top of page

Data Challenge #4


The following data, of over 200 movies of sperm swimming taken using light microscopy, was provided by Gerardo Mendizabal-Ruiz of Conceivable Life Sciences.


The 900MB data set can be downloaded here:

Data - Sperm Swimming Challenge


Here are some ideas for Hackathon projects:


  1. Sperm Motility Classification:

    • Develop a machine learning model to classify sperm motility patterns based on the videos.

    • Train the model to distinguish between categories such as progressive motility, non-progressive motility, and immobility.

    • Evaluate the model's performance and provide insights into the characteristics of healthy and abnormal sperm movement.

  2. Motility Parameter Extraction:

    • Create an algorithm to extract quantitative motility parameters from the videos, such as velocity, straight-line distance, and curvilinear distance.

    • Analyse how these parameters vary among different sperm samples and identify potential correlations with fertility.

  3. Deep Learning for Abnormality Detection:

    • Train a deep learning model to detect abnormalities or irregularities in sperm movement.

    • This could include identifying tail abnormalities, irregular trajectories, or other atypical behaviours that might impact fertility.

  4. Temporal Analysis of Sperm Behaviour:

    • Perform a temporal analysis of sperm movement patterns over the duration of the videos.

    • Identify any trends, periodicities, or changes in motility over time that could be associated with specific stages of sperm life.

  5. Real-time Motility Monitoring:

    • Develop a real-time monitoring system for sperm motility.

    • Utilize the trained model to analyse live video streams, providing immediate feedback on sperm quality and motility.

bottom of page