Javad Khaghani1,2, Saqib Basar1,2, Siavash Khallaghi1,2, Yosef Chodakiewitz2, and Sam Hashemi1,2
1 Vigilance Health Imaging Network Inc, Vancouver, Canada
2 Prenuvo Inc, Vancouver, Canada
Purpose:
This study aims to leverage deep learning insights applied to MR pelvis scans to investigate the impact of metabolic dysfunction on the number of follicles and endometrial thickness in women of reproductive age.
Materials and Methods:
A semantic segmentation network extracted the endometrium and ovaries from sagittal T2 pelvis scans as part of a 1.5T whole body MRI protocol across various sites in US and Canada. Endometrial thickness was measured by applying a distance transform on the sagittal slice with the largest endometrial area. An instance detection solution identified the bounding boxes of individual follicles within each ovary to count the total number of follicles per patient. Metabolic dysfunction was assessed from the patient’s medical history through a medical intake form. One-way Analysis of Variance (ANOVA) and Kruskal-Wallis (KW) tests analyzed the effects of each metabolic dysfunction on follicle count and endometrial thickness in reproductive age women (n=1536). Patients with missing information for any metabolic dysfunction were excluded.
Results:
Significant differences in total follicle count were observed across different blood pressure grades (ANOVA: F=4.95, p=0.01; KW: H=8.31, p=0.02), blood sugar grades (ANOVA: F=3.35, p=0.04; KW: H=7.87, p=0.02), and diabetes status (ANOVA: F=3.21, p=0.04; KW: H=8.62, p=0.01). Specifically, the mean follicle counts were 29.09 (normotensive), 24.49 (elevated), and 25.2 (Stage 1 hypertension) for blood pressure grades, and 29.26 (Normal), 24.25 (Prediabetes), and 28.43 (Diabetic) for blood sugar grades. Further, mean follicle counts were 29.48 (No Diabetes), 39.3 (Type 1 Diabetes), and 22.27 (Type 2 Diabetes) for diabetes classification.
No significant differences in total follicle count were found for thyroid hormone grades (ANOVA: F=0.05, p=0.82; KW: H=0.25, p=0.62) or thyroid conditions (ANOVA: F=0.59, p=0.62; KW: H=1.33, p=0.72). The mean follicle counts were 29.1 (Normal) and 29.44 (Abnormal) for thyroid hormone grades, and 29.61 (No thyroid condition), 28.0 (Hyperthyroidism), 29.1 (Hypothyroidism), and 23.0 (Parathyroidism) for thyroid condition groups.
No significant differences in endometrial thickness were observed across any metabolic dysfunction groups.
Conclusion:
While metabolic factors and conditions such as blood pressure, blood sugar, and diabetes influence follicle count, they do not significantly impact endometrial measurements.
Clinical Relevance Statement:
Metabolics dysfunction is a complex entity that affects multiple organs and systems across the body. The effect of metabolic dysfunction on women’s reproductive health provides valuable insights in fertility assessment and improving patient outcomes.
A) Summary statistics of number of follicles in different metabolic function groups in reproductive age women.
B) Statistical analysis results for the patient dataset. The F-value for ANOVA analysis and the H-value for Kruskal-Wallis analysis are reported, along with the corresponding p-values.
Initial Tables:
Population:
Extra tables:
