Medical device and method for diagnosis and treatment of disease

Abstract

The present disclosure provides a medical device for diagnosing a nervous system disease of a patient, configured to be used in a pre-hospital environment, comprising: a memory; the camera is configured as an image / video data sensor for acquiring a patient and comprises a microphone for acquiring natural language data of the patient; a processor communicatively connected to the memory, the sensor, and the camera, the processor configured to: execute computer instructions to obtain patient data input; performing at least one primary computational algorithm and / or analysis process to evaluate the patient data input and determine a primary diagnosis based on the patient input data; in response to the primary diagnosis indicating that a further diagnosis is necessary, performing at least one auxiliary computational algorithm and / or analysis process to evaluate the patient data input, and determining one or more additional diagnoses based on the patient data input; and determining a final diagnosis and / or treatment regimen for the patient in response to the primary diagnosis and the one or more additional diagnoses.

Description

[0001] This application is a divisional application of Chinese Patent Application No. 201980071177.1, with the application date of August 28, 2019 and the invention title of "A Medical Device and Method for Diagnosing and Treating Diseases", and claims priority to U.S. Provisional Patent Application No. 62 / 723,593 filed on August 28, 2018 and U.S. Provisional Patent Application No. 62 / 816,239 filed on March 11, 2019. Both of these patent applications are incorporated herein by reference as if fully set forth herein. In the event of any conflict between the incorporated materials and the specific ideas of the present disclosure, the latter shall prevail. Similarly, in the event of any conflict between the definition of a word or phrase understood in the technical field and the definition specifically set forth in the present disclosure, the latter shall prevail. Technical Field

[0002] The present disclosure provides a medical device and method for diagnosing and treating diseases. Background Art

[0003] It is necessary for medical professionals to make a correct diagnosis of a patient so that they can prescribe appropriate treatments for the patient's disease and manage it properly. It is difficult to make an accurate diagnosis of some diseases, especially when a patient has an initial consultation. This problem becomes even more difficult when the effective treatment window for a disease is limited, such as in the case of stroke and heart attack. Since diagnosing a disease takes time, emergency treatments for different types of acute strokes (especially ischemic strokes), such as using "thrombolytic" recombinant tissue plasminogen activator (rtPA), are rarely performed. Although stroke is the leading cause of disability and the second most common cause of death globally, in the United States, nearly 800,000 strokes occur each year, with healthcare costs of approximately $37 billion. More than 17 million strokes occur each year, 5.7 million of which are fatal, but medical professionals do not yet have tools that can diagnose and treat quickly and effectively. Neurological emergencies (such as acute stroke) can cause time-dependent brain damage. For the above reasons, there is an urgent need to find a device and method that can diagnose and treat diseases quickly and effectively, but it seems insoluble. Summary of the Invention

[0004] Accordingly, an object of embodiments of the present invention is to overcome the above-mentioned deficiencies and drawbacks related to the current technology.

[0005] Currently, ischemic stroke can be treated with recombinant tissue plasminogen activator (rtPA) even before reaching the hospital. When a patient diagnosed with acute stroke is brought to the hospital, only routine neuroimaging examinations are required to confirm the final diagnosis, and then rtPA can be administered to the patient by infusion. Given the strong correlation between earlier treatment and better prognosis, the disclosed device enables on-site diagnosis without the need for in-hospital physicians to re-evaluate, thus saving an average of 83 minutes for rtPA infusion treatment. The time saved means that the effectiveness of rtPA treatment can be increased by 37%, and the number of patients treated can be doubled. Since the medical device disclosed herein is telescopic, it can be used in approximately 81,000 ambulances in the United States and 178,000 ambulances in Europe, for example.

[0006] According to one embodiment, the disclosed artificial intelligence diagnosis (AID) medical device has natural language and visual recognition / computer vision, is a cloud-based artificial neural network, and diagnoses neurological emergencies by communicating directly with the patient. After diagnosis, the medical device will convey the treatment information to the medical staff and guide the transportation of the patient to an appropriate hospital. Patients without a clear diagnosis may be referred to the on-duty neurologist for evaluation.

