AI-Powered system Achieves First Pregnancy in Long-Term Azoospermia Case
A groundbreaking system developed by Columbia University researchers has led to the first triumphant pregnancy in a couple with a 19-year history of infertility, specifically a condition where the male partner had no detectable sperm – a condition known as azoospermia. Called STAR (Sperm Tracking and Recovery), the technology utilizes a powerful combination of deep learning and precision microfluidics to locate viable sperm cells that would be virtually unachievable to find through conventional methods.
Led by Dr. Zev Williams, director of the Columbia fertility center, the STAR system was inspired by techniques used in astrophysics. The core idea is to apply the same AI algorithms used to identify stars amongst billions of celestial bodies to the task of finding sperm within a vast sea of cellular debris. Sperm are the smallest cells in the human body, making manual detection incredibly challenging.
The system operates by continuously analyzing a semen sample flowing at 300 frames per second, processing up to 1.1 million images per hour – with some reports indicating it can handle over 8 million images in under an hour. an object detection model, trained on annotated sperm images, predicts the location of potential sperm cells in real-time. To ensure accuracy, a detection is only confirmed if the same object is identified as sperm in at least three consecutive frames.
In a compelling presentation of its capabilities, STAR identified 44 sperm in just one hour in a sample where embryologists had spent two days searching manually without success. The system doesn’t just find the sperm; it also controls the fluid flow to isolate and recover them for insemination.
The landmark pregnancy occurred after the STAR system analyzed a 3.5 mL semen sample that had previously shown no sperm in manual analysis. The AI processed 2.5 million images in approximately two hours and successfully detected seven sperm. Two of these motile sperm were used in intracytoplasmic sperm injection (ICSI), resulting in two viable embryos. thirteen days after embryo transfer, a positive pregnancy test confirmed the success, ultimately leading to a confirmed clinical pregnancy.
Dr. Williams describes STAR as revealing things previously “blind” to medical professionals,offering a non-invasive and automated alternative for patients facing previously untreatable infertility.The system can be likened to an “ultra-fast and microscopic metal detector,” efficiently sifting through a sample to identify and isolate sperm, unlike the laborious manual process of searching for them under a microscope.
