PSYCHIATRIC DISORDER

CROSS REFERENCE TO RELATED APPLICATIONS

[001] This application claims the benefit of priority of U.S. Provisional Patent Application No. 63/307,844, filed February 8, 2022, the contents of which are all incorporated herein by reference in their entirety.

FIELD OF THE INVENTION

[002] The present invention relates generally to the field of assistive diagnostics. More specifically, the present invention relates to using determining treatment of a psychiatric disorder.

BACKGROUND OF THE INVENTION

[003] Bipolar disorder (BD) is a psychiatric disorder with high morbidity and mortality.

Lithium (Li), a prominent mood stabilizer, is fully effective in roughly 30% of BD patients. The rest of the patients respond partially or do not respond at all. Another drug used in BD is anticonvulsant valproic acid (VPA), also referred to herein as valproate. Many efforts have been made to understand how these drugs affect the patients’ neurons.

SUMMARY OF THE INVENTION

[004] We have performed electrophysiological (EP) recordings in patient-derived dentate gyrus (DG) granule neurons for three groups: control individuals, BD patients who respond to Li treatment, and BD patients who do not respond to Li treatment. The recordings were analyzed by the statistical tools of modem information theory, which enabled us to recognize new relationships between the EP features. These added features included the entropy of several EP measurements and the mutual information between different types of EP measurements and provided further knowledge about the distribution of the EP entities and how these affect each other.

[005] This newly added knowledge enabled a significant improvement in our ability to distinguish the patients from the control individuals and the Li responders from the non- responders using Support Vector Machine (SVM) classification algorithms. [006] Additionally, these new tools allowed us to quantify how neurotypical the patients’ neurons became after in vitro treatment with Li and VP A in a robust manner compared to our previous analysis using just the average and standard deviation of the EP data. This quantification yielded that neurons derived from BD patients who are Li responsive become more neurotypical after chronic Li treatment but not after chronic VP A treatment (despite reducing their hyperexcitability with VPA treatment). The neurons derived from BD Li non-responsive patients became less excitable after treatment with VPA but not neurotypical when observing the complete statistical distributions of the EP recordings.

[007] When using machine-learning (ML) based classifiers or predictors such as random forest or SVM models, we were able to improve the accuracy of prediction by using features that were calculated using information theory; entropy of some of the features and mutual information between the features. These are parameters that we use as features to the classifier, but when adding their entropies as features and also the mutual information between these parameters there is a significant improvement in the performance of the classifier.

[008] We have used this scheme for 2 classifications: In the first one, we classified healthy individuals and BD patients. In the second, we classified BD patients who respond to lithium treatment and the patients who do not respond to lithium treatment. In both these prediction schemes, the prediction improved from approximately 80% accuracy to more than 95% accuracy with the added features of entropy and mutual information (MI).

[009] Embodiments of the invention may include a method of determining treatment of a patient by at least one processor.

[010] According to some embodiments, the at least one processor may obtain, from whole-cell patch clamps, electrophysiological (EP) signals, representing EP activity (e.g., action potentials) in at least one patient-derived neuron. The at least one processor may analyze the EP signals, to obtain values of a plurality of EP features, characterizing action potentials in the at least one neuron, and compute at least one entropy value, representing entropy of at least one respective EP feature.

[011] According to some embodiments, the at least one processor may classify the patient-derived neuron (and hence - the patient) according to predefined classes of a psychiatric disorder based on (i) the plurality of EP feature values, and/or (ii) the at least one entropy value. Additionally, or alternatively, the at least one processor may determine treatment of the patient based on said classification, as elaborated herein.

[012] According to some embodiments, the at least one processor may classify the patient by inferring a pretrained machine-learning (ML) based classifier on (i) the plurality of EP feature values, and/or (ii) the at least one entropy value, to classify the patient according to the predefined classes of the psychiatric disorder.

[013] Additionally, or alternatively, the at least one processor may select one or more groups of EP features of the plurality of EP features; and for each selected group, computing a Mutual Information (MI) value, representing probabilistic dependence between members of the group. The at least one processor may subsequently infer the classifier also on the MI values of the one or more selected groups of EP features, to classify the patient according to the predefined classes of the psychiatric disorder.

[014] Additionally, or alternatively, at least one processor may employ rule-based logic, to categorize the patient-derived neuron (and hence - the patient) according to the classes of a psychiatric disorder based on (i) the plurality of EP feature values, (ii) the at least one entropy value, and/or the MI values of the one or more selected groups of EP features.

[015] According to some embodiments, the psychiatric disorder may be a Bipolar Disorder (BD). In such embodiments, the predefined classes may include for example (a) non-BD patients, (b) BD patients, responsive to Lithium treatment, (c) BD patients, not responsive to Lithium treatment, (d) BD patients, responsive to anticonvulsant valproic acid (VPA) treatment, and (e) BD patients, not responsive to VPA treatment. Additional psychiatric disorders and predefined classes are also possible, according to the specific application of the invention.

[016] According to some embodiments, the at least one processor may be configured to train the ML model.

[017] For example, the at least one processor may receive a training dataset that may include information derived from each patient of a cohort of patients, that may include: (i) EP feature values, and (ii) entropy values of neuron cells. The at least one processor may also receive annotation data, labeling each patient (or patient derived neuron ) of the cohort according to the predefined psychiatric disorder classes. The at least one processor may subsequently use the annotation data as supervisory information, to train the classifier based on (i) the EP feature values and/or (ii) entropy values of the training dataset, to classify patients of the cohort according to the predefined classes of psychiatric disorder.

[018] Additionally, or alternatively, the training dataset may include information derived from each patient of a cohort of patients, that includes (i) EP feature values, (ii) entropy values of neuron cells, and (iii) MI values of one or more selected groups of EP features. The at least one processor may also receive annotation data, labeling each patient of the cohort according to the predefined psychiatric disorder classes, and may use the annotation data as supervisory information, to train the classifier based on (i) the EP feature values, (ii) the entropy values of the training dataset, and/or (iii) MI values of the training dataset, to classify patients of the cohort according to the predefined classes of psychiatric disorder.

[019] According to some embodiments, the EP features may include, for example values of: Na current, action potential spike height, action potential spike threshold voltage, slow Potassium (K) current, action potential spike width, action potential spike rise time, cell capacitance, cell excitability, fast K current under a first clamp voltage, fast K current under a second clamp voltage, Afterhyperpolarization (AHP) voltage at a first timing, and AHP voltage at a second timing.

