Categories
Change Password!
Reset Password!
Machine learning-based heatmaps can improve precision in myomectomy decisions, potentially leading to better patient outcomes and more targeted treatments for infertility.
Uterine myomas (also known as uterine fibroids) are a known cause of infertility, and determining when myomectomy is necessary can be complex due to their varying size and location. To address this challenge, a recent study published in ‘Reproductive Sciences’ has developed a new visualization tool to assist patients with infertility in making more informed decisions about surgical removal of fibroids i.e. myomectomy.
One hundred and ninety-one women with fibroids were included, with 124 undergoing myomectomies. Among these, 65 (52.4%) achieved pregnancy within an average of 17.6 months post-surgery, and 54 (83.1%) delivered live births. Researchers created a logistic regression model to predict pregnancy rates, achieving a high validation accuracy of 74.6% based on factors such as age, type of myomas, and size, etc.
Takuya Yokoe et al. developed a nomogram to visualize how each factor impacts pregnancy outcomes, using preoperative magnetic resonance imaging data and machine learning through a convolutional neural network. The model's classification precision was 71.4% and 77.7% for sensitivity and specificity, respectively. Gradient-weighted class activation mapping generated via heatmaps distinguished between fibroids that needed surgery and those that did not. This pilot study shows machine learning could improve myomectomy decisions, though more clinical studies with larger samples are needed to validate the outcomes.
Reproductive Sciences
Monogram and Heat Map on Magnetic Resonance Imaging to Evaluate the Recommendation for Myomectomy in Patients with Infertility: A Pilot Study
Takuya Yokoe et al.
Comments (0)