NU 2023-107
INVENTORS
Shuhua Zheng*
Eric Donnelly
Jonathan Strauss
SHORT DESCRIPTION
Machine learning-based new unified risk classification score (NU- CATs) for patients with endometrial cancer.
BACKGROUND
Clinically relevant classification of endometrial cancer is important for prognosis and predictions of treatment success. Traditionally, risk stratification of endometrial cancer is done using historical clinical pathological features; this method, however, is prone to inter-rater reliability errors. Adding molecular subgroups of endometrial carcinoma via patient genetic factors increases reproducibility and clincal relevance of these classifications. Though, commonly used standards for risk stratification in endometrial cancer imperfectly represent diverse populations.
ABSTRACT
Northwestern researchers have developed a machine learning-based algorithm which incorporates clinopathologic and molecular information from a diverse population to produce a risk score, termed the New Unified risk classifiCATion score (NU-CATS), for endometrial carcinoma prognosis. This model utilizes inputs from diverse datasets, including clinicopathological (i.e, age, race, histology.) and genetic (i.e, TP53, MMR, mutation count.) data. These data correlated with phenotypes representative of disease progression (i.e., metastasis formation, number of metastatic lesions). A NU-CATS score indicative of increased risk of distant metastasis is produced by a 5-layer deep artificial neural network trained, tested, and validated on publically available clinical datasets. NU-CATS was shown to outperform previously used models for estimating risk of Stage I/II disease progression and survival in African-American endometrial cancer patients.
APPLICATIONS
Risk classification for patients with endometrial cancer
Prognosis for other types of cancer with additional datasets
Tailored therapies based on NU-CATS score
ADVANTAGES
Incorporates diverse clinicalpathologic and molecular variables of endometrial cancer
Utilizes a diverse population of cases, expanding on previously used classification datasets
Utilizes publically available datasets for training, validating, and testing model
Yields superior prognostication of the risk of nodal involvement, distant metastasis formation, disease progression, and overall survival
PUBLICATION
Zheng S. et. al. (2023) A cost-effective, machine learning-based new unified risk-classification score (NU-CATS) for patients with endometrial cancer. Gynecologic oncology. 175: 97-106.
IP STATUS
US patent application filed
Using machine learning classification score to predict survival in endometrial cancer patients; a higher NU-CATS score represents high risk endometrial cancer patients.