[020] Additionally, or alternatively, the at least one entropy values may include, for example entropy of Na current, entropy of fast K current under the first clamp voltage, entropy of slow K current, entropy of fast K current under the second clamp voltage, entropy of cell excitability, entropy of cell capacitance, and entropy of action potential spike height. [021] Additionally, or alternatively, the groups of EP features (e.g., for calculating MI information) may include, for example (i) Na current and fast K current under the first clamp voltage, (ii) cell capacitance and fast K current under the second clamp voltage, (iii) cell excitability and fast K current under the second clamp voltage, (iv) cell capacitance and slow K current, (v) cell excitability and slow K current, (vi) Na current, slow K current, and cell excitability, (vii) Na current, cell excitability, and fast K current under the first clamp voltage, (viii) cell capacitance and fast K current under the first clamp voltage, (ix) cell excitability and fast K current under the first clamp voltage, (x) Na current and fast K current under the second clamp voltage, (xi) Na current and cell excitability, (xii) cell capacitance and cell excitability, (xiii) Na current, cell excitability, and fast K current under the second clamp voltage, (i) cell capacitance and Na current, (xiv) Na current, action potential spike height, and fast K current under the first clamp voltage, (xv) Na current, action potential spike height, and fast K current under the second clamp voltage, and (xvi) Na current, slow K current, and action potential spike height.

[022] Embodiments of the invention may include a system for determining treatment of a patient. Embodiments of the system may include whole-cell patch clamps, configured to (a) apply an excitation signal to at least one neuron, and (b) measure electrophysiological (EP) signals in response to the excitation signal; a non-transitory memory device, wherein modules of instruction code may be stored; and at least one processor associated with the memory device, and configured to execute the modules of instruction code.

[023] Upon execution of said modules of instruction code, the at least one processor may be configured to: obtain EP signals, in at least one patient-derived neuron; analyze the EP signals, to obtain values of a plurality of EP features, characterizing action potentials in the at least one neuron; compute at least one entropy value, representing entropy of at least one respective EP feature; infer an ML based classifier on (i) the plurality of EP feature values, and (ii) the at least one entropy value, to classify the patient according to predefined classes of a psychiatric disorder, and determine treatment of the patient based on said classification.

[024] Embodiments of the invention may include a method of determining treatment of a patient by at least one processor. According to some embodiments, the at least one processor may use, or may communicate (e.g., via wired or wireless communication) with whole-cell patch clamps, to obtain EP signals, in at least one patient-derived neuron.

[025] The at least one processor may analyze the EP signals, to obtain values of a plurality of EP features, characterizing action potentials in the at least one neuron, and select one or more groups of EP features of the plurality of EP features, as elaborated herein. For one or more (e.g., each) selected group, the at least one processor may compute a Mutual Information (MI) value, representing probabilistic dependence between members of the group, and classify, or categorize the neuron (and hence - the patient) according to predefined classes of a psychiatric disorder. This classification may be based on (i) the plurality of EP feature values, and/or (ii) the MI values of the one or more selected groups of EP features. The at least one processor may subsequently determine treatment of the patient based on said classification. [026] For example, the at least one processor may classify the patient by inferring a pretrained machine-learning (ML) based classifier on (i) the plurality of EP feature values, and/or (ii) the MI values of the one or more selected groups of EP features, to classify the patient according to the predefined classes of the psychiatric disorder.

[027] Additionally, or alternatively, the at least one processor may compute at least one entropy value, representing entropy of at least one respective EP feature. The at least one processor may subsequently inferring the classifier also on the at least one entropy value, to classify the patient according to the predefined classes of the psychiatric disorder.

BRIEF DESCRIPTION OF THE DRAWINGS

[028] The subject matter regarded as the invention is particularly pointed out and distinctly claimed in the concluding portion of the specification. The invention, however, both as to organization and method of operation, together with objects, features, and advantages thereof, may best be understood by reference to the following detailed description when read with the accompanying drawings in which:

[029] Fig. 1 is a block diagram, depicting a computing device which may be included in a system for determining treatment of a psychiatric disorder according to some embodiments;

[030] Fig. 2 is a block diagram, depicting a system for determining treatment of a psychiatric disorder, according to some embodiments;

[031] Figs. 3 A and 3B are graphs depicting experimentally obtained values of electrophysiologic features’ entropy, according to some embodiments of the invention;

[032] Figs. 4A-4D are graphs depicting experimentally obtained values of mutual information of electrophysiologic features, according to some embodiments of the invention;

[033] Figs. 5A-5C depict a non-limiting example of implementation of a machine learning (ML) based classifier, and corresponding characteristics of classification by that classifier, according to some embodiments of the invention;

[034] Figs. 6A-6C depict another, non-limiting example of implementation of an ML- based classifier, and corresponding characteristics of classification by that classifier, according to some embodiments of the invention; [035] Fig. 7 A is a flow diagram, depicting a method of determining treatment of a psychiatric disorder by at least one processor, according to some embodiments; and

[036] Fig. 7B is a flow diagram, depicting another method of determining treatment of a psychiatric disorder by at least one processor, according to some embodiments.

[037] It will be appreciated that for simplicity and clarity of illustration, elements shown in the figures have not necessarily been drawn to scale. For example, the dimensions of some of the elements may be exaggerated relative to other elements for clarity. Further, where considered appropriate, reference numerals may be repeated among the figures to indicate corresponding or analogous elements.

DETAILED DESCRIPTION OF THE PRESENT INVENTION

[038] One skilled in the art will realize the invention may be embodied in other specific forms without departing from the spirit or essential characteristics thereof. The foregoing embodiments are therefore to be considered in all respects illustrative rather than limiting of the invention described herein. Scope of the invention is thus indicated by the appended claims, rather than by the foregoing description, and all changes that come within the meaning and range of equivalency of the claims are therefore intended to be embraced therein.

[039] In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be understood by those skilled in the art that the present invention may be practiced without these specific details. In other instances, well-known methods, procedures, and components have not been described in detail so as not to obscure the present invention. Some features or elements described with respect to one embodiment may be combined with features or elements described with respect to other embodiments. For the sake of clarity, discussion of same or similar features or elements may not be repeated.

[040] Although embodiments of the invention are not limited in this regard, discussions utilizing terms such as, for example, “processing,” “computing,” “calculating,” “determining,” “establishing”, “analyzing”, “checking”, or the like, may refer to operation(s) and/or process(es) of a computer, a computing platform, a computing system, or other electronic computing device, that manipulates and/or transforms data represented as physical (e.g., electronic) quantities within the computer’s registers and/or memories into other data similarly represented as physical quantities within the computer’s registers and/or memories or other information non-transitory storage medium that may store instructions to perform operations and/or processes.

[041] Although embodiments of the invention are not limited in this regard, the terms “plurality” and “a plurality” as used herein may include, for example, “multiple” or “two or more”. The terms “plurality” or “a plurality” may be used throughout the specification to describe two or more components, devices, elements, units, parameters, or the like. The term “set” when used herein may include one or more items.

[042] Unless explicitly stated, the method embodiments described herein are not constrained to aa particular order or sequence. Additionally, some of the described method embodiments or elements thereof can occur or be performed simultaneously, at the same point in time, or concurrently.

