The invention discloses a
muscle quality assessment aided decision-making method based on double-source features, which comprises the following steps: firstly, collecting multi-
modal data of a patient, extracting
modal features, then adopting
time sequence enhancement, obtaining DXA
whole body, DXA leg and
blood test features, and then carrying out
modal correlation calculation, namely, calculating the correlation among the modal features of the DXA
whole body, the DXA leg and
blood test three
modes; then, a
relation graph between
modes is constructed through graph
convolution, mode relation fusion is carried out, finally,
muscle quality prediction is carried out, and
muscle quality evaluation is completed. According to the method, the DXA
whole body and leg images and the double-source features of blood indexes are utilized, multi-modal relation modeling and
time sequence enhancement are combined, high-precision explainable prediction of muscle quality is achieved, the whole body and
leg muscle features can be distinguished,
blood test changes can be dynamically reflected, and auxiliary decision support is provided for
sarcopenia screening.