2022-RA-1299-ESGO How to predict preoperative risk of lymph node metastasis in vulvar cancer patients the Morphonode Predictive Model. (20th October 2022)

• Record Type: Journal Article
• Authors: Fragomeni, Simona Maria; Moro, Francesca; Palluzzi, Fernando; Federico, Alex; Bove, Sonia; Mascilini, Floriana; Pozzati, Federica; Scambia, Giovanni; Testa, Antonia Carla; Garganese, Giorgia
• Is Part Of: International journal of gynecological cancer. Volume 32(2022)Supplement 2
• Page Start: A445
• Page End: A446
• Publication Date: 2022-10-20
• Subjects: Generative organs, Female -- Cancer -- Periodicals; 616.99465
• DOI: 10.1136/ijgc-2022-ESGO.961
• Languages: English
• ISSNs: 1048-891X

Abstract

Abstract : Introduction/Background: Preoperative evaluation of inguinal lymph nodes in vulvar cancer patients is still a challenge. Our aim was to build a robust, multi-modular ultrasound model based on artificial intelligence. Methodology: From March 2017 to April 2020, 127 women were included at our center and 237 inguinal regions were studied before surgery by ultrasound experienced examiners. Ultrasound features defined in previous studies were prospectively collected. Histopathology was considered the reference standard. Fourteen informative features were used to train and test the machine, in order to obtain a diagnostic model. The following data classifiers were integrated into the predictive model: 1) random forest classifiers (RCF); 2) decisional tree (DT); 3) regression binomial model (RBM); 4) similarity profiling (SP). A predictive tool was implemented in the open-source R package, available on line as 'Morphonode Predictive Model' at https://github.com/Morphonodepredictivemodel. Results: The tool provides four output modules: 1) the binary malignancy prediction (Morphonode-RFC), distinguishing between malignant and benign lymph nodes with an accuracy of 93.3% and a negative predictive value of 97.1% (95%CI 83.8–100.0); 2) the risk signature (Morphonode-DT), identifying 4 specific signatures correlated with the risk of metastases: metastatic signature (MET), high metastatic risk (HMR), moderate metastatic risk (MMR) and low metastatic risk (LMR); the point risk of metastasis for each signature is 100%, 81%, 16% and 4% respectively; MET signature correlates with higher risk for multiple metastatic nodes (frequency of 45.7%); 3) the point malignancy risk, providing a point risk estimate in each specific lymph node described (Morphonode-RBM); 4) a selection of the top- 5 similar profiles in the study series, supporting the clinician to integrate output analysis (Morphonode-SP). Conclusion: Our findings indicate that Morphonode Predictive Model is a simple and observer-independent tool. It could be easily integrated in the clinical routine for preoperative stratification of vulvar cancer patients.