Medical Imaging | RISE Research

This paper presents and tests image-analysis techniques for detecting lung cancer in Computed Tomography (CT) scans, with the goal of providing radiologists with an objective computational aid for early tumour detection. Its premise is that interpreting medical scans varies with a practitioner's experience and subjective judgement, and that as the volume of scans rises, there is growing value in mathematical tools that can support consistent interpretation. The work treats a CT image as a matrix in which each pixel carries a numerical value corresponding to tissue density, which means mathematical rather than purely visual techniques can be used to locate abnormal regions. It implements and tests three classical image-analysis methods, region growing, edge detection and thresholding, on a dataset of lung images to identify regions that may indicate tumours. A central methodological idea is examining the overlap (intersection) of the regions flagged by the different techniques, on the reasoning that areas identified by multiple methods are more likely to be genuine abnormalities, thereby increasing the accuracy of tumour identification. The paper's focus is to demonstrate that combining multiple traditional image-processing techniques can improve detection accuracy while remaining transparent and easily understandable for clinical decision support, in contrast to opaque black-box approaches. Framed around mathematics and medical imaging, it emphasises early and accurate detection as central to improving cancer survival and argues that interpretable, mathematically grounded algorithms can serve as practical, trustworthy tools to assist radiologists rather than replace their judgement, illustrating how matrix-based analysis can objectively surface suspicious structures within a scan.