System and method for estimating a relative substance composition of a portion of a body of a patient

A computer-implemented method for estimating a relative substance composition (RSC) of a portion of a body of a patient in a field of view for a medical image to be taken from the patient in a medical imaging scan, the computer-implemented method comprising:

• receiving at least one piece of patient information data (PPID);
• receiving at least one piece of field-of-view information data (PFID), wherein the PFID is indicative of the field of view with which the medical image is to be taken of the patient;
• inputting the received at least one PPID and the received at least one PFID into a trained machine learning algorithm (MLA), wherein the trained MLA is configured and trained to receive the at least one PPID and the at least one PFID as an input and to generate as an output at least one output signal indicating a relative substance composition (RSC) of the body of the patient for the medical image based on the at least one PPID and the at least one PFID;
• and
• outputting at least the at least one output signal.

A computer-implemented method for training a machine learning algorithm (MLA) for use in a system for estimating a relative substance composition (RSC) of a portion of a body of a patient in a field of view for a medical image to be taken from the patient in a medical imaging scan, the system comprising an input interface configured to receive at least one piece of patient information data (PPID) and receive at least one piece of field-of-view information data (PFID), wherein the PFID is indicative of the field of view with which the medical image is to be taken of the patient, a computing device configured to implement a trained machine learning algorithm, wherein the trained MLA is configured and trained to receive the at least one PPID and the at least one PFID received by the input interface as input, and generate as output at least one output signal indicating an RSC of a portion of the body of the patient for the medical image based on the at least one PPID and the at least one PFID, and an output interface configured to output at least the at least one output signal, the computer-implemented method comprising:

• providing training samples, each of the training samples comprising a set of input parameters comprising one or more PPID and one or more PFID in accordance with pieces of information that the MLA is configured to receive, and each of the training samples being labelled with a corresponding RSC;
• and
• training the MLA with supervised learning using the provided training samples.

A method for providing training data for use in a computer-implemented method for training a machine learning algorithm (MLA), the method comprising:

• providing at least one family member of a virtual family, wherein each family member of the at least one family member comprises a plurality of voxels for which an individual relative substance composition (RSC) is known;
• virtually positioning each family member of the at least one family member in a plurality of positions and with a plurality of fields-of-view with regard to a medical imaging scan;
• determining, for each position of the plurality of positions and field-of-view of the plurality of fields-of-view for each family member of the at least one family member of the virtual family, a total RSC for a respective field of view based on the individual RSCs of the voxels of the family member in the respectively field of view at the respective position;
• generating sets of input parameters comprising at least one PPID and at least one PFID based on the at least one family member, the plurality of positions, and the plurality of fields-of-view;
• labelling the generated sets of input parameters with the corresponding determined total RSC, such that training samples are generated.

In a non-transitory computer-readable storage medium that stores instructions executable by one or more processors to estimate a relative substance composition (RSC) of a portion of a body of a patient in a field of view for a medical image to be taken from the patient in a medical imaging scan, the instructions comprising:

• receiving at least one piece of patient information data (PPID);
• receiving at least one piece of field-of-view information data (PFID), wherein the PFID is indicative of the field of view with which the medical image is to be taken of the patient;
• inputting the received at least one PPID and the received at least one PFID into a trained machine learning algorithm (MLA), wherein the trained MLA is configured and trained to receive the at least one PPID and the at least one PFID as an input and to generate as an output at least one output signal indicating a relative substance composition (RSC) of the body of the patient for the medical image based on the at least one PPID and the at least one PFID;
• and
• outputting at least the at least one output signal.