Principal Investigator:
In Collaboration with: Dipartimento Scienze della salute della donna, del bambino e di sanità pubblica
Ricercatori coinvolti:
Dr. Fernando Palluzzi, Dr. Luciano Giaco’, Ing. Tina Pasciuto, Dr.ssa Iolanda Mozzetta
Obiettivo:
The Morphonode Predictive Model (https://github.com/Morphonodepredictivemodel) is an ensemble methodology for the prediction of inguinal lymph node metastasis before surgery. The R-based package is composed by four modules, including: random forest classifiers (Morphonode-RFC) for malignancy prediction, robust binomial regression (Morphonode-RBM) for malignancy risk estimation, decision trees (Modphonode-DT) for the detection of signatures of malignancy risk and metastasis frequency, and a function for similarity profiling (Morphonode-SP) to search for patients with similar ultrasound characteristics, risk level and signature. This ensemble method revealed a higher performance than subjective assessment (93.3% vs. 76.4% of predictive accuracy, respectively) and high robustness to missing data, demonstrating the key importance of computational approaches in personalized medicine and surgery.