[043] Bipolar disorder (BD) is a psychiatric disorder characterized by episodes of abnormal mood, typically mania and depression. During manic states, people feel euphoric or irritable, grandiose, with increased energy and decreased need for sleep. In depression, they are sad or feel empty, with associated changes in energy, cognitive functions, sleep, and appetite.

[044] Early diagnosis and proper long-term treatment are critical since the illness can lead to suicide and typically affects education, occupational status, family connections, social interactions, and quality of life. Different pharmacological and psychosocial therapies are being used to treat this disorder. However, these have been limited in their influence on the patientsS. The popular mood stabilizer lithium (Li) has shown a therapeutic effect in about one-third of BD patients with prophylactic treatment (these patients will be abbreviated as Li Responsive (LR) patients throughout the text). The remaining patients either respond partially or do not respond at all. We will denote these patients as Non- responsive (NR) patients throughout this study. Several studies have examined Li mechanisms of action, but these mechanisms generally remain elusive despite the efforts made to understand them.

[045] Anticonvulsant valproic acid (VPA) has also been found beneficial in some BD patients and is a commonly used alternative to Li. In a study, VPA was found to enhance the region of growth cones in cultured sensory neurons.

[046] In this study, we use techniques of information theory to analyze the EP characteristics of BD DG granule neurons. Adding this analysis gives us new perspectives on how Li and VP A affect DG neurons of BD patients compared to the Control group. Furthermore, using these information theory techniques enhances the prediction of drug response significantly. Information theory incorporates probabilistic reasoning and representation to comprehend the enriched transmission of information and processing between systems.

[047] The advantage of using information theory techniques is their independence of a specific probabilistic model, thus enabling quantifying a far wider variety of interactions and events than would be feasible by parametric model methods. Information theory can recognize both linear and non-linear relationships between variables. Even when the data contains various types, information theory techniques may yield meaningful estimates of their correlations that cannot be reached otherwise. It possesses several metrics designed to quantify system behavior, and the analysis can be multi-dimensional. Using information theory in neuroscience allows for simple comparative analysis among various entities such as cells, brain areas, activities, models, or topics.

[048] Embodiments of the invention may include explicitly calculating entropy and/or mutual information of recorded electrophysiological (EP) features, As elaborated herein embodiments of the invention may subsequently using the calculated entropy and MI data as predictive information, to improve categorization of subjects according to predefined classes of psychiatric disorders.

[049] Reference is now made to Fig. 1 , which is a block diagram depicting a computing device, which may be included within an embodiment of a system for determining treatment of a psychiatric disorder, according to some embodiments.

[050] Computing device 1 may include a processor or controller 2 that may be, for example, a central processing unit (CPU) processor, a chip or any suitable computing or computational device, an operating system 3, a memory 4, executable code 5, a storage system 6, input devices 7 and output devices 8. Processor 2 (or one or more controllers or processors, possibly across multiple units or devices) may be configured to carry out methods described herein, and/or to execute or act as the various modules, units, etc. More than one computing device 1 may be included in, and one or more computing devices 1 may act as the components of, a system according to embodiments of the invention.

[051] Operating system 3 may be or may include any code segment (e.g., one similar to executable code 5 described herein) designed and/or configured to perform tasks involving coordination, scheduling, arbitration, supervising, controlling or otherwise managing operation of computing device 1, for example, scheduling execution of software programs or tasks or enabling software programs or other modules or units to communicate. Operating system 3 may be a commercial operating system. It will be noted that an operating system 3 may be an optional component, e.g., in some embodiments, a system may include a computing device that does not require or include an operating system 3.

[052] Memory 4 may be or may include, for example, a Random-Access Memory (RAM), a read only memory (ROM), a Dynamic RAM (DRAM), a Synchronous DRAM (SD-RAM), a double data rate (DDR) memory chip, a Flash memory, a volatile memory, a non-volatile memory, a cache memory, a buffer, a short term memory unit, a long term memory unit, or other suitable memory units or storage units. Memory 4 may be or may include a plurality of possibly different memory units. Memory 4 may be a computer or processor non-transitory readable medium, or a computer non-transitory storage medium, e.g., a RAM. In one embodiment, a non-transitory storage medium such as memory 4, a hard disk drive, another storage device, etc. may store instructions or code which when executed by a processor may cause the processor to carry out methods as described herein.

[053] Executable code 5 may be any executable code, e.g., an application, a program, a process, task, or script. Executable code 5 may be executed by processor or controller 2 possibly under control of operating system 3. For example, executable code 5 may be an application that may determine treatment of a psychiatric disorder, as further described herein. Although, for the sake of clarity, a single item of executable code 5 is shown in Fig. 1, a system according to some embodiments of the invention may include a plurality of executable code segments similar to executable code 5 that may be loaded into memory 4 and cause processor 2 to carry out methods described herein.

[054] Storage system 6 may be or may include, for example, a flash memory as known in the art, a memory that is internal to, or embedded in, a micro controller or chip as known in the art, a hard disk drive, a CD-Recordable (CD-R) drive, a Blu-ray disk (BD), a universal serial bus (USB) device or other suitable removable and/or fixed storage unit. Data pertaining to EP recordings in patient-derived dentate gyrus (DG) granule neurons may be stored in storage system 6, and may be loaded from storage system 6 into memory 4 where it may be processed by processor or controller 2. In some embodiments, some of the components shown in Fig. 1 may be omitted. For example, memory 4 may be a non-volatile memory having the storage capacity of storage system 6. Accordingly, although shown as a separate component, storage system 6 may be embedded or included in memory 4.

[055] Input devices 7 may be or may include any suitable input devices, components, or systems, e.g., a detachable keyboard or keypad, a mouse and the like. Output devices 8 may include one or more (possibly detachable) displays or monitors, speakers and/or any other suitable output devices. Any applicable input/output (I/O) devices may be connected to Computing device 1 as shown by blocks 7 and 8. For example, a wired or wireless network interface card (NIC), a universal serial bus (USB) device or external hard drive may be included in input devices 7 and/or output devices 8. It will be recognized that any suitable number of input devices 7 and output device 8 may be operatively connected to Computing device 1 as shown by blocks 7 and 8.

[056] A system according to some embodiments of the invention may include components such as, but not limited to, a plurality of central processing units (CPU) or any other suitable multi-purpose or specific processors or controllers (e.g., similar to element 2), a plurality of input units, a plurality of output units, a plurality of memory units, and a plurality of storage units.

