MACHINE LEARNING IN PERSONALIZED MEDICINE FOR BREAST CANCER DIAGNOSIS

Abstract

Aim: The aim of this work is to to improve the prognostic precision in breast cancer by integrating machine learning with proposed methods..

Background: In personalized medicine, the utilization of learning approaches has become increasingly significant as a result of evaluation of complex, large-scale, and unstructured data and information. In recent years, a significant number of researchers have shown an interest in personalized medicine, which entails the development of one-of-a-kind treatments for each individual patient on the basis of their shared traits, which may include their DNA, their heredity, and their way of life.

Problem: a In addition to being one of the malignancies, breast cancer is characterized by a wide range of subtypes that exhibit a diversity of clinical outcomes. Breast cancer has a number of diverse biological origins. As a consequence of this, proper disease stratification is very necessary in order to provide individually tailored therapeutic therapy for breast cancer.

Methodology:  For the purpose of developing prognostic models for the advancement of breast cancer, this study utilized machine learning in conjunction with random optimization (RO) to incorporate glucose metabolism markers and other prognostic characteristics into a dataset. There were many different performance metrics that were utilized in order to evaluate the models.

Results:  There was a high level of analytical performance among the ML-proposed models, with AUC values of 0.75 or above; among these models, the ML-RO-0 model had the greatest relative significance for glucose metabolism features. It was determined that the combined DSS model was successful with a c-statistic of 0.84.