sections), which allows for the selection of the most viable cells and increases the likelihood of successful fertilization.
3. Is AI used in IVF with egg donation? Yes, in programs using donor material (both in egg donation and in IVF with double donation), AI helps ensure strict quality control and accurate assessment of male biomaterial obtained via biopsy.
4. How accurate is AI in biopsy analysis? In a study by Toho University, the model demonstrated 80% accuracy. This allows doctors to process data more quickly and eliminate subjective errors in classification using the Johnson scale.
5. Does a doctor need to be a programmer to work with AI? No. Modern platforms, such as Google AutoML Vision, allow medical professionals to harness the power of machine learning without writing code, using intuitive interfaces.
Leave a Reply