AI has a multitude of potential applications in the field of radiology: image normalization, image enhancement, artifact correction, quality control, registration, anatomy segmentation, anomaly detection, and automatic report generation. All of these applications present unique challenges that AI is well-suited to tackle. Multiple solutions for prostate cancer MR assist have been released containing various levels of AI.

Artificial intelligence used for segmentation in prostate MR
Regarding Prostate MR software, Invivo Corp. (Gainesville, FL, USA), now of Philips (Best, The Netherlands), was first to the scene in 2011 with their DynaCAD software to reduce workload, aid interpretation, and help improve clinical accuracy. DynaCAD includes an automatic, model-based prostate segmentation process that the user can manipulate. They then facilitated treatment with their DynaTrim MRI-compatible interventional device for Trans-Rectal prostate intervention guided by their DynaLoc interventional software targeting system for in-bore MR guided biopsies and focal RF ablation. This software is an example of a treatment planning system to facilitate manual guidance of a biopsy needle or treatment probe.
Philips introduced another software, UroNav, that is an elegant decision software that fuses pre-biopsy MR images of the prostate with ultrasound-guided biopsy images in real time by combining electromagnetic tracking and navigation with an onboard computer and a real-time imaging interface. Siemens Healthineers (Erlangen, Germany) introduced their AI-Rad Companion Prostate MR software, which performs similarly as the Philips UroNav software. Both of these software products employ sophisticated registration algorithms to overlay real-time ultrasound images with imported MR images; thus, facilitating MRI-guided biopsies and treatments without the encumbrances and additional run-time expense of the MR system.

Other products such as Plexo (Ezra, New York, NY, USA) were designed to make radiologists’ workflows more efficient when interpreting Prostate MR scans. A key feature includes prostate segmentation and volumetric measurements as outputs to PACs systems or their own stand-alone displays while the system organizes the various imaging sequences conveniently on the display.

All of the above represent examples of limited AI used for organ recognition and segmentation as these are learned processes; however, the remaining elements such as structured displays to assist in throughput are achieved through intelligent software coding. These systems lack the trained AI for automated lesion detection, lesion risk scores and segmentation and/or report writing.

Artificial Intelligence Homepage