An Adaptable SVM Model for Abnormalities Detection in Chest X-ray Reports

Abstract

In Nigeria, prose format is used to present and perform analysis on chest x-ray reports and this often results in delayed response from the clinicians. Therefore, with a view to developing a system for analyzing chest x-ray reports for diagnosing cardiomegaly, linear support vector machine algorithm was utilized to formulate an adaptable model with a train-test split of 70:30 for six hundred and fifty (650) de-identified patients' information. Attributes relevant to cardiomegaly from the collected dataset were extracted using Term frequency/inverse document frequency technique. This work provides an adequate requirement for diagnosis design with accuracy of 93.69%. Its implementation in software application has the potential to reduce delay in attending to patients and can also help the clinicians focus on the findings from chest x-ray reports.