Relationship of Image Quality to Decision Accuracy for ARVC/D in MRI Elizabeth Krupinski, PhD Theron Ovitt, MD University of Arizona
ARVC/D Arrhythmogenic right ventricular cardiomyopathy / dysplasia RV muscle replaced by fatty &/or fibrous tissue Progressive arrhythmic death
ARVC/D Most common 15 – 40 Premature ventricular beats Ventricular tachycardia left bundle branch block morphology Sudden cardiac death uncommon but may be 1 st sign Familial 30 – 50% cases
Use of MRI Differential diagnosis - RV outflow tract tachycardia (RVOT) Characterizes fat in free wall by signal intensity Shows contraction abnormalities of RV
Rationale Acquisition protocols vary Image quality not studied Artifacts not studied Reader performance not studied
Cases 45 static MRI films – ARVC/D registry UofA – 7 met Task Force criteria – 31 suspected clinical signs – 7 imaged other reasons
Readers 12 experienced cardiac MRI radiologists 1 cardiologist expertise in MRI 13 different institutions US, Canada, Europe
Reading Protocol Cases randomly 4 groups Mailed independently ~ 1 month between mailings Pre-formatted response sheet 6 clinical questions 3 image quality questions
Image Quality Rating Task Force Suspect ARVC/D No ARVC/D Excellent 7% 8% Good 37% 38% 45% Fair 44% 42% 38% Poor 12% 13% 9%
Motion Artifacts Rating Task Force Suspect ARVC/D No ARVC/D None 22% 16% 15% Mild 54% 43% 32% Moderate 22% 32% 45% Severe 9% 8% 2%
Other Artifacts Rating Task Force Suspect ARVC/D No ARVC/D None 52% 53% 65% Mild 13% 17% 25% Moderate 30% 22% 9% Severe 5% 8% 1%
IQ & Motion Artifacts Inverse relationship – No mild = good excellent – Moderate severe = fair poor X 2 = , df = 9, p <
IQ & Other Artifacts Inverse relationship – No mild = good excellent – Moderate severe = fair poor X 2 = , df = 9, p <
ARVC/D Decisions ARVD Present ARVD Absent Task Force 47% 53% Suspected 39% 61% No ARVC/D 29% 71% X 2 = 6.756, df = 2, p = 0.034
Decisions x IQ PoorFairGoodExcel TP 8% 43% 42% 7% FN 16% 42% 35% 7% FP 8% 42% 46% 4% TN 10% 35% 45% 10% X 2 = df = 9 p =
Reader Agreement 1/45 (2%) complete agree – No ARVD – All agree absent Kappa = – Very low – Poor discriminability
Other Indicators If present => ARVC/D present – Fat in myocardium – Enlarged RV chamber size – RV configuration abnormal
Other Indicators No relationship with ARVC/D present – LV chamber size – Myocardium fat location Anterior wall RV OT RV apex
Conclusions Inter-observer variability high ARVC/D absent > present Task Force = Suspected Certain signs probability ARVC/D reported present but inconsistent
Conclusions No established imaging protocol – Reduces IQ – Optimal views? – Motion & other artifacts Different MRI equipment Lack cine for wall motion abnormalities