AI-AIF: artificial intelligence-based arterial input function for quantitative stress perfusion cardiac magnetic...

Area of Science:

• Cardiovascular Imaging
• Artificial Intelligence in Medicine
• Biomedical Engineering

Background:

• Quantifying myocardial blood flow (MBF) using stress perfusion cardiac magnetic resonance (CMR) is hindered by signal saturation effects in the arterial input function (AIF).
• The non-linear relationship between gadolinium concentration and MR signal complicates accurate AIF estimation, impacting MBF quantification.
• Accurate AIF is critical for reliable MBF measurements in CMR, especially in clinical settings.

Purpose of the Study:

• To develop and validate a deep learning model for predicting an unsaturated arterial input function (AIF) from standard cardiac magnetic resonance (CMR) images.
• To enable accurate quantification of myocardial blood flow (MBF) using a single-sequence acquisition and a single contrast injection.
• To overcome signal saturation challenges in AIF estimation for stress perfusion CMR.

Main Methods:

• A 1D U-Net deep learning model was trained to predict unsaturated AIF from saturated AIF derived from standard CMR images.
• Training data utilized reference dual-sequence acquisition AIFs (DS-AIFs) from 201 patients.
• The model was tested on independent cohorts from two centers (n=44), comparing AI-AIF derived MBF with DS-AIF quantified MBF.

Main Results:

• The AI-AIF method showed no statistically significant difference in MBF quantification compared to the DS-AIF method (2.77 vs. 2.79 mL/min/g, P=0.33).
• Bland-Altman analysis revealed minimal bias (-0.11 mL/min/g) between AI-AIF and DS-AIF for quantitative MBF.
• MBF diagnosis classification using AI-AIF agreed with DS-AIF in 95% of myocardial segments.

Conclusions:

• Stress perfusion CMR for MBF quantification is feasible using a single-sequence acquisition and single contrast injection.
• An AI-based correction of the arterial input function (AIF) effectively addresses signal saturation issues.
• This deep learning approach facilitates accurate and efficient MBF assessment in clinical practice.