Diagnostic accuracy of quantitative perfusion CMR for detecting coronary artery disease: comparison of model-based deconvolution techniques

Presenting Author(s)

• NH

  Naoki Hashimoto, MD

  Doctor
  Mie University Hospital
  Tsu, Mie, Japan

Primary Author(s)

• NH

  Naoki Hashimoto, MD

  Doctor
  Mie University Hospital
  Tsu, Mie, Japan

Co-Author(s)

• 

  Masaki Ishida, MD, PhD, FSCMR

  Associate Professor
  Mie University Hospital
  Tsu, Mie, Japan
• MT

  Masafumi Takafuji, MD, PhD

  Assistant Professor
  Mie University Hospital
  Tsu, Mie, Japan
• HI

  Haruno Ito, MD, PhD

  Fellow
  Mie University Hospital
  Tsu, Mie, Japan
• TK

  Takanori Kokawa, MD

  Assistant Professor
  Mie University Hospital, Japan
• SA

  Suguru Araki, MD

  Fellow
  Mie University Hospital
  Tsu, Mie, Japan
• SY

  Shintaro Yamaguchi, MD

  Resident
  Mie University Hospital, Japan
• MF

  Miyuko Fujita, MD

  Resident
  Mie University Hospital, Japan
• SN

  Satoshi Nakamura, MD, PhD

  Associate Professor
  Mie University Hospital
  Tsu, Mie, Japan
• TK

  Tairo Kurita, MD, PhD

  Associate Professor
  Mie University Hospital
  Tsu, Mie, Japan
• KD

  Kaoru Dohi, MD, PhD

  Professor
  Mie University Hospital
  Tsu, Mie, Japan
• 

  Hajime Sakuma, MD, PhD

  Professor
  Mie University Hospital
  Tsu, Mie, Japan

Background: Quantitative perfusion (QP) CMR allows for quantification of myocardial blood flow (MBF) and myocardial perfusion reserve (MPR) and serves as an alternative approach to visual assessment in the management of coronary artery disease (CAD). Recently, with the advent of dual-sequence T1 saturation correction and automated post-processing pipeline [1], pixel-wise MBF mapping has become available at the clinical level utilizing model-based deconvolution such as Fermi function, Tofts and modified Tofts models [2,3]. However, few studies have compared the diagnostic accuracy of QP CMR across different model-based deconvolution techniques. The purpose of this study was to compare the diagnostic accuracy of QP CMR among three model-based deconvolutions (Fermi function, Tofts and Modified Tofts) against invasive coronary angiography (CAG).

Methods: Sixty-two patients (44 men, 67±17 years) with known or suspected CAD who underwent both CMR and CAG within 3 months were studied. CMR protocol included cine, stress and rest dual-sequence QP, and late gadolinium-enhancement CMR (Elision 3.0T, Philips). QP CMR was conducted with standard gadolinium contrast injection scheme (0.05 mmoL/kg, Gd-DOTA). Pharmacological stress was achieved using adenosine triphosphate (160 μg/kg/min). MBF maps were generated inline and analyzed offline. Mean MBF were calculated using a 32-subsegment scheme derived from the AHA 16 segments. To assess the diagnostic performance of stress QP CMR, the lowest stress MBF and MPR in the absence of infarction within each coronary territory were used. The significant CAD was defined as luminal diameter reduction of >70% on quantitative CAG.

Results: Thirty of 186 (16.1%) vessels had significant CAD. In the vessel-based analysis for stress MBF, the area under the ROC curve (AUC), sensitivity, specificity and cut-off were 0.88, 83%, 92% 1.98 mL/min/g for Fermi function; 0.78, 70%, 79%, 2.04 mL/min/g for Tofts; 0.70, 73%, 62%, 1.94 mL/min/g for modified Tofts, respectively. For MPR, AUC, sensitivity, specificity and cut-off were 0.81, 71%, 90%, 1.46 for Fermi function; 0.75, 50%, 92%, 1.30 for Tofts; 0.73, 53%, 85%, 1.16 for modified Tofts, respectively. AUC of Fermi function was significantly higher than that of Tofts and modified Tofts (p< 0.05) for stress MBF (Figure 1) and than Tofts for MPR (p< 0.05) (Figure 2). AUC of MPR was higher when the interval of stress and rest was >10min, compared to when the interval was < 10 min (0.68 vs. 0.95 (p=0.02) for Fermi (Figure 3), 0.67 vs. 0.89 (p=0.11) for Tofts, 0.63 vs. 0. 82 (p=0.20) for modified Tofts).

Conclusion: Dual-sequence stress QP CMR with Fermi function model-based deconvolution provides high diagnostic accuracy in the detection of significant CAD and outperform Tofts and modified Tofts models. To utilize MPR, rest perfusion should be performed >10 min after stress perfusion.

Figure 1. ROC curves of stress MBF for detecting significant CAD on vessel-based analysis
Figure 2. ROC curves of MPR for detecting significant CAD on vessel-based analysis
Figure3. ROC curves of MPR with Fermi function model for detecting significant CAD: impact of stress-rest interval (>10min vs. <10 min)