Oocyte hunter by machine learning

Artificial Intelligence

Oocyte hunter by machine learning


IVF begins with finding eggs

At present, more than 5% of newborns in developed countries have been conceived through IVF technology. IVF involves many microscopic operations in the laboratory. First, eggs are separated from follicular fluid extracted from the patient's body under a microscope as soon as possible. The operation time is long, so experienced embryologists must perform the operation. The procedure usually costs 10,000 NTD per IVF cycle in most reproductive medicine centers in Taiwan.

Automation for efficiency and quality

The purpose of this study is to automate the process of finding eggs. The study also aims to reduce egg omission rates and exposure to light, which may harm eggs, in addition to saving manpower. Our first goal is to identify cumulus-oocyte-complexes (COCs), which contain oocytes. Ideally, you could find COC directly under microscopy. These are about 2mm in diameter. However, the follicle fluid often contains debris and blood impurities, making it difficult to find them.

Our machine learning model Oocyte Hunter could find oocytes precisely

The images of egg cumulus cell complexes are annotated on the HTC DeepQ platform using nVidia DGX2. As well as finding eggs in static images, the model can circle eggs in real-time dynamic images. Eventually, the model will be combined with a robot system to build IVF-on-a-chip system.

Artificial intelligence modeling tools in a smart hospital

Automatic AI generation of image report

Electrocardiogram-Based Prediction of Myocardial Infarction and an Automated Echocardiographic Inter