[0007] The term "disease" as used herein is a general synonym and is used interchangeably with the terms "disorder" and "condition" (medical conditions), as they all reflect abnormal conditions resulting from impairment of the normal function of the human body, animal, or a part thereof, typically manifested by distinct signs and symptoms, and leading to a reduction in the survival time or quality of life of a person or animal, and may also include dysfunction or abnormal functioning of an organ, part, structure, or system of the body, caused by genetic or developmental errors, infections, poisoning, nutritional deficiencies or imbalances, toxic or adverse environmental factors, illness, discomfort, or pain.

[0008] The invention of the present disclosure relates to a method, a system, and a medical device. The device includes a memory and a processor connected to and communicable with the memory; and the processor is configured to execute instructions to evaluate one or more input patient data related to a first specific disease; the processor uses at least one computational algorithm to compare the one or more input data with a set of numerical values from at least one database; trains at least one computational algorithm to estimate a patient's diagnosis based on the first specific disease; uses at least one computational algorithm to determine a first diagnostic score of the patient for this specific disease; when the first diagnostic score of the first specific disease is higher than a first numerical value, diagnoses the patient as having this specific disease, and displays or provides this diagnosis of the patient as an output. According to another embodiment, the processor is further configured to execute instructions to diagnose the patient as not having the specific disease when the diagnostic score is lower than a second numerical value and the second numerical value is lower than the first numerical value. According to another embodiment, the processor is further configured to execute instructions to display the patient's diagnosis as inconclusive when the diagnostic score is between the first numerical value and the second numerical value. According to yet another embodiment, the medical device further includes the following devices: a mirror capable of measuring a reflective area of at least one square foot; a clock; a refrigerator; a toilet; a chair; a bed; a television; a microwave oven; a floor lamp or a counter lamp; and a ceiling-mounted lamp. According to another embodiment, the medical device further includes straps for fixedly wearing the medical device on the wrist. According to another embodiment, the input data is one of, or a combination of, demographic data, symptoms, medical history elements, examination results, and / or diagnostic test results. According to another embodiment, the input data is input by one of the patient or a third party and is automatically acquired by the medical device. According to another embodiment, the medical device interacts with the patient through voice prompts. According to another embodiment, when there is a positive symptom of a specific disease, the likelihood of diagnosing the disease increases, and when there are different symptoms of a similar disease, the likelihood of diagnosing the disease decreases. According to another embodiment, the first specific disease is one of neurological abnormalities, congestive heart failure, asthma, myocardial infarction, and infection. According to another embodiment, the specific disease is a neurological abnormality, including one of, or a combination of, acute ischemic stroke, transient ischemic attack, seizure, demyelinating disease, multiple sclerosis, brain trauma, and brain tumor. According to another embodiment, the medical device automatically evaluates one or more initial signs of a specific disease of the patient and automatically triggers a more comprehensive evaluation when an initial sign is detected.According to another embodiment, an abnormal body temperature is rated as an initial sign of infection, a change in one or more of gait, speech, and limb movement is rated as an initial symptom and sign of a neurological abnormality, a change in respiratory rate, a pause in speech, or both together are rated as an initial symptom and sign of worsening congestive heart failure, shortness of breath, dyspnea, or both together are rated as an initial symptom and sign of an impending or ongoing asthma attack, and one or more of clutching the chest, a pained facial expression, rapid breathing, flushing, and / or sweating are rated as an initial symptom and sign of a myocardial infarction. According to another embodiment, when the device diagnoses a patient with a first specific disease, the device determines the appropriate medication for the patient and notifies the patient, or determines the appropriate medication for the patient, notifies the patient, and then also directly dispenses the medication. According to another embodiment, the processor is further configured to execute instructions to evaluate the likelihood that the patient has a second disease, the second disease being similar to the first disease, and when the likelihood of diagnosing this similar symptom as another disease is higher than 25%, 50%, 75%, or 90%, the diagnosis of having the first disease is accompanied by an alarm. According to another embodiment, at least one of the computational algorithms includes an artificial neural network, a support vector machine (SVM), Nu-SVM, linear SVM, naive Bayes (NB) algorithm, Gaussian NB, polynomial NB computational algorithm, multi-class SVM, directed acyclic graph SVM (DAGSVM), structured SVM, least squares SVM (LS-SVM), Bayesian SVM, transductive SVM, support vector clustering algorithm (SVC), classification SVM type 1 (C-SVM classification), classification SVM type 2 (nu-SVM classification), regression SVM type 1 (ε-SVM regression), and regression SVM type 2 (nu-SVM regression). According to another embodiment, the processor is further configured to send the diagnosis to a healthcare facility via a wired or wireless network, and the medical device further includes means for communicating the diagnosis via the network. According to another embodiment, the processor is further configured to execute a second computational algorithm to determine a second diagnostic score for the patient when it is determined that the first diagnostic score is below a first value, or below the first value and higher than a second value, or below both the first value and the second value. According to another embodiment, the processor is further configured to execute multiple computational algorithms, each algorithm using data from multiple databases, to determine a diagnostic score for the patient's first specific disease for each computational algorithm, and when the diagnostic scores of most or all of the computational algorithms for the specific disease are higher than the first value, the patient is diagnosed with that specific disease.According to another embodiment, the processor is further configured to input one or more syndrome elements covered by the historical definition of a classical syndrome related to a first particular disease, assign a syndrome element score proportional to the prevalence in a known or documented population of patients with the classical syndrome, determine whether the patient has the syndrome elements, calculate the diagnostic probability that the patient has the classical syndrome by dividing the total score representing the syndrome elements identified in the patient by the total score of all syndrome elements covered by the historical definition of the classical syndrome, and input the diagnostic probability of the classical syndrome as data input related to the first disease. According to another embodiment, the first calculation algorithm and the second calculation algorithm are part of a plurality of calculation algorithms that iteratively improve the diagnosis of the patient in a serial manner. According to another embodiment, the calculation of the first calculation algorithm is based on a common set of data from multiple databases, and wherein the calculation of the second calculation algorithm is based on all data from a single database. According to another embodiment, the second calculation algorithm is selected from a plurality of calculation algorithms, wherein each calculation algorithm in the plurality of calculation algorithms is based on all data from a different database, and wherein the selection of the second calculation algorithm is based on the similarity between the patient's data input and the database data used by the second calculation algorithm.

