Synthetic hematocrit estimation for myocardial ECV


MR image-based synthetic hematocrit estimation for myocardial ECV




Define-AI ID



Akos Varga-Szemes


Akos Varga-Szemes

Panel Chair

Carlo N. De Cecco

Panel Reviewers

Cardiac Panel


Creative Commons 4.0 


Public Comment

Clinical Implementation

Value Proposition

Cardiac magnetic resonance imaging (MRI) based extracellular volume (ECV) assessment has been shown to have clinical value for the evaluation of various myocardial pathologies including myocarditis and diffuse myocardial involvement in amyloidosis. ECV calculation requires native and post-contrast T1 measurements of the blood and the myocardium, as well as the level of blood hematocrit (Hct). Considering that inaccurate Hct may significantly influence ECV and the fluctuation in Hct may reach ±8% within 24 hours, it is particularly important to perform the Hct measurement in the same day and as close to the MRI study as possible. Hct measurements are usually available through blood sampling or using a point-of-care (POC) device. The majority of patients arriving for a cardiac MRI, however, do not have a same day lab test and radiologists do not order labs in general. The availability of POC devices is also not widespread.

Recent research indicates that Hct can also be calculated without the need for any blood sampling using the native R1 (inverse T1) value of the blood. Such method assumes a linear correlation between blood native R1 and Hct, and the equation derived from the correlation can be used to estimate synthetic Hct as long as the MRI is performed on the same scanner platform using the same pulse sequence variant and scheme. While the proposed approach would be promising, its inaccuracy and the consequent deviation in ECV has been shown in multiple studies. Such inaccuracy can be related to other factors that may influence blood R1 independently from Htc.

An AI algorithm would be able to address the inaccuracy of Hct estimation by accounting for additional features beyond R1 and estimate blood Hct levels in order to avoid blood sampling prior to cardiac MRI studies involving ECV analysis.


All patients undergoing MRI-based ECV assessment will have automated Htc assessment, therefore Htc input will not be necessary and the ECV calculation process can be fully automated.

Workflow Description

  1. Patient undergoes contrast enhanced MRI of the heart with native and post-contrast T1 mapping included

  2. Algorithm retrieves inputs:

    1. Image data

    2. Patient characteristics that may influence native T1 measurements: age, gender, BMI, heart rate

    3. Acquisition parameters that may influence native T1 measurements: vendor, field strength, pulse sequence type and scheme

  3. Algorithm executes:

    1. Automatically segments the LV blood compartment on native T1 datasets (another segmentation algorithm can perform this action)

    2. Measures the average blood T1 and calculates R1

    3. Algorithm takes into account native R1 and other input parameters to estimate Hct

    4. Algorithm outputs Hct and feeds it into an ECV calculation formula

Considerations for Dataset Development

Age: Varied

Sex at birth: Male, Female

Blood disorders: Varied

BMI: Varied

Heart rate: Varied

Scanner Vendor: Varied

Field Strength: [0.2 T, 7T]

Pulse sequence type: {spin echo sequences, inversion recovery sequences, gradient echo sequences, diffusion weighted sequences, saturation recovery sequences, echo-planar pulse sequences, spiral pulse sequences}

Pulse sequence scheme: Varied

Technical Specifications






Data Type




Body Region


Anatomic Focus




Primary Outputs


RadElement ID

to be defined


measure of the red blood cells in a patient’s blood

Data Type


Value Set



percentage of red blood cells in sample

Future Development Ideas

  • integrate algorithm with automated ECV calculation