TEST TO DIAGNOSE, MONITOR OR STRATIFY DISEASES DIRECTLY OR INDIRECTLY ASSOCIATED WITH THE PATHOLOGIES OF THE CHOLINERGIC SYSTEM

Abstract

A test to diagnose, monitor or stratify diseases directly or indirectly associated with cholin ergic system pathologies, including Parkinson's disease, which involves a comparison of supplied pupil light reflex data against known normative data of the same kind, characterised by the fact that the supplied data contain the pupil light reflex data, which include at least one sample comprising at least one supplied parameter measured by a known device used to measure pupil light reflexes, and data which include at least one characteristic of the examined patient. The probability of disease occurrence is determined using machine learning algorithms, which comprise at least one neuronal network algorithm (SSN) and/or at least one mathematical function which is not a part of the neuronal network (SSN).

Claims

1-2. (canceled)

3. A method to diagnose Parkinson's disease, comprising comparing data for a patient comprising supplied pupil light reflex data comprising at least one sample of pupil light reflex data for said patient against a training data set comprising pupil light reflex data for known healthy and ill individuals, the method performed with the use of at least one computer device, wherein the patient data comprise pupil light reflex data comprising at least one supplied parameter measured by a known device used to measure pupil light reflexes, and the at least one characteristic of the examined patient comprising the patient's age and sex, wherein the method comprises, when at least one empty value occurs in the patient data, substituting the missing value with an average value from the training data set, and the pupil light reflex data is used to identify specific parameters comprising: the initial pupil diameter (R1), latency of the onset of constriction (T1), minimal pupil diameter (R2), amplitude (R2-R1), maximum constriction velocity, maximum constriction acceleration, and time to maximal constriction, wherein the method comprises analysing the patient data using one or more machine learning algorithms comparing measured values against training data acquired from patients with Parkinson's disease and healthy individuals in order to determine the probability of Parkinson's disease in the patient, wherein the machine learning algorithms comprise a logistic regression model, and wherein the method comprises determining a borderline point for the logistic regression model describing the probability of disease occurrence using a ROC, and wherein the method further comprises calculating the effectiveness of prediction of the probability of disease occurrence by the machine learning algorithm using a confusion matrix and/or by evaluating the precision and the FI score of the machine learning algorithm using a test sample.

4. The method of claim 3, wherein the specific parameters are selected from: initial pupil diameter (R1), latency of the onset of constriction (T1), minimal pupil diameter (R2), amplitude (R2-R1), maximum constriction velocity, maximum constriction acceleration, and time to maximal constriction.

5. The method of claim 4, wherein the specific parameters include all of: initial pupil diameter (R1), latency of the onset of constriction (T1), minimal pupil diameter (R2), amplitude (R2-R1), maximum constriction velocity, maximum constriction acceleration, and time to maximal constriction.

6. The method of claim 3, wherein the at least one characteristic of the examined patient comprises the patient's age and/or sex.

7. The method of claim 3, wherein the machine learning model comprises a logistic regression model.

8. The method of claim 3, wherein the specific parameters are determined using a pupillometer, through stimulation with light and/or other external stimuli.

9. The method of claim 3, wherein the specific parameters are determined using a video of the patient's pupil constricting during its reaction to light.

10. The method claim 3, wherein the method comprises: calculating the effectiveness of prediction of the probability of disease occurrence by a machine learning algorithm using a confusion matrix and/or by evaluating the precision and the FI score of the machine learning algorithm using a test sample.

11. The method of claim 3, wherein at least one parameter is derived from the pupil center position.

12. The method of claim 3, wherein at least one processing unit improves the pupil diameter measurement accuracy through at least one machine learning model.

13. The method of claim 3, wherein at least one processing unit improves the pupil diameter measurement accuracy through at least one mathematical operation such as deconvolution.

14. The method of claim 3, wherein a video of a patient is acquired from at least 40 cm distance enabling capturing of the patients face and at least one eye.

15. The method of claim 3, wherein at least one parameter is inferred from the conscious or unconscious eyeball movements.

Description

BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS

[0017] The invention was presented in a drawing in which:

[0018] FIG. 1 shows the tree describing the functioning of the diagnostic mechanism;

[0019] FIG. 2. shows the comparison of part of the parameters considered at the examination and their comparison between healthy and ill individuals, and the analysis of the deviation from the healthy norm for those parameters;

[0020] FIG. 3. shows the ROC curve and the analysis of model functions describing the assignment to a disease with different borderline points;

[0021] FIG. 4. shows a scheme of main elements of basal nuclei (basal ganglia) and their elements.

DESCRIPTION OF SPECIFIC EMBODIMENTS OF THE INVENTION

[0022] According to the model representation of the invention development, in order to create the diagnostic test, the averaged PLR values distributed according to the specifications of the Gaussian distribution were used, with 122 coming from PD patients and 101 from healthy controls.

