Detecting objects, a significant task in computer vision, is accompanied with many challenges. When we focus on medical images, the challenges of detecting an organ or a tumour exhibit their own specific difficulties. Automatic finding of the prostate gland in magnetic resonance (MR) images, for instance, is a needed task in some clinical procedures. In this paper we propose a novel method for detection of the prostate gland in MR images. We use a multi-stage intra-patient approach to the template matching which is based on a trained dataset. We conduct experiments with the images of 100 patients and report preliminary results that seem to be promising.