[057] The term neural network (NN) or artificial neural network (ANN), e.g., a neural network implementing a machine learning (ML) or artificial intelligence (Al) function, may be used herein to refer to an information processing paradigm that may include nodes, referred to as neurons, organized into layers, with links between the neurons. The links may transfer signals between neurons and may be associated with weights. A NN may be configured or trained for a specific task, e.g., pattern recognition or classification. Training a NN for the specific task may involve adjusting these weights based on examples. Each neuron of an intermediate or last layer may receive an input signal, e.g., a weighted sum of output signals from other neurons, and may process the input signal using a linear or nonlinear function (e.g., an activation function). The results of the input and intermediate layers may be transferred to other neurons and the results of the output layer may be provided as the output of the NN. Typically, the neurons and links within a NN are represented by mathematical constructs, such as activation functions and matrices of data elements and weights. At least one processor (e.g., processor 2 of Fig. 1) such as one or more CPUs or graphics processing units (GPUs), or a dedicated hardware device may perform the relevant calculations. [058] Reference is now made to Fig. 2, which depicts a system 100 for determining treatment of a psychiatric disorder, according to some embodiments.

[059] According to some embodiments of the invention, system 100 may be implemented as a software module, a hardware module, or any combination thereof. For example, system may be or may include a computing device such as element 1 of Fig. 1 , and may be adapted to execute one or more modules of executable code (e.g., element 5 of Fig. 1) to determine treatment of a psychiatric disorder, as further described herein.

[060] Arrows in Fig. 2 may represent flow of one or more data elements to and/or from system 100 and/or among modules or elements of system 100. Some arrows have been omitted in Fig. 2 for the purpose of clarity.

[061] As shown in Fig. 2, system 100 may be connected, e.g., via wired or wireless connection to whole-cell patch clamps 20, or “clamps 20” for short. Clamps 20 may be configured to (a) apply an excitation signal 20E to at least one neuron 25, and (b) measure EP signals 25EP in response to the excitation signal, as known in the art. Alternatively, system 100 may include patch clamps 20. Such embodiments of the system are denoted in Fig. 2 as system 100’. As used herein, the term “Electrophysiological signal” or “EP signal” may refer to a signal derived from neural EP activity, including for example voltage of an action potential spike, current derived from the AP spike, and the like.

[062] According to some embodiments, system 100 may use the whole-cell patch clamps 20 to obtain EP signals, in at least one patient-derived neuron 25. The terms “neuron”, “cell” and “neural cell” may be used herein interchangeably in this context.

[063] The terms “patient”, “individual” and “subject” may be used herein interchangeably in reference to a person or animal to which neuron 25 pertains. Additionally, classification or categorization of neuron 25 (e.g., according to classes of a medical condition or psychiatric disorder) may also be regarded as applicable to the corresponding subject.

[064] Embodiments of the invention may include extracting, or obtaining patient- derived neuron cells 25.

[065] For example, during experimental application of system 100, BD participants were selected as part of an ongoing genetic research while healthy volunteers or married-in relatives of certain probands served as control subjects. B-lymphocytes obtained from all of these individuals were immortalized with Epstein-Barr virus (EBV) and reprogrammed to form induced pluripotent stem cells (iPSCs), as known in the art. These iPSC lines went through validation of quality control management requirements before further differentiation and characterization into neuronal cells.

[066] In some embodiments, the iPSC colonies that met predefined quality criteria were differentiated into primed neural progenitor cells (NPCs) and then into hippocampal DG granule-cell-like neurons. More than 48% of differentiated neurons expressed the Proxl gene, a proxy for dentate granule cells. These differentiated neurons were infected with the Proxl: eGFP lentiviral vector on day 12 of the post differentiation period, thereby obtaining the neural cells of interest, denoted herein as neurons 25.

[067] Embodiments of the invention may include performing EP recordings of Action Potential (AP) signals 25EP on neurons 25 using whole-cell patch-clamps 20 within a predefined period (e.g., 10-45 days) of post differentiation.

[068] Neuronal cultures may be treated chronically (e.g., repeatedly, over time) with 1 mM LiCl or 1 mM VPA starting 14 days post differentiation until the patch-clamp experiments.

[069] EP recordings of EP signals 25EP of these Li or VPA treated cells 25 may be performed between 22 and 30 days post differentiation periods.

[070] According to some embodiments, on the 12th day of differentiation, neurons 25 may be infected with the Proxl ::eGFP lentiviral vector. Neurons 25 may be moved to a recording chamber using a recording medium containing (in mM): 10 HEPES, 4 KC1, 2 CaC12, 1 MgC12, 139 NaCl, and 10 D-glucose (310 mOsm, pH 7.4). Whole-cell patch- clamps 20 may be used to record AP from DG-like neurons 25 expressing Proxl ::eGFP, typically during 22-30 days of differentiation but ranging from 10-45 days. Internal solution containing (in mM): 130 K-gluconate, 6 KC1, 4 NaCl, 10 Na-HEPES, 0.2 K-EGTA, 0.3 GTP, 2 Mg- ATP, 0.2 cAMP, 10 D-glucose, 0.15 % biocytin, and 0.06 % rhodamine may be used to fill patch clamp 20 electrodes. Internal solution pH and osmolarity may be adjusted to physiological values (pH 7.3, 290-300 mOsm) (pipette tip resistance was usually 10-15 MΩ). [071] According to some embodiments, system 100 may include a preprocessing module 105, configured to analyze EP signals 25EP, to produce digital EP data 105D, representing EP activity in neural cells 25. [072] For example, preprocessing module 105 may include an amplification module, such as a Multiclamp700B amplifier (Sunnyvale, California, USA), adapted to enhance received EP signals 25EP.

[073] Additionally, or alternatively, preprocessing module 105 may include a sampling module, configured to collect EP signal data 25EP at a predefined (e.g., 20 kHz) sampling rate.

[074] Additionally, or alternatively, preprocessing module 105 may include a digital to analog (D2A) converter, and a recording system, such as the Axon Instruments' Clampex 10.2 program (Union City, California, USA), configured to record the EP activity as digital EP data 105D (e.g., in storage 6 of Fig. 1). Additionally, or alternatively EP data 105D may be processed by preprocessing module 105 using the currently available Clampfit-10 and MATLAB software kits (release 2014b; The MathWorks, Natick, MA, USA).

[075] During EP recordings, clamps 20 may operate in a current-clamp mode, as known in the art. In this mode, neuron cells 25 may be injected with an excitation signal 20E that includes a holding current, as required to hold cell 25 membrane voltage at -60 millivolt (mV). Excitation signal 20E may then include current injection steps of 3 pico-Amperes (pA), given to the patched cells 25 with a duration of 400 milliseconds (ms), beginning from ~12 pA below the holding current. Thirty-five steps of current injections may be performed by excitation signal 20E. The total number of action potentials in EP signal 25EP (also referred to herein as spikes) in the 35 steps may be counted. This number is referred to herein as the total evoked action potentials, or cell excitability.

[076] Additionally, or alternatively, clamps 20 may operate in a voltage clamp mode, as known in the art. In this mode, clamps 20 may be used to obtain Sodium (Na) and Potassium (K) currents values. Excitation signal 20E may hold neurons 25 at -60 mV, and periodic voltage steps of 400 ms may be given between -90 and 80 mV. These currents may be normalized by the cell capacitance to compensate for cell size. Additionally, or alternatively, in the Li-treated neurons 25 no normalization may be performed.

