Advanced methods in reproductive medicine: Application of optical nanoscopy, artificial intelligence-assisted quantitative phase microscopy and mitochondrial DNA copy numbers to assess human sperm cells

Abstract

Declined fertility rate and population is a matter of serious concern, especially in the developed nations. Assisted Reproductive Technologies (ART), including in vitro fertilization (IVF), have provided great hope for infertility treatment and maintaining population growth and social structure. With the help of ART, more than 8 million babies have already been born so far. Despite the worldwide expansion of ART, there is a number of open questions on the IVF success rates. Male factors for infertility contribute equally as female factors, however, male infertility is primarily focused on the “semen quality”. Therefore, the search of new semen parameters for male fertility evaluation and the exploration of the optimal method of sperm selection in IVF have been included among the top 10 research priorities for male infertility and medically assisted reproduction. The development of imaging systems coupled with image processing by Artificial Intelligence (AI) could be the revolutionary step for semen quality analysis and sperm cell selection in IVF procedures. For this work, we applied optical nanoscopy technology for the analysis of human spermatozoa, i.e., label-based Structured Illumination Microscopy (SIM) and non-invasive Quantitative Phase Microscopy (QPM). The SIM results demonstrated a prominent contrast and resolution enhancement for subcellular structures of living sperm cells, especially for mitochondria-containing midpiece, where features around 100 nm length-scale were resolved. Further, non-labeled QPM combined with machine learning technique revealed the association between gradual progressive motility loss and the morphology changes of the sperm head after external exposure to various concentrations of hydrogen peroxide. Moreover, to recognize healthy and stress-affected sperm cells, we applied Deep Neural Networks (DNNs) to QPM images achieving an accuracy of 85.6% on a dataset of 10,163 interferometric images of sperm cells. Additionally, we summarized the evidence from published literature regarding the association between mitochondrial DNA copy numbers (mtDNAcn) and semen quality. To conclude, we set up the high-resolution imaging of living human sperm cells with a remarkable level of subcellular structural details provided by SIM. Next, the morphological changes of sperm heads resulting from peroxidation have been revealed by QPM, which may not be explored by microscopy currently used in IVF settings. Besides, the implementation of DNNs for QPM image processing appears to be a promising tool in the automated classification and selection of sperm cells during IVF procedures. Moreover, the results of our meta-analysis showed an association of mtDNAcn in human sperm cells and semen quality, which seems to be a relevant sperm parameter for routine clinical practice in male fertility assessment.