AI-Based Augmented Reality Microscope for Real-Time Sperm Detection and Tracking in Micro-TESE

Area of Science:

• Reproductive Medicine
• Medical Imaging
• Artificial Intelligence

Background:

• Non-obstructive azoospermia (NOA) is a severe male infertility cause.
• Microdissection testicular sperm extraction (Micro-TESE) is time-consuming and labor-intensive for NOA diagnosis.
• Manual sperm searching in Micro-TESE can take hours, leading to embryologist fatigue.

Purpose of the Study:

• To develop and evaluate an augmented reality (AR) microscope system with AI for accelerated sperm retrieval in Micro-TESE.
• To improve the efficiency and accuracy of sperm identification during NOA treatment.

Main Methods:

• Integration of a deep learning model (YOLOv5) for real-time sperm detection.
• Implementation of a multi-object tracker (DeepSORT) for sperm tracking and motility analysis.
• Overlay of sperm data (position, motility) onto the microscope eyepiece view via AR for embryologist guidance.

Main Results:

• The YOLOv5 model demonstrated superior accuracy and speed in sperm detection compared to classical methods (precision 0.81, recall 0.52).
• The AR microscope system approximately doubled the sperm detection rate (66.9% vs. 30.8%) in experimental trials.
• The system provided real-time visual guidance, highlighting sperm and potentially reducing Micro-TESE procedure duration.

Conclusions:

• The AR microscope system with AI significantly aids embryologists in sperm retrieval for NOA.
• This technology shows promise for enhancing efficiency and outcomes in assisted reproductive technology (ART).
• Real-time AI-driven sperm identification can alleviate the burden of Micro-TESE procedures.