[077] As shown in Fig. 2, system 100 may include a feature extraction module 110 (or “module 110”, for short), configured to analyzing the EP signals 25EP of EP activity (now EP data 105D), to obtain values of a plurality of EP features 110F, characterizing action potentials in the at least one neuron 25. [078] For example, module 110 may measure amplitude of incoming currents (e.g., signals 25EP) in the voltage-clamp mode in different testing potentials. It has been observed that a strong capacitive transient occurs immediately after a depolarization phase, interfering with the measurements. The membrane in the voltage clamp can be approximated as a resistor and a capacitor in a parallel electrical setup. The current in the capacitor operates roughly as the derivative of the change in potential (I=C*dVc/dt) during a voltage step and is much stronger than the currents in the resistor during fast transients. As a result, we can assume a capacitive impedance that scales nearly linear with the voltage step for fast transitions in the membrane potential (dVc). Module 110 may use this as a reference capacitive current, by measuring the current (e.g., signal 25EP) with a -10 mV voltage step from -60 mV to -70 mV, where almost no voltage-gated channels are open.

[079] Module 110 may then generalize this uniformly with the voltage step (for example, multiplying the current provided by the -10 mV step by -2 for a 20 mV voltage step) and may subtract this scaled current from the measured amplitude of signal 25EP, to remove capacitive transient current.

[080] Feature extraction module 110 may calculate at least one EP feature 110F that is an Na current maximal amplitude feature 110F (also referred to herein as Sodium current 110F, and Na current 110F), at a specific test excitation signal 20E potential, such as -20 mV. These specific excitation signal 20E potentials may be chosen according to where significant differences between groups or drug treatment responsiveness were found.

[081] In other words, when examining classification of a psychiatric disorder that is a Bipolar Disorder (BD), excitation signal 20E potential may be selected to distinguish between neurons 25 pertaining to predefined BD classes. Such classes may include, for example (a) non-BD patient, (b) BD patient, responsive to Lithium treatment, (c) BD patient, not responsive to Lithium treatment, (d) BD patient, responsive to anticonvulsant valproic acid (VPA) treatment, and/or (e) BD patient, not responsive to VPA treatment.

[082] Additionally, or alternatively, feature extraction module 110 may calculate at least one EP feature 110F that is a cell excitability feature 110F, representing a number of AP spikes, in response to application of excitation signal 20E on neuron 25.

[083] Additionally, or alternatively, feature extraction module 110 may calculate at least one EP feature 11 OF pertaining to Potassium (K) current, including for example: (i) a maximal amplitude of a fast Potassium (K) current, under a first clamp voltage (e.g., 20 mV), also referred to herein as fast K current feature 110F (first voltage); (ii) a maximal amplitude of a fast K current under a second clamp voltage (e.g., 0 mV), also referred to herein as fast K current feature 110F (second voltage); and (iii) a maximal amplitude of slow K current, also referred to herein as slow K current feature 110F.

[084] Feature extraction module 110 may extract K currents from EP signals 25EP (now EP data 105D). Module 110 may divide K currents into fast potassium currents, and slow potassium currents, to produce corresponding EP features 110F:

[085] One such EP feature 110F may include a fast potassium current EP feature 110F, representing a maximal current immediately after a depolarization step, generally within a few milliseconds.

[086] Another EP feature 110F may include a slow potassium current EP feature 110F, which may be measured as a current amplitude after 400 ms of depolarization.

[087] Such Potassium current AP information features 110F may be measured, and/or calculated in response to excitation signals 20E having specific test potentials, such as 0 mV and 20 mV. These specific potentials may be chosen according to where significant differences between groups or drug treatment were found, e.g., to discern between the predefined BD classes.

[088] Additionally, or alternatively, module 110 may calculate at least one EP feature 110F of input conductance. For example, module 110 may determine input conductance feature 110F by calculating the current in signal 25EP with the cell held in voltage-clamp mode at -70 mV and then at -50 mV. The measured input conductance may be calculated as the difference in currents, divided by the change in membrane potential (20 mV).

[089] Additionally, or alternatively, module 110 may calculate at least one EP feature 110F that is a cell capacitance feature 110F. In some embodiments, cell capacitance feature 110F may represent capacitance of cell 25, and may be measured during the recordings by using software such as the currently available “Clampex” software.

[090] Additionally, or alternatively, module 110 may calculate at least one EP feature 110F that is a spike shape feature 110F, that may characterize a shape of an AP spike. For example, module 110 may evaluate a first (e.g., a chronologically first) AP spike in EP signal 25EP for calculating one or more spike shape features 110F, with the lowest injected current 20E needed for eliciting a spike. [091] Additionally, or alternatively, module 110 may calculate at least one EP feature 110F that is a spike threshold feature 110F, e.g., a membrane potential that significantly increases the slope of the depolarizing membrane potential, leading to a spike. The spike threshold feature 110F may be calculated, for example, as the first maximum in the second derivative of the voltage vs. time function of EP signal 25EP.

[092] Additionally, or alternatively, module 110 may calculate at least one EP feature 110F that is an Afterhyperpolarization (AHP) feature, characterizes an AHP potential. As known in the art, AHP may represent a hyperpolarizing phase of a neuron's action potential where the cell's membrane potential falls below the normal resting potential. This is also commonly referred to as an action potential's undershoot phase. Module 110 may calculate a first EP feature 110F that is a fast AHP feature 110F, representing voltage at a first timing, and a second EP feature 110F representing AHP voltage at a second timing.

[093] For example, a 5-ms fast AHP EP feature 110F, and a 1-ms fast AHP EP feature 110F may be calculated as the difference between the threshold for spiking and the value of the membrane potential, 5 ms and 1 ms respectively, after the potential returned to cross the threshold value at the end of the action potential.

[094] Additionally, or alternatively, module 110 may calculate at least one EP feature 110F that is an action potential maximal amplitude feature 110F, also referred to herein as spike height or spike amplitude feature 110F. For example, spike height feature 110F value may be calculated as the amplitude difference between the maximum membrane potential during a spike and the spike threshold voltage.

[095] Additionally, or alternatively, module 110 may calculate at least one EP feature 110F that is a spike width feature 110F. For example, spike width feature 110F value may be calculated as the time it takes the membrane potential to reach half the spike height in the rising part of the spike to the descending part of the spike (full width at half-maximum).

[096] Additionally, or alternatively, module 110 may calculate at least one EP feature 110F that is a rise time feature 110F. Rise time 110F may be calculated as the time taken by membrane potential to reach the peak value of membrane potential during occurrence of an AP spike, from the spike threshold value.

