Survival Analysis of Oncological Patients Using Machine Learning Method

• Published in: Healthcare (Basel) Vol. 11; no. 1; p. 80
• Main Authors: Fryan, Latefa Hamad Al, Alazzam, Malik Bader
• Format: Journal Article
• Language: English
• Published: Switzerland MDPI AG 27-12-2022; MDPI
• Subjects: Algorithms; Analysis; Baghdad Teaching Hospital; Breast cancer; Cancer; Cancer patients; Cancer therapies; Care and treatment; Classification; Data mining; data mining machine learning; Decision trees; Deep learning; Disease; Epidemiology; Forecasts and trends; Health aspects; Histopathology; hospital cancer registries; Hospitals; Machine learning; Medical records; Metastasis; Methods; Oncology; Optimization techniques; Prognosis; Radiation; Registries (in medicine); Statistical analysis; Statistical methods; Survival analysis; Trends; Variables; Iraq; data mining machine learning; cancer; survival analysis; hospital cancer registries; Baghdad Teaching Hospital
• ISSN: 2227-9032; 2227-9032
• DOI: 10.3390/healthcare11010080

Currently, a considerable volume of information is collected and stored by large health institutions. These data come from medical records and hospital records, and the Hospital Cancer Registry is a database for integrating data from hospitals throughout Iraq. The data mining (DM) technique provides knowledge previously not visible in the database and can be used to predict trends or describe characteristics of the past. DM methods can include classification, generalisation, characterisation, clustering, association, evolution, pattern discovery, data visualisation, and rule-guided mining techniques to perform survival analyses that take into account all the patient's medical record variables. For four of the eleven groups examined, this accuracy was relatively high. The database of patients treated by the Baghdad Teaching Hospital between 2018 and 2021 was examined using a classification of the most crucial variables for event prediction, and a distinctive pattern was found. Machine learning techniques allow a global assessment of the data that is available and produce results that can be interpreted as significant information for epidemiological studies, even in cases where the sample is small and there is a lack of information on several variables.