Extracting 3D Prostate Geometry from 2D Optically-Tracked Transrectal Ultrasound Images

The technical challenges of traditional transrectal ultrasound-guided prostate biopsy, combined with the limited availability of more advanced prostate imaging techniques, have exacerbated existing differences in prostate cancer outcomes between high-resource and low-resource healthcare settings. The objective of this paper is to improve the tools available to clinicians in low-resource settings by working towards an inexpensive ultrasound-guided prostate biopsy navigation system. The principal contributions detailed here are the design, implementation, and testing of a system capable of generating a 3D model of the prostate from spatially-tracked 2D ultrasound images. The system uses open-source software, low-cost materials, and deep learning to segment and localize cross-sections of the prostate in order to produce a patient-specific 3D prostate model. A user study was performed to evaluate the performance of the system using a simulated patient and ultrasound probe. The results showed that non-clinical users were able to use the system to generate a 3D model of the prostate with minimal training. The capacity to generate a 3D prostate model using readily available technologies has the potential to improve access to navigated biopsy in low-resource settings. Low-cost navigational tools have an important role to play in democratizing access to reliable prostate cancer diagnosis and reducing the burden of disease among those most affected by prostate cancer.