Seeking consensus on ultrasound-based diagnosis and phenotyping of PCOS in real-world settings
How are colleagues operationalizing ovarian ultrasound criteria for PCOS given variability in equipment, patient factors, and emerging biomarkers? Specific points for discussion:
- Transducer frequency and FNPO thresholds: With high-resolution transvaginal probes (≥8-9 MHz), do you adhere to an FNPO cutoff of ≥20 follicles per ovary, and how do you adjust thresholds when using lower-frequency transvaginal probes or transabdominal scans in non-sexually active patients or those intolerant of TVS?
- Ovarian volume vs follicle counts: In cycles with a dominant follicle or functional cyst, do you default to ovarian volume (≥10 mL) rather than FNPO? What is your protocol for excluding corpora lutea and large follicles from volume calculations, and do you average across cycles if there is discordance between ovaries?
- Cycle timing and hormonal contraception: Are you standardizing scans to early follicular phase, or accepting any cycle day with documentation? How long after stopping combined hormonal contraception, progestin-only methods, or postpartum/lactation do you wait before relying on ultrasound morphology?
- Adolescents and early post-menarche: For patients within 8 years of menarche, are you avoiding ultrasound-based PCOM labeling entirely, or using ovarian volume z-scores/age-adjusted references to reduce false positives?
- AMH as a surrogate: When ultrasound quality is suboptimal, are you substituting AMH for PCOM, and if so, which assay-specific thresholds and quality standards are you using to avoid overdiagnosis?
- Stromal features and Doppler: Do you incorporate stromal volume/area ratios, echogenicity, or stromal Doppler indices to support hyperandrogenic phenotypes when FNPO is borderline? Any validated cutoffs or reproducibility data you trust?
- Automation and reproducibility: Experience with 3D automated follicle counting (e.g., SonoAVC) or AI-assisted segmentation-has it reduced interobserver variability? What intraclass correlation coefficients are you seeing versus manual counts?
- Special populations and interventions:
- Lean phenotype with normal FNPO but increased stromal echogenicity-how are you documenting and communicating diagnostic uncertainty?
- After metabolic or pharmacologic interventions (e.g., GLP-1 receptor agonists, significant weight loss), have you observed consistent longitudinal changes in FNPO or ovarian volume, and do you reclassify morphology over time?
- Post-ovulation induction cycles or after ovarian drilling-what washout period do you use before reassessment?
- Documentation and reporting: What minimum dataset do you include in reports (probe specs, gain settings, number/size distribution of follicles, ovarian volumes, limitations)? Any templates you recommend for consistency across sonographers?
Would appreciate shared protocols, machine settings that improve follicle conspicuity without overcounting (e.g., gain, harmonics, focal zones), and any institution-specific thresholds that have proven reliable across devices and operators.