Lawrence N. Tanenbaum MD FACR

The topic of artificial intelligence (AI) is prominent throughout the healthcare enterprise, particularly in imaging services. When employed in the process of image reconstruction, machine learning (ML) based techniques can have an enormous impact on patient comfort and safety.

Both GE Healthcare and Canon Medical Systems are marketing FDA cleared ML methods of image reconstruction on their CT (computed x-Both GE Healthcare and Canon Medical Systems are marketing FDA cleared ML methods of image reconstruction on their CT (computed x-ray tomography) scanners. Initially, CT scanners employed primitive filtered back projection (FBP) image reconstruction techniques which required high x-ray doses to overcome high (assumption and error driven) noise. These were gradually replaced by iterative reconstruction techniques (IR) that reduced dose requirements but created an alteration in image ‘look’ that some found objectionable. Model-based IR techniques followed which further reduced the radiation dose necessary for diagnosis but had an even more atypical appearance.  The new techniques from GE – TrueFidelity and Canon – AiCE (Advanced intelligent Clear IQ Engine) promise to maintain or even further reduce the dose necessary for diagnostic CT. Since they leverage ML techniques trained on traditional CT these images have an appearance more familiar to interpreting physicians which should speed acceptance by the imaging community.

Patients find imaging studies quite stressful with up to 30% suffering frank anxiety reactions during MR imaging. The best way to improve the exam experience is to reduce the time the patient must spend on the imaging table. MedicVision is already marketing an FDA cleared ML-enhanced image space based iterative reconstruction package for MR imaging that offers an approximately 30% reduction in scan times.  GE Healthcare’s MR division is preparing AIRecon – a deep learning (DL) k-space-based image reconstruction tool that increases image spatial resolution and boosts acquisition speed by reducing noise.  Subtle Medical’s deep learning (DL) driven MRI tool, yet to be approved, speeds scan times, shows the potential to reduce contrast dose, and enhance the resolution of MRI studies. Their Subtle-PET application already has FDA clearance for accelerated bedtimes and submission is pending for the indication of reduced radiotracer dose. AI will bring additional patient-focused benefits to MR imaging. Several entities are working on tools that will eliminate signal from motion on MR images.  Eliminating motion can reduce the need for repeat scans which prolong the imaging process. MR scanner vendors are developing DL-based tools for automatic MRI scanning coverage and angulation optimization which should also lead to a more concise and less stressful imaging event.

Questions remain as to the best models for implementation of these techniques and financial models suitable to vendors and the imaging enterprise. We will take them up next time.

Author Bio

Dr. Lawrence N. Tanenbaum is the Chief Technology Officer, Director of Advanced Imaging, Vice President and Medical Director Eastern Region, RadNet Inc.