Artificial Intelligence in Cardiac MRI: Is Clinical Adoption Forthcoming?

Artificial intelligence (AI) refers to the area of knowledge that develops computerised models to perform tasks that typically require human intelligence. These algorithms are programmed to learn and identify patterns from “training data,” that can be subsequently applied to new datasets, without being explicitly programmed to do so. AI is revolutionising the field of medical imaging and in particular of Cardiovascular Magnetic Resonance (CMR) by providing deep learning solutions for image acquisition, reconstruction and analysis, ultimately supporting the clinical decision making. Numerous methods have been developed over recent years to enhance and expedite CMR data acquisition, image reconstruction, post-processing and analysis; along with the development of promising AI-based biomarkers for a wide spectrum of cardiac conditions. The exponential rise in the availability and complexity of CMR data has fostered the development of different AI models. Integration in clinical routine in a meaningful way remains a challenge. Currently, innovations in this field are still mostly presented in proof-of-concept studies with emphasis on the engineering solutions; often recruiting small patient cohorts or relying on standardised databases such as Multi-ethnic Study on atherosclerosis (MESA), UK Biobank and others. The wider incorporation of clinically valid endpoints such as symptoms, survival, need and response to treatment remains to be seen. This review briefly summarises the current principles of AI employed in CMR and explores the relevant prospective observational studies in cardiology patient cohorts. It provides an overview of clinical studies employing undersampled reconstruction techniques to speed up the scan encompassing cine imaging, whole-heart imaging, multi-parametric mapping and magnetic resonance fingerprinting along with the clinical utility of AI applications in image post-processing, and analysis. Specific focus is given to studies that have incorporated CMR-derived prediction models for prognostication in cardiac disease. It also discusses current limitations and proposes potential developments to enable multi-disciplinary collaboration for improved evidence-based medicine. AI is an extremely promising field and the timely integration of clinician's input in the ingenious technical investigator's paradigm holds promise for a bright future in the medical field.

366 The use of dedicated long-axis views focused on the left atrium improves the accuracy of left atrial volumes and function measured by cardiovascular magnetic resonance

Aims The use of apical views focused on the left atrium (LA) has improved the accuracy of LA volume evaluation by two-dimensional echocardiography. However, routine cardiac magnetic resonance (CMR) evaluation of LA volumes still uses standard 2- and 4-chamber cine images focused on the left ventricle. To investigate the potential of LA-focused CMR cine images, we compared LA maximal (LAVmax) and minimal (LAVmin) volumes, and emptying fraction (LA-EF) calculated on both standard and LA-focused long-axis cine images with LA volumes obtained by short-axis cine stacks covering the LA. Methods and results LA volumes and LA-EF were obtained from 108 consecutive patients by applying the biplane area-length algorithm to both standard and LA-focused 2- and 4-chamber cine images. Manual segmentation of a short-axis cine stack covering the LA was used as the reference method. Compared to the reference method, the standard approach significantly underestimated LA volumes (LAVmax: bias −13 ml; LOA = +11 ml, −37 ml; LAVmin; bias −10 ml, LOA: +9 ml, −28ml), and overestimated LA-EF (bias= 5%, LOA: +23%, −14%). Conversely, LA volumes (LAVmax bias −0.03 ml; LOA: +10 ml, −10 ml. LAVmin bias = −1.5 ml; LOA: +7 ml, −10 ml), and LA-EF (bias 2%, LOA: +11%, −7%) by LA-focused cine images were similar to those measured using the reference method. Moreover, LA volumes by LA-focused images were obtained faster than using the reference method (1.2 vs. 4.5 min, P < 0.001). Conclusions LA volumes and LA-EF measured using dedicated LA-focused long-axis cine imaging are more accurate than using standard (LV-focused) cine images.

Myocardial changes on 3T cardiovascular magnetic resonance imaging in response to haemodialysis with fluid removal

