6 CMR service improvement via deployed service-level rapid CMR protocols with integrated AI. (28th January 2023)

• Record Type: Journal Article
• Authors: Artico, Jessica; Laymouna, Reem; Fox, Paige; Shiwani, Hunain; Kurdi, Hibba; Abioudin, Aderonke; Pierce, Iain; Davies, Rhodri; Xue, Hui; Kellman, Peter; Westwood, Mark; Manisty, Charlotte; Treibel, Thomas; Moon, James
• Is Part Of: Heart. Volume 109(2023)Supplement 1
• Page Start: A5
• Page End: A6
• Publication Date: 2023-01-28
• Subjects: Heart -- Diseases -- Treatment -- Periodicals; Cardiology -- Periodicals; 616.12
• DOI: 10.1136/heartjnl-2022-BSCMR.6
• Languages: English
• ISSNs: 1355-6037

Abstract

Abstract : Background: Demand for CMR has been increasing year on year, and has been exacerbated by the pandemic. The need for rapid, focused protocols to increase through-put, improve cost-effectiveness and reduce waiting lists is now essential. We designed and implemented two new rapid CMR protocols incorporating AI approaches in daily clinical practice and measured their impact. Methods: As part of service level improvements, we implemented two protocols: Protocol 1: rapid Perfusion CMR for ischaemia/viability; Protocol 2: rapid non-contrast CMR for cardiotoxicity. These protocols used inline perfusion mapping with AI analysis, 1 and inline ventricular analysis using the Mycardium AI super-human analysis approach. 2 3 A total of 260 patients were recruited and allocated to either rapid or standard CMR. Scanning times (first to last image timestamp), and image quality (consensus of 2 observers) were assessed. Results: Protocol 1: Conventional stress imaging took an average of 36 minutes (range 24–52minutes, n=80). Rapid perfusion CMR took an average of 23 minutes (range 14 to 31minutes, n=120), an average saving of 13minutes (p<0.001) Protocol 2: Conventional non-contrast CMR took 15.0 (range 11 to 20) minutes. Rapid non-contrast CMR took 9.9 (range 5–13) minutes, including inline analysis, an average saving of 5.1minutes – but this shorter scan included the inline AI analysis, an additional reporting saving. For both protocol 1 and 2, the scan quality was considered similar (3/3, good). Conclusion: Rapid CMR protocols incorporating AI approaches permit major savings on scan duration without any apparent image quality penalties. Scans can consistently be performed in less than 25 minutes and less than 10 for non-contrast). These can be implemented in NHS clinical services. Sequelae: Following this trial, rapid CMR approaches have become routine for ~1/4 patients at our site and booking slots have been reduced to 50minutes (from 1 hour) on all CMR booking diaries with a daily increase in activity of +3 patients a day with consequent benefits to waiting lists. References: Xue H, Davies RH, Brown LAE, et al . Automated Inline Analysis of Myocardial Perfusion MRI with Deep Learning. Radiol Artif Intell . 2020 Oct 21;2 (6):e200009. Hui Xue, Jessica Artico, Rhodri H Davies, et al . Automated In-Line Artificial Intelligence Measured Global Longitudinal Shortening and Mitral Annular Plane Systolic Excursion: Reproducibilityand Prognostic Significance. J Am Heart Assoc . 2022 Feb 15;11 (4):e023849. Rhodri H Davies, João B Augusto, Anish Bhuva, et al . Precision measurement of cardiac structure and function in cardiovascular magnetic resonance using machine learning. J Cardiovasc Magn Reson . 2022 Mar 10;24 (1):16.