[097] As known in the art, entropy can be intuitively understood as a measure of uncertainty, or a quantity of information provided by a given variable. The Entropy H(X) of a discrete random variable X, having possible values xEX may be calculated according to Eq. 1, below:

Eq. 1 where ρ() is the probability distribution of X. [098] As shown in Fig. 2, system 100 may include an entropy calculation module 130, configured to compute at least one entropy value 130E, representing entropy of at least one respective EP feature 110F.

[099] According to some embodiments, entropy calculation module 130 may categorize EP data 105D pertaining to the cohort of patients into characteristic groups or classifications. These classifications may include for example control (e.g., non-RB) subjects, LR subjects, LR subjects with Li treatment, LR subjects with VP A treatment, NR subjects, NR subjects with Li treatment, and NR subjects with VP A treatment.

[0100] Entropy calculation module 130 may uniformly bin values of one or more (e.g., each) EP feature 110F (e.g., relating to cell capacitance, cell excitability, Na current, K current, etc.) into a predetermined number (e.g., ten) bins, and calculate probability distribution over these bins (e.g., p(x) of Eq. 1), assuming ergodicity.

[0101] Entropy calculation module 130 may calculate entropy 130E of the one or more (e.g., all) EP features 110F based on Eq. 1.

[0102] Reference is now made to Figs. 3A and 3B which are graphs depicting experimentally obtained values of electrophysiologic features’ entropy, according to some embodiments of the invention. Fig. 3 A shows a comparison between entropy 130E of fast K current (where excitation signal 20E was set to 20 mV), between LR subjects and NR subjects. Fig. 3B shows a comparison between entropy 130E of slow K current (where excitation signal 20E was also set to 20 mV), between LR subjects and NR subjects. It may be appreciated by these comparisons that entropy 130E of fast K current feature 110F and slow K current feature 110F may serve to distinguish between LR subjects and NR subjects. [0103] As elaborated herein, system 100 may thus classify, or categorize new (e.g., beyond the training cohort) EP signals 25EP, as pertaining to LR subjects or NR subjects, based on the entropy data elements 130E (e.g., entropy 130E of fast K current feature 110F and/or entropy 130E of slow K current feature 110F), as depicted in the examples of Figs. 3A and 3B. [0104] According to some embodiments, entropy values 130E may be calculated in relation to features 110F of a training dataset of feature values 110F, to determine features 110F having the most discriminative entropy values 130E between categories of BD subjects. Selection module may subsequently select, or receive (e.g., via input 7) a selection of these most discriminative features 110F, as a subset 120E for calculating entropy values 130E.

[0105] For example, experimental results have shown subsets 120E of EP features 110F that include features 110F, which were most discriminative for classifying or categorizing BD subjects. Such entropy values include for example, entropy 130E of cell capacitance feature 110F, entropy 130E of fast K current feature 110F under the first clamp voltage, entropy 130E of fast K current feature 110F under the second clamp voltage, entropy 130E of slow K current feature 110F, entropy 130E of Na current feature 110F, and entropy 130E of action potential spike height feature 110F.

[0106] In a subsequent applicative, or inference stage, system 100 may calculate entropy values 130E of the subset 120E of features 110F, to determine classification 150C of new, incoming (e.g., beyond the training dataset) EP data elements 105D, derived from EP signals 25EP of new cells 25.

[0107] As known in the art, Mutual Information (MI) can be intuitively understood as a reduction in the uncertainty in one variable by knowing the value of another variable. The mutual information I(X,Y) between discrete variables X and Y may calculated according to Eq. 2, below:

Eq. 2 where ρ(x,y) is the joint probability distribution of X and Y, H(X) and H(Y) are the entropies of variables X and Y respectively, and H(X,Y) is the joint entropy of X and Y (e.g., obtained by replacing ρ(x) with ρ(x, y) in Eq. 1).

[0108] As shown in Fig. 2, system 100 may include an MI calculation module 140, configured to compute at least one MI value 140M, representing probabilistic dependence between two or more preselected EP features 110F.

[0109] According to some embodiments, MI calculation module 140 may categorize EP data 105D pertaining to the cohort of patients into characteristic groups or classifications. These classifications may include for example the control (e.g., non-RB) subjects, LR subjects, LR subjects with Li treatment, LR subjects with VP A treatment, NR subjects, NR subjects with Li treatment, and NR subjects with VP A treatment.

[0110] Additionally, system 100 may include a feature selection module 120, configured to receive (e.g., via input 7 of Fig. 1) or select a group of features 120G, as elaborated herein. [0111] According to some embodiments, MI calculation module 140 may receive preselected group 120G of features 110F (e.g., X and Y of Eq. 2) from feature selection module 120, and uniformly bin one or more (e.g., each) EP feature 110F (e.g., relating to cell capacitance, cell excitability, Na current, K current, etc.) of the selected group 120G into a predetermined number (e.g., ten) bins. MI calculation module 140 may calculate probability distributions over the bins of the preselected features 110F (e.g., p(x), p(y) of Eq. 2), and may also calculate joint probabilities of the members of preselected group of features 110F (e.g., p(x,y) of Eq. 2), to obtain mutual information 140M (e.g., I(X,Y) of Eq. 2) members of preselected group of features 110F (e.g., variables X and Y) as elaborated in Eq. 2.

[0112] Reference is now made to Figs. 4A-4D which are graphs depicting experimentally obtained values of mutual information 140M of electrophysiologic features 110F, according to some embodiments of the invention.

[0113] Fig. 4 A shows a comparison between MI values 140M of LR and NR subjects, where the MI values represent probabilistic dependence among a group 120G of EP features 110F of (i) cell capacitance feature 110F and (ii) fast K current feature 110F (where excitation signal 20E was set to 20 mV);

[0114] Fig. 4B shows a comparison between MI values 140M of LR and NR subjects, where the MI values represent probabilistic dependence among a group 120G of EP features 110F of (i) cell capacitance feature 110F and (ii) slow K current feature 110F (where excitation signal 20E was set to 20 mV);

[0115] Fig. 4C shows a comparison between MI values 140M of LR and NR subjects, where the MI values represent probabilistic dependence among a group 120G of EP features 110F of (i) cell excitability feature 110F and (ii) fast K current feature 110F (where excitation signal 20E was set to 20 mV); and

[0116] Fig. 4D shows a comparison between MI values 140M of LR and NR subjects, where the MI values represent probabilistic dependence among a group 120G of EP features 110F of (i) cell excitability feature 110F, (ii) slow K current feature 110F, and (iii) Na current feature 110F (where excitation signal 20E was set to 20 mV).

[0117] As clearly shown in the non-limiting examples of Figs. 4A-4D, it may be possible to differentiate between LR and NR subjects based on the MI values 140M of member features 110F of group 120G. Therefore, and as elaborated herein, system 100 may classify, or categorize new (e.g., beyond the training cohort) EP signals 25EP, as pertaining to LR subjects or NR subjects, based on MI values 140M among a group 120G of preselected EP features 110F (e.g., as depicted in the examples of Figs. 4A-4D).

