MOCRA: A multi-algorithm clinical decision support system for the early detection of ovarian cancer

Early and accurate triage of adnexal masses remains challenging due to the heterogeneous presentation of ovarian cancer and the fragmented nature of existing diagnostic tools. While several validated algorithms exist-such as NICE NG12, HSE, IOTA Simple Rules, O-RADS v2022, RMI2, and ROMA-each evaluates different aspects of risk, and none provide an integrated, clinically actionable output. We developed MOCRA (Multivariate Ovarian Cancer Risk Assessment), a deterministic, rule-based clinical decision support system (CDSS) designed to harmonize these tools into a unified risk stratification.

This retrospective, single-center diagnostic accuracy study included 68 analyzable patients with adnexal masses. MOCRA encoded six validated diagnostic algorithms using object-oriented architecture and combined their outputs through transparent precedence rules to produce a four-level risk classification (low, intermediate, high, indeterminate). Diagnostic performance was evaluated against physician-confirmed outcomes (histopathology or ≥ 6-month follow-up). Functional reliability was assessed using predefined test cases, and usability was evaluated by 15 gynecologic oncologists using the Post-Study System Usability Questionnaire (PSSUQ).

Among 68 patients (7 malignant, 61 benign), MOCRA achieved an accuracy of 97.1%, sensitivity 100.0%, specificity 96.7%, F1-score 87.5%, and AUC 0.984. No malignancies occurred in the MOCRA low-risk category. Compared with single algorithms, MOCRA reduced false negatives while maintaining high specificity by cross-validating symptom, biomarker, and ultrasound signals. Functional testing confirmed deterministic and stable performance (mean reliability 4.8/5). Usability ratings were uniformly positive (overall PSSUQ score 4.6/5), with clinicians highlighting the interpretability of the four-tier risk level and the clarity provided by side-by-side algorithm outputs.

MOCRA demonstrates strong diagnostic performance and high clinician usability in this pilot evaluation, suggesting that deterministic integration of multiple validated algorithms can improve consistency and reduce missed high-risk cases. However, the small, single-center dataset-particularly the limited number of malignant cases-warrants cautious interpretation. Larger multicenter and prospective studies with extended follow-up are needed to confirm generalizability and real-world clinical impact.