The invention discloses a
valvular heart disease severity AI grading diagnosis method, which comprises the following steps of executing multi-
source data acquisition, and acquiring a clinical
medical record text,
cardiac ultrasound image data and a
valvular heart disease interventional therapy prognosis
data set of a patient to be diagnosed; wherein the clinical
medical record text comprises patient subjective symptom description information, and the
interventional therapy prognosis
data set comprises post-treatment symptom
remission rate, complication occurrence rate and lifetime data of patients with different severity levels. The limitation that the prior art only depends on single objective image data is broken through; according to the method, symptom quantitative
processing and targeted image
feature extraction are combined, comprehensive capturing of multi-dimensional features is achieved, then a grading threshold value is reversely calibrated through a
prognosis prediction result, grading deviation is effectively corrected, and the accuracy and clinical matching degree of grading diagnosis are remarkably improved.