[0118] According to some embodiments, MI values 140M may be calculated in relation to features 110F of a training dataset of feature values 110F, to determine EP features 110F having the most discriminative MI 140M values between categories of BD subjects. Selection module may subsequently select, or receive (e.g., via input 7) a selection of these most discriminative features 110F, to form group 120G.

[0119] For example, experimental results have shown groups 120G of EP features 110F that include pairs of features 110F, which were most discriminative for classifying or categorizing BD subjects. These groups 120G included for example pairs such as: (i) cell capacitance and Na current, (ii) cell capacitance and cell excitability, (iii) cell capacitance and fast K current under the first clamp voltage, (iv) Na current and fast K current under the first clamp voltage, (v) cell excitability and fast K current under the first clamp voltage, (vi) cell capacitance and fast K current under the second clamp voltage, (vii) Na current and fast K current under the second clamp voltage, (viii) cell excitability and fast K current under the second clamp voltage, (ix) Na current and cell excitability. Therefore, groups 120F of features 110F may include pairs of features 110F such as example pairs (i)-(ix) above.

[0120] In another example, experimental results have shown groups 120G of EP features 110F that include groups of three features 110F, which were most discriminative for classifying or categorizing BD subjects. These groups 120G of three features included for example: (x) Na current, action potential spike height, and fast K current under the first clamp voltage, (xi) Na current, action potential spike height, and fast K current under the second clamp voltage, and (xii) Na current, slow K current, and action potential spike height. [0121] In a subsequent applicative, or inference stage, system 100 may calculate MI values 140M of the group 120G of features 110F, to determine classification 150C of new, incoming (e.g., beyond the training dataset) EP data elements 105D, derived from EP signals 25EP of new cells 25.

[0122] According to some embodiments, entropy 130E and MI 140M between the EP features 110F may be calculated through the neuronal maturation period of neurons 25 to follow a developmental trajectory and understand the strength of interaction between the features 110F of the different groups and the changes brought by treatment.

[0123] As shown in Fig. 2, system 100 may include a rule-base module 150’. Rule-base module 150’ may be configured to classify or categorize subjects or patients to according to categories or classes 150C’ of a psychiatric disorder (e.g., non-BD subjects, LR subjects, LR subjects with Li treatment, LR subjects with VPA treatment, NR subjects, NR subjects with Li treatment, and NR subjects with VPA treatment), based on EP features 110F derived from recorded EP data 150D.

[0124] For example, rule-base module 150’ may be implemented as a decision function, that may receive as input a combination of values of one or more EP features 110F and/or entropy values 130E, and produce a classification value 150C’ representing pertinence of the subject of EP data 150D to a class of psychiatric disorder (e.g., LR subjects, NR subjects) as depicted in the examples of Figs. 3A and 3B.

[0125] In another example, rule-base module 150’ may be implemented as a decision function, that may receive as input a combination of values of one or more EP features 110F, entropy values 130E, and/or MI values 140M, and produce a classification value 150C’ representing pertinence of the subject of EP data 150D to a class of psychiatric disorder (e.g., LR subjects, NR subjects) as depicted in the examples of Figs. 4A-4D.

[0126] Additionally, or alternatively, system 100 may include a machine learning (ML) based classifier model 150. Classifier model 150 may be, or may include for example a Support Vector Machine (SVM) model and/or a random forest classifier model, pretrained to distinguish between different categories or classes 150C of a psychiatric disorder (e.g., non-BD subjects, LR subjects, LR subjects with Li treatment, LR subjects with VPA treatment, NR subjects, NR subjects with Li treatment, and NR subjects with VPA treatment), based on features 110F derived from recorded EP data 150D.

[0127] In other words, system 100 may infer pretrained ML based classifier 150 on the plurality of EP feature 110F values to predict classification 150C to classify the patient according to predefined classes of a psychiatric disorder. [0128] Reference is now made to Figs. 5A-5C which depict a non-limiting example of implementation of an ML based classifier 150, and corresponding characteristics of classification 150C by classifier 150, according to some embodiments of the invention. As shown In Fig. 5 A, classifier 150 may predict, or provide classification 150C of patient based on EP feature values 110F.

[0129] As shown in Figs. 5B and 5C, the area Under the Curve (AUC) in the Receiver Operating Characteristics (ROC) were used to assess performance of classifier 150. The mean accuracy, the mean AUC and the standard deviation of the AUC scores for each of the test sets were calculated. In Fig. 5B, the performance of classifier 150 was evaluated in relation to classification 150C between LR and NR subjects. In Fig. 5C, the performance of classifier 150 was evaluated in relation to classification 150C between BD and non-BD

(control) subjects.

[0130] Reference is also made to Figs. 6A-6C which depict a second, non-limiting example of implementation ML based classifier 150, and corresponding characteristics of classification 150C, according to some embodiments of the invention.

[0131] As shown In Fig. 6A, classifier 150 may predict, or provide classification 150C of patient based on (i) EP feature values 110F, and (ii) the at least one entropy value 130E of EP features 110F, to provide classification of the patient or subject according to the predefined classes of a psychiatric disorder.

[0132] Additionally, or alternatively, feature selection module 120 may select, or may receive a selection of one or more groups 120G of EP features 110F of the plurality of EP features 110F, and collaborate with MI calculation module 140 to compute, for one or more (e.g., each) selected group 120G an MI value 140M representing probabilistic dependence among members of that group. As shown In Fig. 6A, system 100 may predict, or provide classification 150C of the patient by further inferring classifier model 150 on the MI values 140M of the one or more selected groups of EP features 110F, to classify the patient according to the predefined classes of the psychiatric disorder.

[0133] In Fig. 6B, the performance of classifier 150 was evaluated in relation to classification 150C between LR and NR subjects. In Fig. 6C, the performance of classifier 150 was evaluated in relation to classification 150C between BD and non-BD (control) subjects. [0134] By comparing Fig. 5B to Fig. 6B, it may be appreciated that specific selection of a group 120G of EP features 110F, and further inference of ML model 150 on calculated entropy 130E and/or calculated MI 140M of these features 110F may improve performance of classification 150C of patients in relation to LR Vs. NR classes.

[0135] In a similar manner, by comparing Fig. 5C to Fig. 6C, it may be appreciated that specific selection of a group 120G of EP features 110F, and further inference of ML model 150 on calculated entropy 130E and/or calculated MI 140M of these features 110F may improve performance of classification 150C of patients in relation to non-BD (control) subject Vs. BD subject classes.

[0136] It may therefore be appreciated by a person skilled in the art, that by explicitly calculating, and adding entropy values 130 as input to classifier 150, system 100 may add information pertaining to distribution of each of EP features 110F, that includes more than the EP features’ 110F mean values.