[0009] According to another embodiment, the presently disclosed invention relates to devices, systems, and methods that include generating a numerical value representing a syndrome related to a medical disease, where the syndrome can be a collection of symptoms, historical elements, examination results, and diagnostic test results related to the medical disease; generating a set of biometric values representing a patient, providing each value representing the syndrome and the set of biometric values to a machine learning system to provide an output value representing the likelihood that the patient has the medical disease, and generating an output derived from the output value for a user.

[0010] Preferred embodiments of the medical device will be described in detail below and will be accompanied by drawings, so that the various objects, features, aspects, and advantages of the present invention will be more apparent. In the drawings, the same numerical values represent the same components. The present invention can solve one or more problems and deficiencies of the current technology discussed above. However, it is contemplated that the present invention may prove useful in solving other problems and deficiencies in many technical fields. Therefore, the claimed invention should not be construed as limited to solving any particular problem or deficiency discussed herein. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings incorporated in and forming a part of this specification illustrate various embodiments of the invention and, together with the general description of the invention given above and the detailed description of the drawings given below, explain the principles of the invention. It is to be understood that the drawings are not necessarily drawn to scale, since the emphasis is placed on illustrating the principles of the invention. The invention will now be described by way of example with reference to the drawings, in which:

[0012] Figure 1 is a schematic diagram of the external symptoms and internal nerve damage of a classic ischemic stroke syndrome, the lateral medullary syndrome. Damage to the lateral part of the medulla oblongata (the darker part on the left side of the cross-section of the medulla oblongata as shown) impairs certain neuroanatomical structures, resulting in some unique symptoms and physical examination abnormalities. The acute injury can be caused by the occlusion of a specific artery, also known as ischemic stroke;

