Transrectal ultrasound (TRUS) is regularly used to guide prostate biopsies, which constitute the current gold standard for cancer diagnosis. Unfortunately, TRUS suffers from low sensitivity, leading to a high rate of false negative results. Therefore, MR images – and more recently also Ga-labelled PSMA (Prostate Specific Membrane Antigen) PET – are used to identify suspicious areas in the prostate. The challenge is now to fuse pre-interventional PET/MRI with interventional TRUS both accurately enough to allow targeting with the biopsy needle and fast enough not to impede clinical routine with too much (time-consuming) user interaction.
In this project, we proposed a novel method to first automatically segment the prostate in TRUS using a Hough transform based random forest approach. Then, a elastic surface registration is performed to fuse (PET/)MRI with TRUS, relying on the Coherent Point Drift algorithm. The minimal computation time for the entire pipeline (less than five minutes) allows implementation in the clinical routine.
Prostate image fusion, including registration and often segmentation, remains to be a very active topic in the research community. For details on our method as well as the impressive work of other groups around the world, please refer to our publications and the references therein:
In: 2017 IEEE International Symposium on Biomedical Imaging (ISBI), pp. 853–857, IEEE 2017.
In: Hamlyn Symposium on Medical Robotics, London, UK, pp. 21–22, 2015.
In: International Journal of Computer Assisted Radiology and Surgery, 10 (12), pp. 1997–2007, 2015.
In: Journal of Nuclear Medicine, 56 (supplement 3), pp. 510, 2015.
In: European Urology Supplements, 2 (14), pp. e217, 2015.
In: Clinical Image-Based Procedures. Translational Research in Medical Imaging, pp. 1–8, Springer LNCS, 2014.