Method for creating a predictive model for predicting glaucoma risk in a subject, method for determining glaucoma risk in a subject using such predictive model, device for predicting glaucoma risk in a subject, computer program and computer readable medium

A system for predicting glaucoma development and/or progression risk in a subject using a predictive model for predicting glaucoma development and/or progression risk, comprising:

• an eyeball parameters measuring and recording device configured to obtain eyeball parameters continuously or at intervals over a 24-hour period, wherein the eyeball parameters include a circumference at a corneoscleral limbus of an eyeball and/or an intraocular pressure;
• a cardiovascular system parameters measuring and recording device configured to obtain cardiovascular system parameters continuously or at intervals over the 24-hour period, the cardiovascular system parameters including a systolic arterial pressure, a diastolic arterial pressure, a mean arterial pressure, and a heart rate;
• a control circuit having a communication connection with the devices;
• a processor installed in the control circuit;
• a memory installed in the control circuit and operatively coupled to the processor;
• an output device for presenting results having a communication connection with the control circuit;
• wherein the processor is configured to execute a program code stored in the memory for performing the following steps based on a set of measurement data series provided by the devices, the set of measurement data series being for a plurality of subjects:
• a step a of applying for each of the measurement data series a procedure of dividing the recorded 24-hour profile of the eyeball parameters, provided by the eyeball parameters measuring and recording device, wherein said procedure comprises:
• identifying, for the recorded 24-hour profile of the eyeball parameters, reference time points including:
• session start (START) and session end (END) corresponding, respectively, to the beginning and the end of the recorded 24-hour profile, as well as a reference point indicating assuming a horizontal position for sleep (SLEEP) and a reference point indicating assuming a vertical position after sleep (WAKE);
• identifying, for the recorded 24-hour profile of the eyeball parameters, and with reference to the identified reference time points (START, SLEEP, WAKE, END), time points (TPI, TP2, TP3) such that at least the following subperiods are defined:
• an initial subperiod (START—TP) from session start to 5 hours before assuming a horizontal position for sleep;
• a subperiod preceding assuming a horizontal position for sleep (TPI-SLEEP) from 5 hours before assuming a horizontal position to assuming a horizontal position for sleep;
• a subperiod following assuming a horizontal position for sleep (SLEEP—TP2) from assuming a horizontal position for sleep to assuming a horizontal position for sleep+2 hours;
• a subperiod preceding assuming a vertical position after sleep (TP2—WAKE) from assuming a horizontal position for sleep+2 hours to assuming a vertical position after sleep;
• a subperiod following assuming a vertical position after sleep (WAKE—TP3) from assuming a vertical position after sleep to assuming a vertical position after sleep+2 hours;
• a final subperiod (TP3—END) from assuming a vertical position after sleep+2 hours to the session end;
• wherein the above division is chosen to reflect relationships between the eyeball parameters and the cardiovascular system parameters characteristic for a pathomechanism of glaucoma development;
• a step b) of reducing each of the measurement data series, for each one of the plurality of subjects, by transformations including determining, in each subperiod, features describing a single subject in the form of one or more aggregating attributes, wherein the one or more aggregating attributes are selected from the group consisting of:
• a sum of the area under the curve in a subperiod, a slope angle of a linear regression line in a subperiod, a total variation in a subperiod, representative values of the discrete Fourier transform in a subperiod, and calculating correlations between the eyeball parameters and the cardiovascular system parameters, in the form of correlation parameters, provided by the cardiovascular system parameters measuring and recording device;
• a step c) of creating a plurality of records each corresponding to one of the plurality of subjects, each record containing the one or more aggregating attributes and the calculated correlation parameters corresponding to said one of the plurality of subjects;
• a step d) of collecting a set of training records, labelled for creating the predictive model for predicting glaucoma development and/or progression risk wherein collecting the set of training records includes assigning, to each of the created records in step c) using the processor, a label indicating a glaucoma diagnosis (glaucomatous neuropathy positive/negative) based on predetermined medical data for a respective subject of the plurality of subjects;
• and
• a step e) of creating the predictive model for predicting glaucoma development and/or progression risk, based on the set of training records created for the plurality of subjects, using supervised machine learning mechanisms based on one or more algorithms selected at least from regression algorithms, decision trees, Bayesian algorithms, ensemble algorithms and support vector-based algorithms, wherein the created predictive model reflects correlations between the eyeball parameters and the cardiovascular system parameters characteristic for a pathomechanism of glaucoma development in each defined subperiod used for training the predictive model;
• wherein the set of training records are selected to control the creation of the predictive model by satisfying control parameters of sensitivity >0.85 and AUC >0.85 such that it provides high sensitivity to individual variability and an individual biorhythm of subjects, and embedding the predictive model for predicting glaucoma development and/or progression risk in the device memory;
• wherein the processor is further configured to perform an additional step:
• the created predictive model receiving measurement data series for an individual subject to be examined, wherein said measurement data series are next processed in the same way as described in the steps a) to c) to produce an individual subject record containing the one or more aggregating attributes and the calculated correlation parameters corresponding to said one of the plurality of subjects created in step c) and outputting a probability of said record to belong to a group of glaucomatous neuropathy development and/or progression risk positive or glaucomatous neuropathy development and/or progression risk negative subjects, wherein an output of the predictive model indicating glaucomatous neuropathy development and/or progression risk for the examined subject is generated by processing, using the predictive model, the individual subject record including correlations between the eyeball parameters and the cardiovascular system parameters characteristic for a pathomechanism of glaucoma development and/or progression in each defined subperiod for the individual subject.