Machine Learning Based-Personalized Prediction of Sperm Retrieval Success Rate

Peking University

Status: Completed

Conditions: Azoospermia, Nonobstructive; Infertility, Male

Treatments: Diagnostic Test: Machine learning-based predictive model

Study type: Observational

Funder types: Other

Identifiers

NCT06358794

IRB00006761-M2022692

Details and patient eligibility

About

Non-obstructive azoospermia (NOA) stands as the most severe form of male infertility. However, due to the diverse nature of testis focal spermatogenesis in NOA patients, accurately assessing the sperm retrieval rate (SRR) becomes challenging. The current study aims to develop and validate a noninvasive evaluation system based on machine learning, which can effectively estimate the SRR for NOA patients. In single-center investigation, NOA patients who underwent microdissection testicular sperm extraction (micro-TESE) were enrolled: (1) 2,438 patients from January 2016 to December 2022, and (2) 174 patients from January 2023 to May 2023 (as an additional validation cohort). The clinical features of participants were used to train, test and validate the machine learning models. Various evaluation metrics including area under the ROC (AUC), accuracy, etc. were used to evaluate the predictive performance of 8 machine learning models.

Enrollment: 2,612 patients

Sex: Male

Ages: 20 to 60 years old

Volunteers: No Healthy Volunteers

Inclusion criteria

• diagnosed with nonobstructive azoospermia
• underwent microdissection testicular sperm extraction

Exclusion criteria

• without intact clinical information
• low data quality

Trial design

2,612 participants in 2 patient groups

Training cohort

Description:

2,438 patients diagnosed with NOA were included for model training and validation

Treatment:

Diagnostic Test: Machine learning-based predictive model

External validation cohort

Description:

174 participants from January 2023 to May 2023 were included as the external validation cohort for online platform