Placental Blood-Flow Velocity Quantification From Diffusion MRI

Area of Science:

• Biomedical Imaging
• Computational Biology
• Medical Physics

Background:

• Altered placental capillary blood flow is a key indicator of obstetric complications.
• Quantifying blood flow at the capillary scale is challenging due to disordered microvascular beds.
• Existing imaging methods struggle to accurately measure velocity in these complex networks.

Purpose of the Study:

• To develop and validate a novel method for estimating placental capillary blood velocity using diffusion MRI (dMRI).
• To combine Monte Carlo (MC) simulations with machine learning to overcome limitations of current imaging techniques.
• To provide a more accurate assessment of microvascular blood flow in the placenta.

Main Methods:

• Developed MC simulations incorporating perfusion and diffusion to create signal dictionaries.
• Employed supervised machine learning regressors (Random Forest and Multilayer Perceptron) for parameter estimation.
• Applied the trained models to simulated and in vivo dMRI data, comparing with Intravoxel Incoherent Motion (IVIM) methods.

Main Results:

• The MC-based machine learning approach outperformed traditional IVIM methods in simulated data.
• The MC-based Multilayer Perceptron (MLP) model provided physiologically reasonable velocity estimates in vivo.
• The new method demonstrated higher sensitivity in detecting blood flow parameters compared to IVIM.

Conclusions:

• The proposed method offers an alternative to conventional IVIM models for characterizing capillary perfusion, especially in the placenta.
• This technique can potentially aid in identifying pathological placental blood flow conditions, such as preeclampsia and fetal growth restriction.
• The approach may be applicable to other organs where microscale blood flow changes indicate underlying pathologies.