In this study from Norway it is suggested that the processing of Semen samples (Direct Spermatobioscopy) in the future will be almost Automated, but it is not yet fully carried out despite the fact that there are many of the most modern Casa software out there, to mention one of the best known has a margin of error of almost 10% compared to a human who does it manually in the andorology laboratory. This automatic processing is obtained through the understanding of images. Michael Riegler, a young ESHRE ambassador and member of the Oslo research group, shows that advanced machine learning methods for analyzing videos of semen samples can be a useful tool in researching male infertility. (1) proposes a different perspective to understand it.
At this year’s ESHRE annual meeting, there were several presentations on the topic of machine learning (artificial intelligence) and reproductive outcomes. Although promising, most of the current research in human reproduction is, from the point of view, still in its infancy.
Manual semen analysis is critical to researching male infertility, but it is time consuming and requires extensive training for reproducible results. Automatic analysis started in the 1980s, but was very challenging due to factors such as other types of cells or particles that could not be distinguished.
This study analyzed microscopic videos of semen samples from 85 participants and patient-related data and was limited to information obtained such as age, body mass index (BMI), and days of sexual abstinence.
The objective was to determine if the inclusion of these personal data could help to predict the percentage of progressive and non-progressive sperm motility and immobile sperm.
The results indicated that the selected deep learning algorithms did not lose or gain predictive power, even when sperm concentration was included in the analysis, in contrast to Computer Aided Sperm Analysis (CASA) systems where it is known that concentration is a co-founding variable. Furthermore, we found that incorporating the time from collection to analysis, which inevitably influences sperm motility, represents a major advantage over all classical machine learning methods. The best method outperformed the baseline (mean motility of the dataset as a prediction, also called a null model or ZeroR baseline) by an average mean absolute error of 4.20% for predicting motility. Importantly, our method was able to make the prediction in five minutes, including sample preparation, in contrast to extensive manual analysis.
This paper evaluated automated learning algorithms so it must be done carefully, as high evaluation metric scores are often not an indicator of a proper and working algorithm. Cross validation is an efficient data approach to avoid this and should be included in each analysis of this type of method.
Future research should explore whether additional data from participants, such as fatty acids, genomics, or activity level, could be used and how they should be combined in multimodal analysis to increase predictive power.
Finally, it is worth mentioning that this method allows others to reproduce these results and carry out further analysis on this topic.
Overall, the results indicate that they open up a wide range of possibilities within the field of human reproduction. Furthermore, the quality and thoroughness of evaluation of these methods must be considered to a high standard by the entire medical community.
You still don’t have your Spermatobioscopy study done? By the time you know what conditions you are in, you are sure that you are fertile, knowing how you are can change your life.
Visit us for an assessment consultation for Biology of Human Reproduction
Hicks SA, Andersen JM, Witczak O, et al. Machine learning based analysis of sperm videos and participant data for male fertility prediction. Nature Sci Rep 2019; 9, 16770. doi: 10.1038 / s41598-019-53217-y 2. Riegler MA, Andersen JM, Hammer HL, et al. Artificial intelligence as a tool to predict the mobility and morphology of sperm. Hum Reprod 2019; 34: suppl. 1, P-116. Witczak O, Andersen JM, Hicks SA, et al. Artificial intelligence predicts sperm motility from fatty acids in sperm. Hum Reprod 2019; 34: suppl. 1, P-120. Haugen TB, Hicks SA, Andersen JM, et al. VISEM - A Human Sperm Multimodal Video Data Set. In Proceedings of the 10th ACM 2019 Multimedia Systems Conference: pp 261-266). Topol EJ. High performance medicine: the convergence of human and artificial intelligence. Nat Med 2019; 25: 44-56.