Background Mapping of left ventricular (LV) native T1 is a promising non-invasive, non-contrast imaging biomarker. Native myocardial T1 times are prolonged in patients requiring dialysis, but there are concerns that the dialysis process and fluctuating fluid status may confound results in this population. We aimed to assess the changes in cardiac parameters on 3T cardiovascular magnetic resonance (CMR) before and after haemodialysis, with a specific focus on native T1 mapping. Methods This is a single centre, prospective observational study in which maintenance haemodialysis patients underwent CMR before and after dialysis (both scans within 24 h). Weight measurement, bio-impedance body composition monitoring, haemodialysis details and fluid intake were recorded. CMR protocol included cine imaging and mapping native T1 and T2. Results Twenty-six participants (16 male, 65 ± 9 years) were included in the analysis. The median net ultrafiltration volume on dialysis was 2.3 L (IQR 1.8, 2.5), resulting in a median weight reduction at post-dialysis scan of 1.35 kg (IQR 1.0, 1.9), with a median reduction in over-hydration (as measured by bioimpedance) of 0.75 L (IQR 0.5, 1.4). Significant reductions were observed in LV end-diastolic volume (− 25 ml, p = 0.002), LV stroke volume (− 13 ml, p = 0.007), global T1 (21 ms, p = 0.02), global T2 (− 1.2 ms, p = 0.02) following dialysis. There was no change in LV mass (p = 0.35), LV ejection fraction (p = 0.13) or global longitudinal strain (p = 0.22). On linear regression there was no association between baseline over-hydration (as defined by bioimpedance) and global native T1 or global T2, nor was there an association between the change in over-hydration and the change in these parameters. Conclusions Acute changes in cardiac volumes and myocardial native T1 are detectable on 3T CMR following haemodialysis with fluid removal. The reduction in global T1 suggests that the abnormal native T1 observed in patients on haemodialysis is not entirely due to myocardial fibrosis.

Segmental evaluation of right ventricular systolic function in atrial septal defect (ASD) type II patients

Introduction A left to right (LR) shunt in atrial septal defect (ASD) may cause right heart and pulmonary overfilling, at the expense of the systemic circulation. Purpose The study objective was to evaluate the impact of LR shunt on left (LV) and right ventricular (RV) filling, function, and myocardial strain by using cardiovascular magnetic resonance imaging (CMR). Methods Thirty-five ASD type secundum patients (42±18 y.o.) were compared to a control group (n=40). Cine imaging was used to calculate ventricular volumes and ejection fraction (EF), global longitudinal (GLS) and circumferential (GCS), free wall (FW) and interventricular septal (IVS) longitudinal strain. Phase-contrast imaging was used to calculate pulmonary flow to systemic flow ratio (Qp/Qs). Results Qp/Qs was 2.2±0.60 (range 1.3–3.6), which resulted in higher RV end-diastolic volume/BSA (EDVi, 152±42 vs. 82±11 ml/m2), lower LV EDVi (72±17 vs. 83±10 ml/m2), and higher RV/LV EDVi ratio (2.1±0.5 vs. 1±0.1) compared to controls (all p<0.001) [Figure 1]. Patients also presented with higher RV, but lower LV indexed stroke volumes (both p<0.001), and a strong trend toward lower RVEF (p=0.08). They demonstrated significantly lower RV GLS (p=0.03) and longitudinal IVS strain (p<0.001) [Figure 2]. RV FW strain or RV GCS did not differ among study groups. Shunt severity correlated with RV size and stroke volume, right atrial size and pulmonary trunk diameter (all p<0.001). In contrast, no correlation was identified with functional nor strain parameters. Conclusion Cardiac remodeling in ASD patients with long-standing LR shunt negatively affects RV systolic performance, which is likely related to longitudinal septal dysfunction. FUNDunding Acknowledgement Type of funding sources: None. Figure 1 Figure 2

Flow vorticity relationships with right ventricular geometry in adult tetralogy of Fallot