[0013] Figure 2 is a table summarizing various stroke-like episodes, excluding the symptoms of hemorrhagic stroke (intracranial hemorrhage, subarachnoid hemorrhage);

[0014] Figure 3 is a flowchart of the domain of a computational algorithm serially employed in the patient diagnostic evaluation of an embodiment of the presently claimed invention. In this example of an embodiment of the medical device, to identify ischemic stroke patients suitable for emergency treatment, patients who may have a certain neurological emergency are evaluated through a set of computational algorithms that run in a logical sequence to refine the diagnosis. "Stroke-like episodes" cover neurological diseases often misdiagnosed as stroke, such as seizures. AI = Artificial Intelligence. CA = Computational Algorithm. TIA = Transient Ischemic Attack;

[0015] Figures 4A to 4D are the potential steps involved in the weighted / probability calculation of a classic syndrome using scores, which can be used as discrete data inputs for an embodiment of the disclosed AI-based diagnostic medical device of the medical device. Figures 4A to 4D describes the logic of using score evaluation to identify classic syndromes such as the lateral medullary ischemic stroke syndrome;

[0016] Figure 5Shows the importance based on this data input, the data acquisition tasks (left) of the patient interface of the device, the order of which will be adjustable. Then, one or more computational algorithms process the data input (right), where the strongest data may be the key definition. Taking a stroke diagnosis device as an example, the key definitions will include the acuity of onset, the persistence of symptoms, classical stroke syndromes, and stroke-like episodes (the gray box in the upper left of the center of the figure). Key definitions may require examination results as well as symptoms and triggering events to be established; other data may have an impact, although to a lesser extent. Empirically, the weight (W) of the data input is improved to increase the accuracy of the diagnosis, rather than the "gold standard" type of doctor diagnosis. Then, the calculated probability (netj) can determine, for example, the diagnosis of acute ischemic stroke (AIS), based on which treatment decisions can be made. EMS = Emergency Medical Service; EKG = Electrocardiogram; rtPA = recombinant tissue plasminogen activator;

[0017] Figure 6 Shows a specific example of a hierarchical multi-level analysis device process with a first-level computational algorithm group and three second-level computational algorithm groups, where the first-level computational algorithm first evaluates the patient, and if a diagnosis cannot be generated, then one of the second-level computational algorithms is adopted, applied, or followed to evaluate the patient differently to obtain a diagnosis;

[0018] Figure 7 Shows a flowchart that describes an example of the process of the domain of computational algorithms serially used in the diagnostic evaluation of a patient according to an embodiment of the currently claimed invention;

[0019] Figure 8 Shows a schematic diagram of a serial method for calculating a patient's diagnosis using multiple databases and computational algorithms;

[0020] Figure 9 Shows a schematic diagram of a group method for calculating a patient's diagnosis using multiple databases and computational algorithms;

[0021] Figure 10 Shows a flowchart that describes an example of the process of the domain of computational algorithms serially adopted in the diagnostic evaluation of a patient according to an embodiment of the currently claimed invention;

[0022] Figure 11 Shows an example of a neural network architecture;

[0023] Figure 12 Shows a specific schematic example of a neural network with four neural network layers;

[0024] Figure 13 illustrates a schematic example of a computing device, in accordance with an embodiment of the disclosed subject matter;

[0025] Figure 14 illustrates an example block diagram of the medical device, in accordance with an embodiment of the disclosed subject matter, the medical device including a plurality of sensors coupled to a neural network via an interface;

[0026] Figure 15 illustrates a diagnostic medical device embodiment of the disclosed medical device based on artificial intelligence, and potential steps involved in another embodiment of performing fractional weighting / probability calculation of classical syndromes as discrete data inputs;

[0027] Figure 16 and Figure 17 shows a specific example of an African - American middle - aged male from New York City, and a method of selecting a preferred diagnostic computing algorithm based on demographic characteristics of patients similar to those in the patient records of other patients in the database used to train the computing algorithm, utilizing non - geographic demographic similarities ( Figure 16 ) and geographic similarities ( Figure 17 ); and

