Manual analysis of image is a time-consuming operation that uses a step-by-step methodology and is very skill-dependent. In contrast, computer system analysis of the resulting magnet and radiography image enables professionals to make quick decisions, produces repeatable results and an electronic patient record, enhances the diagnosis process, and aids in patient treatment. Unsupervised learning and computer-aided diagnosis of brain tumours are the subjects of this study. This paper first describes the state-of-the-art methods for segmenting brain tumours, then explains why soft computing is preferable.