@misc{18160,
  author = {Jorunn Andersen and Oliwia Witczak and E.U. Due and Steven Hicks and Vajira Thambawita and Lars Bj{\o}rndal and Michael Riegler and Trine Haugen},
  title = {Sperm motility assessed by deep convolutional neural networks into WHO categories},
  abstract = {BACKGROUND AND AIMSemen analysis is central in infertility investigation. Manual assessment of sperm motility according to the WHO recommendations is the golden standard, and extensive training is a requirement for accurate and reproducible results. We have evaluated the performance of Deep convolutional neural networks (DCNNs) in predicting the proportion of sperm in the WHO motility categories. METHODSTwo models were evaluated using 65 video recordings of wet semen preparations and manually obtained values from an external quality assessment programme for semen analysis. One DCNN model was trained to predict the three categories progressive motility, non-progressive motility, and immotile spermatozoa, and the other model four categories, where progressive motility was differentiated into rapid and slow. RESULTSThe resulting average mean absolute error (MAE) was 0.05 and 0.07, and the average ZeroR baseline was 0.09 and 0.10 for the three-category and the four-category model, respectively. Strongest correlation between the mean manually assessed values and DCNN-predicted motility was observed for \% progressively motile spermatozoa (Pearson{\textquoteright}s r = 0.88, p \< 0.001) and \% immotile spermatozoa (r = 0.89, p \< 0.001). For rapid progressive motility, the correlation was moderate (Pearson{\textquoteright}s r = 0.673, p \< 0.001). The DCNN-predicted value was within the range of the interlaboratory variation of the results for most of the samples. CONCLUSIONSIn conclusion, DCNN models were able to predict the proportion of spermatozoa into the WHO motility categories with significantly lower error than the baseline. More data is needed to train supervised AI approaches to improve accuracy and reliability in the assessment of sperm motility. We have therefore provided a dataset with 20 video recordings of wet semen preparations with manually annotated bounding-box coordinates and a set of sperm characteristics. Unlabelled video clips for analysis via methods such as self- or unsupervised learning, are also made accessible.},
  year = {2024},
  journal = {Reproductive BioMedicine Online},
  publisher = {Elsevier},
}