Background Repaired tetralogy of Fallot adults (rToF) undergo right ventricular (RV) remodeling, in part due to volume overload of residual pulmonary regurgitation volume (PRV). Time-resolved phase-contrast cardiac magnetic resonance imaging (4D Flow MRI) enables the qualitative and quantitative measurement of altered blood flow patterns, including vorticity. Cardiac atlases allow for complex three-dimensional heart shapes to be expressed as morphometric scores. Those scores show the extent of geometrical shift and can help explore uncharted relationships between vorticity and architecture. Purpose We aimed to quantify vorticity, incorporating deep learning to enhance 4D Flow data, and correlate this with global cardiac parameters and morphometric scores. Methods 12 Adult rToF patients and 10 age-matched controls underwent 4D flow MRI and cine imaging. RV interventricular vorticity was calculated for outflow and inflow tracts. EDV, ESV and SV were computed from cines which were also used to build three-dimensional shape models. The biventricular models were projected onto an atlas generated from 95 rToF patients, and twenty-one principal component analysis shape modes were correlated with cardiac metrics and vorticity to identify global shape variations. Association between biventricular shape and vorticity was further analysed using multivariate multiple regression models. Results Strong correlation was found between PRV and the right ventricular outflow tract (RVOT) vorticity. PRV and RVOT vorticity both correlated with the same 3 shape modes (r=−0.55, −0.50 and 0.6 (p<0.05) respectively for PR and r=0.63, −0.82 and 0.60 (p<0.05) respectively for vorticity) i.e., the RV dilates with an increase in basal bulging, apical bulging and tricuspid annulus tilting with more severe regurgitation, as well as a smaller LV, and a paradoxical movement of the septum (Figure 1). However, RV vorticity correlated with 2 modes that did not correlate with PRV, (r=−0.62, −0.69, p<0.05). With higher vorticity the RV was longer, increased tilting of the tricuspid annulus and an increased basal bulge around the tricuspid area. The multivariate analysis model demonstrated that higher vorticity was associated with displacement of the pulmonary valve and change in the RVOT length and direction. A septal displacement towards the left ventricle was observed and increased apical flatness of the RV (Figure 1). Qualitatively, vorticity in rToF group was more heterogeneous than controls (Figure 2). Conclusions Vorticity is a novel marker based on the influence of blood motion providing new insight into early diagnosis and prognosis of cardiac disease. This is the first study to examine the relationships between vorticity and regional RV shape changes in rToF. Mode associations with vorticity were different to associations with PRV. More longitudinal studies are required for standardization of change in vorticity with the disease process. FUNDunding Acknowledgement Type of funding sources: Foundation. Main funding source(s): The New Zealand heart foundation Mode variations and morphometric model Vorticity visualization and analysis

mRNA COVID-19 Vaccination and Development of CMR-confirmed Myopericarditis

Introduction: Several case reports or small series have suggested a possible link between mRNA COVID vaccines and the subsequent development of myocarditis and pericarditis. This study is a prospective collection and review of all cases with a myocarditis/pericarditis diagnosis over a 2-month period at an academic medical center. Methods: Prospective case series from 1st June 2021 until 31st July 2021. Patients were identified by admission and discharge diagnoses which included myocarditis or pericarditis. Inclusion criteria were: in receipt of mRNA vaccine within one month prior to presentation; The CMR protocol included cine imaging, native T1 and T2 mapping, late gadolinium enhancement and post contrast T1 mapping. All CMR studies were read in consensus by two experienced readers. Diagnosis was based on clinical presentation, ECG/echo findings and serial troponins and was confirmed in each case by CMR. Incidence was estimated from total doses of mRNA vaccine administered in the Ottawa region for the matching time-period. This data was obtained from the Public Health Agency of Ottawa. Results: 32 patients were identified over the period of interest. Eighteen patients were diagnosed with myocarditis; 12 with myopericarditis; and 2 with pericarditis alone. The median age was 33 years (18-65 years). The sex ratio was 2 females to 29 male. In 5 cases, symptoms developed after only a single dose of mRNA vaccine. In 27 patients, symptoms developed after their second dose of. Median time between vaccine dose and symptoms was 1.5 days (1-26 days). Chest pain was the commonest symptom, but many others were reported. Non-syncopal non-sustained ventricular tachycardia was seen in only a single case. Median LV ejection fraction (EF) was 57% (44-66%). Nine patients had an LVEF below the normal threshold of 55%. Incidence of myopericarditis overall was approximately 10 cases for every 10,000 innoculations. Summary and Conclusions: This is the largest series in the literature to clearly relate the temporal relationship between mRNA COVID vaccination, symptoms and CMR findings. In most patients, symptom onset began within the first few days after vaccination with corresponding abnormalities in biomarkers and on ECG. Cardiac MRI confirmed acute myocardial and pericardial changes with the presence of edema demonstrated with both tissue mapping and late gadolinium enhancement. Symptoms settled quickly with standard therapy and patients were discharged within a few days. No major adverse cardiac events and no significant arrhythmias were noted during inpatient stay. Further follow up will be required to ascertain the longer-term outcomes of this patient group.