[0137] It may also be appreciated that MI may provide a scope of information that is broader than the EP features’ 110F correlation values. In other words, correlation between EP features 110F may measure a linear dependence between variables (e.g., EP features 110F), whereas MI may provide more general information, such as a measurement of how different a joint distribution between EP features 110F is, compared to the product of the marginal EP feature 110F distributions.

[0138] Therefore, as shown by comparing Figs. 6A-6C to Figs. 5A-5C, by using this added information, system 100 was able to significantly improve the predictive power of classifier 150.

[0139] According to some embodiments, feature selection module 120 may select EP features 110F, entropy data elements 130E and/or groups 120G (for MI 140M calculation) by gradually adding these data elements (110F, 130E, 140M) into a training dataset 50DS, and assessing the contribution of the added data elements (110F, 130E, 140M) to prediction. [0140] For example, if a feature 110F is found to improve accuracy of classification 150C, then it may remain as member of training dataset 50DS. In a complementary manner, if a feature 110F does not improve accuracy of classification 150C, it may be removed from dataset 50DS. This way specific EP features 110F of EP measurements may be selectively added as members of training dataset 50DS, followed by selective addition of entropy data elements 130E of EP as members to training dataset 50DS, and followed by selective addition of MI data elements 140M of combinations of 110F EP features as members to training dataset 50DS.

[0141] As shown in Fig. 2, system 100 may include a recommendation module, adapted to determine, or produce a recommendation 160R for treatment of a patient based on classification 150C.

[0142] Pertaining to the non-limiting example of BD, recommendation 160R may include, for example, a prescription of Li treatment, for patients who have been classified 150C as Lithium LR subjects, or use of VP A for patients who have been classified 150C as Lithium NR subjects.

[0143] According to some embodiments, system 100 may receive (e.g., via input 7 of Fig. 1, during a training stage) a training dataset 50DS, for training classifier 150. Dataset 50DS may include information derived from each patient of a cohort of patients. This information may include, for example EP data 105D, extracted from each of the patients in the cohort.

[0144] Additionally, or alternatively, the received information in training dataset 50DS may include EP feature values 110F extracted from each of the patients in the cohort, as elaborated herein.

[0145] Additionally, or alternatively, the received information in training dataset 50DS may include entropy values 130E of neuron cells 25 pertaining to one or more (e.g., all) patients the cohort of patients.

[0146] Additionally, or alternatively, the received information in training dataset 50DS may include annotation data 50AN, which may associate, or label one or more (e.g., each) patient of the cohort according to the predefined psychiatric disorder classes.

[0147] In such embodiments, system 100 may use the annotation data as supervisory information, to train the classifier based on (i) EP data 105D of training dataset 50DS, and/or EP feature 110F values of training dataset 50DS, and on (ii) entropy values 130E of the training dataset 50DS, to classify patients of the cohort of patient according to the predefined classes of psychiatric disorder.

[0148] Additionally, or alternatively, the received information in training dataset 50DS may include MI values 140M of one or more selected groups 120G of EP features 110F.

[0149] In such embodiments, system 100 may use the annotation data 50AN as supervisory information, to train classifier 150 based on (i) EP data 105D of training dataset 50DS, and/or EP feature values 110F of training dataset 50DS, (ii) on entropy values 130E of the training dataset 50DS, and (iii) on MI values of the training dataset 50DS, to classify patients of the cohort according to the predefined classes of psychiatric disorder.

[0150] Reference is now made to Fig. 7 A, which is a flow diagram depicting an example of a method of determining treatment of a psychiatric disorder, according to some embodiments of the invention.

[0151] As shown in step S1005, the at least one processor (e.g., processor 2 of Fig. 1) may communicate with, or use whole-cell patch clamps 20, to obtain electrophysiological (EP) signals 25EP, in at least one patient-derived neuron 25.

[0152] As shown in step S 1010, the at least one processor 2 may analyze the EP signals 25EP to obtain values of a plurality of EP features 110F, characterizing action potentials in the at least one neuron 25.

[0153] As shown in step S1015, the at least one processor 2 may compute al least one entropy value 130E, representing entropy of at least one respective EP feature 110F.

[0154] As elaborated herein, the at least one processor 2 may classify the neuron (and hence, the patient) according to predefined classes of a psychiatric disorder, based on (i) the plurality of EP feature values, and/or (ii) the at least one entropy value.

[0155] For example, as shown in step S1020, the at least one processor 2 may infer a pretrained ML based classifier 150 on (i) the plurality of electrophysiological feature values, and (ii) the at least one entropy value, to classify the patient according to the predefined classes 150C of a psychiatric disorder.

[0156] Additionally, or alternatively, and as shown in step S1025, the at least one processor 2 may determine treatment of the patient based on said classification.

[0157] Reference is also made to Fig. 7B, which is a flow diagram depicting another example of method of determining treatment of a psychiatric disorder, according to some embodiments of the invention.

[0158] As shown in step S2005, the at least one processor (e.g., processor 2 of Fig. 1) may communicate with, or use whole-cell patch clamps 20, to obtain electrophysiological (EP) signals 25EP, in at least one patient-derived neuron 25.

[0159] As shown in step S2010, the at least one processor 2 may analyze the EP signals 25EP to obtain values of a plurality of EP features 110F, characterizing action potentials in the at least one neuron 25. [0160] As shown in steps S2015 and S2020, the at least one processor 2 may select one or more groups 120G of EP features 110F of the plurality of EP features 110F. As elaborated herein, for one or more (e.g., each) selected group 120G, the at least one processor 2 may compute an MI value, representing probabilistic dependence between members of the group 120G.

[0161] As shown in step S2025, the at least one processor 2 may classify the patient according to predefined classes of a psychiatric disorder based on (i) the plurality of EP feature values, and/or (ii) the MI values of the one or more selected groups of EP features. For example, the at least one processor 2 may infer a pretrained ML based classifier 150 on (i) the plurality of EP feature values, and (ii) the MI values of the one or more selected groups of EP features, to classify the patient according to the predefined classes of the psychiatric disorder.

[0162] Additionally, or alternatively, and as shown in step S2030, the at least one processor 2 may determine treatment of the patient based on said classification.

[0163] Unless explicitly stated, the method embodiments described herein are not constrained to a particular order or sequence. Furthermore, all formulas described herein are intended as examples only and other or different formulas may be used. Additionally, some of the described method embodiments or elements thereof may occur or be performed at the same point in time.

[0164] While certain features of the invention have been illustrated and described herein, many modifications, substitutions, changes, and equivalents may occur to those skilled in the art. It is, therefore, to be understood that the appended claims are intended to cover all such modifications and changes as fall within the true spirit of the invention.

[0165] Various embodiments have been presented. Each of these embodiments may of course include features from other embodiments presented, and embodiments not specifically described may include various features described herein.