Example: Statistical Analysis of Data

[0023] The PD diagnostic test includes stages (a) supplying at least one sample containing at least one parameter measured by the device used to measure the PLR and at least one characteristic of the examined patient. (b) Used data are subsequently, or in parallel, subjected to cleaning where any empty values are substituted with average values from the training data set. In the case of categorical data, it is substituted with the most commonly occurring sample or by at least one other sample that could be significant in the diagnostic process. (c) Generated or acquired data are identified and assigned to specific values. With the help of a suitable tool, such as for example a pupillometer, through stimulation with light and/or other external stimuli, the R1 (initial pupil diameter) and R2 (minimal pupil diameter) parameters as well as the amplitude (R2-R1), maximum velocity (Vmax mm/s), maximum acceleration (ACmax mm/s{circumflex over ( )}2) and other parameters that can be calculated based on the video of the pupil constricting during its reaction to light, are obtained. Additionally, directly from the patient or from their medical record, information about their sex, age, race, prescribed medication and/or diseases which may influence the functioning of the diagnostic test, and any other information directly and/or indirectly related to the patient that may be significant clinically and/or technically is obtained. (d) Obtained data are analysed using at least one mathematical function.

[0024] For the purpose of calculation, deep learning multi-layer neuronal net is used and the error value is calculated based on at least one mathematical model, such as XGBoost, Logistic Regression or a deep learning neuronal net algorithm which consists of at least one hidden layer. All used models have the weight accuracy assigned based on the F1 score. The F1 score is calculated by multiplying by 2 the sensitivity/recall (the ratio between the true positive results to the sum of the true positive and falsely negative results), divided by precision/positive predictive value (equal to the proportion of true positive results among all positive results) divided by the sum of precisions and sensitivities.

F1.sub.score=2×sensitivity×precision/sensitivity+precision

[0025] Such value will determine the error of each analysed model. (e) The resultant set of functions, comprising at least one mathematical function which includes a borderline point which describes the probability of disease occurrence. The borderline point is determined using the area under the curve ROC (AUC), and effectiveness is calculated using the cross table of the diagnostic test (confusion matrix) and/or by evaluating the tested data set (e.g. precision, F1 score, etc.) Currently, doctors use an assessment framework to determine presence of PD which includes: [0026] 1) Assessment framework based on identification of patient characteristics typical for this disease. [0027] Sensitivity: 74% [0028] Specificity: 48% [0029] 2) Following this invention, the framework of assessing the patient based on the algorithm using the PLR data. (Using a healthy control cohort of 45 individuals and 178 PD patients). [0030] Sensitivity: 96% [0031] Specificity: 94%

[0032] FIG. 1. shows the tree describing the functioning of the diagnostic mechanism, especially the sequence of actions completed by the algorithm during the diagnostic decision-making.

[0033] FIG. 2. shows the comparison of some of the parameters considered at the examination and their comparison between healthy and ill individuals, and the analysis of the deviation from the healthy norm for those parameters. In the figure, there is a comparison between 9 parameters. Light grey colour denotes ill individuals and black colour denotes healthy individuals. [0034] Sex [0035] Age [0036] R1 (initial pupil diameter) [0037] T1 latency [0038] R2 (minimal pupil diameter) [0039] Amplitude (R2-R1) [0040] Maximum velocity (mm/sec) [0041] Maximum acceleration (mm/sec{circumflex over ( )}2) [0042] Time to maximal constriction

[0043] In the opposite corner, there is a distribution of samples from healthy and ill individuals for each parameter, while in individual squares there are samples from parameter from the x axis and the y axis, subdivided between healthy and ill individuals. The larger the distance, the stronger the relationship and more significant is given parameter at detecting the pathology of the cholinergic system.

[0044] FIG. 3. shows the ROC curve and the analysis of model functions describing the assignment to a disease with different borderline points. The curve describes true values of the true positives for different borderline points of probability from 0 where all samples are assigned as negative to 1 where all samples are assigned as positive.

[0045] FIG. 4. shows a scheme of the main elements of the basal nuclei and their relative connections where the glutaminergic pathways stem from the thalamus to cortex, from cortex to stratum, from subthalamic nucleus to internal globus pallidus, from subthalamic nucleus to substantia nigra pars compacta to striatum, and where all the remaining pathways are GABAergic. Importantly, a significant loss of cholinergic neurons of the forebrain was confirmed in the brain affected by PD (Whitehouse et al., 1983; Candy et al., 1983). The identification of a larger loss of forebrain neurons in PD patients than in Alzheimer's patients made Arendt et al. (1983) suggest that deficits in the central cholinergic system may be as pronounced in PD as in AD. This explains the correlation between the neurodegeneration occurring in substantia nigra pars compacta in PD and the pathologies of the cholinergic system, based on which the diagnosis is made.

[0046] Table 1

[0047] Drugs which may affect the pupil relexes and their mechanism of affecting the reflex, Kelbsch C, Strasser T, Chen Y, Feigl B, Gamlin P D, Kardon R, Peters T, Roecklein K A, Steinhauer S R, Szabadi E, Zele A J, Wilhelm H and Wilhelm B J (2019) Standards in Pupillography. Front. Neurol. 10:129. DOI: 1.0.3389/fneur.2019.00129.