[0028] Figure 18 and Figure 19 are respectively a schematic diagram of elements in an embodiment of the diagnostic medical device ( Figure 18 ) and a swim - lane diagram of the process flow through these elements when the device is in operation ( Figure 19 ). API = Application Programming Interface; LUIS = Language Understanding Intelligence Service; CA = Computing Algorithm; NLP = Natural Language Processing. DETAILED DESCRIPTION

[0029] The present invention is understood by reference to the following detailed description, which should be read in conjunction with the accompanying drawings. It should be understood that the detailed description of the various embodiments below is by way of example only and is not meant to limit the scope of the present invention in any way. In the above summary, in the following detailed description, in the following claims, and in the drawings, specific features of the present invention (including method steps) are shown by way of reference. It should be understood that the disclosure of the present invention in this specification includes all possible combinations of these specific features, not just those explicitly described. For example, when a particular feature is disclosed in the context of a particular aspect or embodiment or particular claim of the present invention, that feature can also be used or combined, where possible, in the context of other particular aspects and embodiments of the present invention and throughout the invention. The term "comprising" as used herein and its grammatically equivalent synonyms mean that other components, elements, steps, etc. are present optionally. For example, the term "comprising" (or "including") components A, B, and C can consist of (i.e., only contain) components A, B, and C, or can contain not only components A, B, and C but also one or more other components. When a method including two or more defined steps is referred to herein, the defined steps can be performed in any order or simultaneously (unless the context excludes such possibility), and the method can include one or more other steps performed before any of the defined steps, between two of the defined steps, or after all of the defined steps (unless the context excludes such possibility).

[0030] As used herein, the term "at least" followed by a number indicates the start of a range beginning with that number (which can be a range with an upper limit or without an upper limit, depending on the variable being defined). For example, "at least 1" means 1 or greater than 1. The term "at most" followed by a number as used herein indicates the end of a range ending with that number (which can be a range with a lower limit of 1 or 0, or without a lower limit, depending on the variable being defined). For example, "at most 4" means 4 or less than 4, and "at most 40%" means 40% or less than 40%. In this specification, when a range is specified as "(first number) to (second number)" or "(first number)-(second number)", this means a range with the lower limit being the first number and the upper limit being the second number. For example, 25 to 100 millimeters means a range with a lower limit of 25 millimeters and an upper limit of 100 millimeters. The embodiments set forth below represent the necessary information for those skilled in the art to practice the present invention and show the best mode of practicing the present invention. Additionally, the present invention does not require that all advantageous features and all advantages be incorporated into every embodiment of the present invention.

[0031] Now introduce Figures 1 - 19, a brief description of various components of the present invention will be discussed. The inventors have disclosed that a computational algorithm (CA) 2, such as an artificial neural network (ANN), can be used as part of a medical device 3 to predict a diagnosis 20 of a disease 4 based on one or more individual symptoms 6, historical elements 8, examination results 10, and / or diagnostic test results 12 (collectively referred to as "data inputs 14"). Preferably, each data input 14 receives its respective prediction weight 16 for analysis.

[0032] The inventors have observed that certain diseases 4 can be facilitated for identification based on a set of highly concurrent symptoms 6 (e.g., syndrome 18). While the technical definition of "syndrome 18" only includes symptoms 6, the term is used herein in a broader, more colloquial sense as a collection of multiple individual data inputs 14 including, for example, symptoms 6, historical elements 8, examination results 10, and / or diagnostic test results 12. This broader definition is more closely related to medical practice.

