Obstructive sleep apnea syndrome diagnosis method using machine learning

An obstructive sleep apnea syndrome diagnosis method using machine learning, the obstructive sleep apnea syndrome diagnosis method comprising:

• an information elicitation operation of eliciting subject flow characteristic information using an information elicitation machine learning model from three-dimensional (3D) airway shape information of a subject airway;
• and
• a symptom diagnosis operation of eliciting symptom status information using a symptom diagnosis machine learning model from the flow characteristic information elicited in the information elicitation operation and biological characteristic information of the subject;
• wherein the information elicitation operation comprises:
• an information elicitation preparation operation of training the information elicitation machine learning model using training airway shape information of a training airway and training flow characteristic information elicited through computational fluid dynamics (CFD), for a plurality of training airways previously selected for training;
• and
• an information elicitation management operation of eliciting subject flow characteristics information, for at least one newly selected airway for analysis, using the information elicitation machine learning model trained in the information elicitation preparation operation from the plurality of training airways;
• wherein the information elicitation preparation operation comprises:
• a training airway shape information elicitation operation of creating, for each of the plurality of training airways previously selected for training, a 3D model of the training airway from a tomogram of the training airway and eliciting airway shape information of the training airway from the 3D model, for the plurality of training airways previously selected for training, wherein the training airway shape information comprises a length of the training airway, a position of each of a plurality of points spaced apart from each other in a longitudinal direction of the training airway, a diameter of a longer axis at each of the plurality of points, a second diameter of a shorter axis at each of the plurality of points, and a cross-sectional area at each of the plurality of points;
• a training flow characteristic information elicitation operation of eliciting training flow characteristic information through performing a simulation using the CFD by applying a boundary condition to the 3D model of the training airway during the information elicitation preparation operation, wherein the boundary condition comprises a pressure at an inlet or an outlet position of the training airway and a flow rate at the inlet or the outlet position of the training airway, and wherein the flow characteristic information comprises velocity and pressure;
• and
• an information elicitation machine learning model training operation of training the information elicitation machine learning model by using the training airway shape information determined by the training airway shape information elicitation operation as an input parameter and the training flow characteristic information determined by the training flow characteristic information elicitation operation as an output parameter;
• wherein the information elicitation management operation comprises:
• a subject airway shape information elicitation operation of creating, for the at least one newly selected airway for analysis, a 3D model of the newly selected airway from a tomogram of the newly selected airway and eliciting airway shape information of the newly selected airway from the 3D model, for the at least one newly selected airway for analysis using the information elicitation machine learning model;
• and
• a subject flow characteristic information elicitation operation of outputting the subject flow characteristic information by inputting, into the information elicitation machine learning model, the subject airway shape information determined by the subject airway shape information elicitation operation;
• wherein the information elicitation preparation operation is performed while the information elicitation machine learning model is trained, and the information elicitation management operation is performed after the information elicitation machine learning model is trained;
• and
• wherein the symptom diagnosis operation comprises:
• a symptom diagnosis preparation operation of training the symptom diagnosis machine learning model by using the training flow characteristic information of the training airway elicited in the information elicitation operation and biological characteristic information of an individual having the training airway as input parameters, and symptom status information of the individual as an output parameter, for the plurality of training airways previously selected for training among the training airways used in the information elicitation operation, wherein the biological characteristic information comprises age and BMI;
• and
• a symptom diagnosis management operation of outputting the symptom status information of the subject by inputting the subject flow characteristic information of the subject airway determined from the information elicitation management operation and the biological characteristic information of the subject having the subject airway to the symptom diagnosis machine learning model trained in the symptom diagnosis preparation operation, for at least one second airway newly selected for analysis,
• wherein the symptom diagnosis preparation operation is performed while the symptom diagnosis machine learning model is trained, and the symptom diagnosis management operation is performed after the symptom diagnosis machine learning model is trained, and
• wherein the information elicitation machine learning model is trained by performing machine learning by adapting a Gaussian process regression (GPR) algorithm or adapting a multi-variate Gaussian process regression (MV-GP) algorithm during the information elicitation machine learning model training operation, and wherein the symptom diagnosis machine learning model is trained by performing machine learning by adapting a support vector machine (SVM) algorithm during the symptom diagnosis preparation operation.