Prognosis of ovarian cancer multi-level-type for given cohorts to recommend the likelihood of treatment outcome

Abstract

Ovarian cancer (OC) is a lethal cancer and ranks 5th in women cancer. Generally, OC has a high incidence of recurrence and is asymptomatic in nature until discovered in the advanced stage. There are around various types and sub-types and subclassification of OC (STSC) but of reported cases 90% of the cases are of epithelial where 80% are of only serous. A total of 240 patients' CT scan OC radiology reports of patients who underwent chemotherapy (biomarker) amounting to 984 reports are used for this retrospective study with cases from 2018 to 2021. This prototype foresees the prognosis of the disease STSC from the baseline (firsthand) reports and recommends the likelihood of treatment status and duration of treatment based on the age group of the patient. Classification to prognosis the records into any one of the STSC and further treatment outcome mapping is performed using two models such as stratified ensemble discrimination (SED) model and hybrid generative transformer discrimination (HGTD) model. SED model (gradient boosting) secured an accuracy of 81% and F1 score of 83% and outperformed the other model. For serous, with the incident rate of approximately 66%, the system has recommended that the occurrence age ranges from approximately 38 to 57 with a treatment duration of 2 years and 9 months with cases showing regression to treatment being 69% and 30% reported stable. The model can help patients, the radiologist, and the oncologist to automatically prognosis and recommend the likelihood of treatment outcomes for a given cohort in critical decision making.