T1T2 Mapping and Texture Analysis of Cine and T2-weighted Magnetic Resonance Imaging to Detect Myocardial Tissue Changes Associated with Chronic Kidney Disease

Myocardial tissue changes associated with chronic kidney disease (CKD) may lead to heart failure, serious ventricular arrhythmia, and sudden cardiac death. Here, we sought to determine usefulness of T1T2 mapping and texture analysis of T2-weighted short inversion time inversion recovery (STIR) and cine imaging to detect myocardial changes associated with CKD. We examined 34 patients with CKD and 10 controls using a 1.5 T system. T1 and T2 values of the septal myocardium were significantly greater in patients than in controls (P = 0.021 for T1 and P < 0.01 for T2). The open-access texture analysis software, including features reduction methods, selected one particular feature for T2-weighted STIR and another for cine imaging: vertical fraction for STIR and horizontal short-run emphasis for cine imaging, which were significantly lower in the patients with CKD than in the controls (P < 0.01 for both). Texture analysis was not significantly better than T1T2 mapping in detecting the myocardial tissue changes, whereas STIR suffered from image artifacts in 9 patients. In conclusion, additional T1T2 mapping or postprocessing texture analysis may be feasible for detection of myocardial tissue changes associated with CKD.

CMR-based T1-mapping offers superior diagnostic value compared to longitudinal strain-based assessment of relative apical sparing in cardiac amyloidosis

Cardiac amyloidosis (CA) is an infiltrative disease. In the present study, we compared the diagnostic accuracy of cardiovascular magnetic resonance (CMR)-based T1-mapping and subsequent extracellular volume fraction (ECV) measurement and longitudinal strain analysis in the same patients with (a) biopsy-proven cardiac amyloidosis (CA) and (b) hypertrophic cardiomyopathy (HCM). N = 30 patients with CA, N = 20 patients with HCM and N = 15 healthy control patients without relevant cardiac disease underwent dedicated CMR studies. The CMR protocol included standard sequences for cine-imaging, native and post-contrast T1-mapping and late-gadolinium-enhancement. ECV measurements were based on pre- and post-contrast T1-mapping images. Feature-tracking analysis was used to calculate 3D left ventricular longitudinal strain (LV-LS) in basal, mid and apical short-axis cine-images and to assess the presence of relative apical sparing. Receiver-operating-characteristic analysis revealed an area-under-the-curve regarding the differentiation of CA from HCM of 0.984 for native T1-mapping (p < 0.001), of 0.985 for ECV (p < 0.001) and only 0.740 for the “apical-to-(basal + midventricular)”-ratio of LV-LS (p = 0.012). A multivariable logistical regression analysis showed that ECV was the only statistically significant predictor of CA when compared to the parameter LV-LS or to the parameter “apical-to-(basal + midventricular)” LV-RLS-ratio. Native T1-mapping and ECV measurement are both superior to longitudinal strain measurement (with assessment of relative apical sparing) regarding the appropriate diagnosis of CA.

Segment Length in Cine Strain Analysis Predicts Cardiac Resynchronization Therapy Outcome Beyond Current Guidelines

Background: Patients with a class I recommendation for cardiac resynchronization therapy (CRT) are likely to benefit, but the effect of CRT in class II patients is more heterogeneous and additional selection parameters are needed in this group. The recently validated segment length in cine strain analysis of the septum (SLICE-ESS sep ) measurement on cardiac magnetic resonance cine imaging predicts left ventricular functional recovery after CRT but its prognostic value is unknown. This study sought to evaluate the prognostic value of SLICE-ESS sep for clinical outcome after CRT. Methods: Two hundred eighteen patients with a left bundle branch block or intraventricular conduction delay and a class I or class II indication for CRT who underwent preimplantation cardiovascular magnetic resonance examination were enrolled. SLICE-ESS sep was manually measured on standard cardiovascular magnetic resonance cine imaging. The primary combined end point was all-cause mortality, left ventricular assist device, or heart transplantation. Secondary end points were (1) appropriate implantable cardioverter defibrillator therapy and (2) heart failure hospitalization. Results: Two-thirds (65%) of patients had a positive SLICE-ESS sep ≥0.9% (ie, systolic septal stretching). During a median follow-up of 3.8 years, 66 (30%) patients reached the primary end point. Patients with positive SLICE-ESS sep were at lower risk to reach the primary end point (hazard ratio 0.36; P <0.001) and heart failure hospitalization (hazard ratio 0.41; P =0.019), but not for implantable cardioverter defibrillator therapy (hazard ratio, 0.66; P =0.272). Clinical outcome of class II patients with a positive ESS sep was similar to those of class I patients (hazard ratio, 1.38 [95% CI, 0.66–2.88]; P =0.396). Conclusions: Strain assessment of the septum (SLICE-ESS sep ) provides a prognostic measure for clinical outcome after CRT. Detection of a positive SLICE-ESS sep in patients with a class II indication predicts improved CRT outcome similar to those with a class I indication whereas SLICE-ESS sep negative patients have poor prognosis after CRT implantation.