AI Approaches for the Prognosis of the Survival (or Not) of Patients with Bone Metastases

Authors:

Abstract
Cancer spreads in various parts of the body and bones are infected, among others. Bone metastases have severe consequences affecting negatively the patients’ quality of life and putting their survival at stake. By exploiting the experience of experts and available data, AI approaches may be used in tasks such as diagnosis, prognosis and treatment. This paper discusses the use of AI approaches for the prognosis of the survival (or not) of patients with bone metastases after a year from the start of radiotherapy. For this purpose, twenty- seven individual AI approaches were tested. One of them concerned rules elicited from experts and the rest of them concerned approaches whose implementation was based on an available dataset. The best accuracy among the individual approaches was exhibited by the expert rules. Two different types of combined approaches (i.e., loosely coupled sequence and various majority voting classifiers) were also tested. They combined several of the aforementioned individual approaches. Depending on the combined approach used, improvement in accuracy ranged from 2.3% to 9.3% compared to rules working individually. The results demonstrate that combined AI approaches may be useful in the specific domain.

This publication has 23 references indexed in Scilit: