Technique for quantifying a cardiac function from CMR images

A method for outputting a value indicative of an ejection fraction (EF) of a cardiac chamber of a heart based on a temporal sequence of cardiac magnetic resonance (CMR) images of the cardiac chamber using a neural network system, the method comprising:

• receiving, at an input layer of the neural network system, the temporal sequence of a stack of slices of the CMR images along an axis of the heart, the temporal sequence comprising one or multiple consecutive cardiac cycles of the heart;
• and
• outputting, at an output layer of the neural network system, the value indicative of the EF of the cardiac chamber based on the temporal sequence of the stack of slices of the CMR images along the axis of the heart,
• wherein the neural network system has interconnections between the input layer and the output layer;
• and
• wherein the neural network system was trained with a plurality of datasets, each of the datasets comprising an instance temporal sequence of a stack of slices of CMR images along the axis over one or multiple consecutive cardiac cycles for the input layer and an associated instance value indicative of the EF for the output layer.

A non-transitory computer readable storage medium having stored thereon program code executable by a computer for outputting a value indicative of an ejection fraction (EF) of a cardiac chamber of a heart based on a temporal sequence of cardiac magnetic resonance (CMR) images of the cardiac chamber using a neural network system, the storage medium comprising the computer code for:

• receiving, at an input layer of the neural network system, the temporal sequence of a stack of slices of the CMR images along an axis of the heart, the temporal sequence comprising one or multiple consecutive cardiac cycles of the heart;
• and
• outputting, at an output layer of the neural network system, the value indicative of the EF of the cardiac chamber based on the temporal sequence of the stack of slices of the CMR images along the axis of the heart,
• wherein the neural network system has interconnections between the input layer and the output layer;
• and
• wherein the neural network system was trained with a plurality of datasets, each of the datasets comprising an instance temporal sequence of a stack of slices of CMR images along the axis over one or multiple consecutive cardiac cycles for the input layer and an associated instance value indicative of the EF for the output layer