TABLE-US-00001 Drug Mechanism Pupil TOPICAL Pilocarpine cholinergic miosis.sup.a Carbachol cholinergic miosis.sup.a Aceclidine cholinergic miosis.sup.a Atropine anticholinergic mydriasis.sup.b Scopolamine anticholinergic mydriasis.sup.b Tropicamide anticholinergic mydriasis Phenylephrine α.sub.1-adrenoceptor agonist mydriasis Metoxamine α.sub.1-adrenoceptor agonist mydriasis Apraclonidine α.sub.1-adrenoceptor agonist mydriasis.sup.c Dapiprazole α.sub.1-adrenoceptor antagonist miosis Brimonidine α.sub.2-adrenoceptor agonist miosis.sup.d Cocaine noradrenaline uptake inhibitor mydriasis SYSTEMIC Antihistamines H1 histamine receptor antagonists miosis.sup.e ANTIHYPERTENSIVES Prazosin α.sub.1-adrenoceptor agonist miosis.sup.f Clonidine α.sub.2-adrenoceptor agonist miosis.sup.g ANTIARRYTHMICS Disopyramide anitcholinergic mydriasis DRUGS FOR PARKINSON'S DISEASE Anticholinergics blockade of muscarinic receptors mydriasis.sup.h Dopaminergics stimulation of D2 dopamine receptors mydriasis.sup.i ANTIDEPRESSANTS Tricyclic mainly noradrenaline uptake blockade mydriasis.sup.j Reboxetine noradrenaline uptake blockade mydriasis Venlafaxine noradrenaline/serotonin uptake blockade mydriasis SSRIs serotonin uptake blockade no effect.sup.k ANTIPSYCHOTICS Phenothiazines α.sub.1-adrenoceptor antagonist, sedation miosis.sup.l Haloperidol α.sub.1-adrenoceptor antagonist miosis SEDATIVES benzodiazepines GABA receptor agonist .fwdarw. sedation no effect.sup.m PSYCHOSTIMULANTS Amphetamine noradrenaline releaser mydriasis Modafinil dopamine uptake blocker mydriasis.sup.n ANALGESICS Opiates stimulation of inhibitory μ receptors miosis.sup.o ANTIEMETICS Scopolamine anticholinergic mydriasis ANTI-INCONTINENCE DRUG anticholinergic mydriasis.sup.p .sup.aglaucoma treatment. .sup.bmyopia treatment. .sup.cin Horner's syndrome (supersensitive α.sub.1-adrenoceptors). .sup.ddrugreduces noradrenaline release (glaucoma treatment). .sup.efirst generation antihistamines (e.g. diphenhydramine, cyclizine) penetrate into the brain where they block H1 histamine receptors, leading to sedation. .sup.fdrug blocks α.sub.1-adrenoceptors in vascular smooth muscle. .sup.gdrug stimulates inhibitor α.sub.2-adrenoceptors on central noradrenergic neurones, leading to sedation and sympatholysis. .sup.hinclude orphenadrine, procyclidine, trihexyphenidyl. .sup.iD2 dopamine receptor agonists (e.g. pramipexole) stimulate inhibitory D2 receptors on wake-promoting centrl dopaminergic neurones, leading to sedation. This is expected to cause miosis, however, paradoxically, pramipexole causes mydrasis. .sup.jTricyclic antidepressants block the uptake of noradrenaline, potentiating noradrenergic neurotransmission, and this would lead to mydriasis. However, they have some other effects: blockade of muscarinic cholinoceptors would lead to mydriasis and sedation, and blockade of α.sub.1-adrenoceptors would cause miosis. The overall effect reflects the balance between these actions: mydrasis due to noradrenaline uptake blockade and cholinoceptor blockade in counteracted by miosis due to α.sub.1-adrenoceptor blockade and sedation. This explains the variable effects of tricyclic antidepressants on the pupil: imipramine desipramine dilate it, while amitriptyline has little effect on it. .sup.kSelective serotonin reuptake inhibitor (SSRIs) block serotonin receptors in a complex network of serotonergic neurones associated with different excitatory/inhibitory receptors. The overall effect is little or no change in pupil diameter. .sup.lThese drugs (e.g. chlorpromazine, trifluoperazine) also have anticholinergic effects that would lead to mydrasis. However, α.sub.1-adrenoceptors blockade and sedation predominate, leading to miosis. .sup.mParadoxically, alhough the benzodiazepine diazepam is highly sedative, it has no effect on pupil diameter. .sup.nModafinil blocks dopamine uptake at excitatory synapses on central noradrenergic neurones: this leads to increase in arousal and sympathetic activity. .sup.oStimulation of inhibitory receptors on central noradrenergic neurones leads to sedation and sympatholysis. .sup.pThese drugs (oxybutynin, festerodine) inhibit voiding of the urinary bladder by blocking cholinceptors in the detrusor muscle.