[0033] In particular, the diagnosis 20 of certain diseases 4 in the field of neurology can be predicted by identifying a syndrome 18, which itself can be used as an individual data input 14 for the computational algorithm 2 or supplement the diagnostic assessment of the computational algorithm 2. In clinical neuroscience, focal brain injuries typically involve multiple discrete neural structures involved in functional networks, where these structures and networks are closest (if not overlapping) in physical space and otherwise may have little or no functional relationship. Taking the brainstem as an example, the brainstem is used to directly connect the larger forebrain to the rest of the body through cranial nerves and indirectly connect to the rest of the body by projecting to the spinal cord, and most non-cognitive neural functions can be located therein. Therefore, even a minor focal injury to the brainstem may cause obvious symptoms, but manifest in a unique way based on the injured site of the brainstem. A specific example is a unilateral injury to the lateral part of the lower brainstem (medulla oblongata), which causes a set of symptoms 6, examination results 10, and diagnostic examination results 12, called the lateral medullary syndrome 18, as Figure 1As shown. Lateral medullary syndrome 18 is closely related to occlusion of the vertebral artery or the posterior inferior cerebellar artery. Therefore, it almost always represents an ischemic stroke 4, i.e., the disease that the patient 22 has, and will be correctly identified as having syndrome 18. Therefore, it will be considered by clinicians as a typical stroke syndrome 18. Other typical neurological syndromes 18 are more strongly associated with other diseases 4 (such as seizures, demyelinating diseases such as multiple sclerosis or head trauma). Then, a computational algorithm 2, such as an ANN, for the purpose of diagnosing ischemic stroke 4 can identify the diagnosis 20 of ischemic stroke 4, thereby excluding the diagnosis 20 of ischemic stroke 4 for any individual 22 with typical symptoms 18 diagnosed by the medical device 3 as having a non-ischemic stroke condition (referred to as a "stroke-like attack" disease 24).

[0034] According to one embodiment of the currently claimed medical device 3, it is a computational algorithm 2, such as an ANN, where the syndrome 18 is rated as a separate data input 14 and accompanied by one or more symptoms 6, historical elements 8, examination results, and / or diagnostic test results 12 that an individual 22 has. The weight 16 of the input of the typical syndrome 18 is preferably high (e.g., 0.9 or greater), and has a positive predictive effect on the target disease. Then, whether the definition of the classical syndrome 18 is met or not will be considered a "key definition 26" for the ANN diagnostic evaluation. Taking acute ischemic stroke (AIS) as an example, the classical stroke syndrome 18 will be used as the key definition 26 for evaluating the patient 22, as Figure 5 shown.

[0035] Similarly, other key definitions 26 may include the definitions / diagnostic criteria / characteristics of medical conditions or syndromes similar to the target disease. Using an acute ischemic stroke diagnostic computational algorithm as an example, the stroke-like attack condition 24 will include diseases that are commonly misdiagnosed as acute ischemic stroke in a diagnostic setting, such as seizures, hemorrhagic stroke, migraine, or head trauma. See Figure 2 . The classical stroke-like attack syndrome 24 as the key definition 26 will preferably have a high weight 16 (e.g., 0.9 or greater), but has a negative predictive effect on the target disease 4 (here, acute ischemic stroke).

[0036] In another embodiment, the currently claimed invention includes a computational algorithm 2, such as an ANN, where for diagnostic purposes, other key definitions 26 will be considered as data inputs 14. These include predefined definitions for whether the following conditions are met or not: (i) acute / sudden onset; (ii) persistence or remission of symptoms 6, examination results, and / or diagnostic test results.

[0037] In another embodiment, the computational algorithm 2 may relatively rank the weights 16 of the data inputs 14 as follows: the key definition 26 is greater than the single symptom 6, the single symptom 6 is greater than the single test result 10, and the single test result 10 is greater than the medical history element 8. In some embodiments, the diagnostic test result 12 may be used as the key definition 26. For example, the absence of intracranial hemorrhage in a CT scan or other diagnostic test evaluation is defaulted or excluded and defined as "ischemia".

[0038] In other embodiments, meeting or not meeting the definition of the classical syndrome 18 is not used by the computational algorithm 2 as a data input 14, but rather to confirm the diagnosis 20 of the computational algorithm or invalidate the diagnosis 20 of the computational algorithm, or generate a disease with an indeterminate diagnosis 20 that requires further evaluation by the on-duty doctor 28. In other embodiments, the definition of the classical syndrome 18 must be met first before the computational algorithm 2 can evaluate the data input 14 to arrive at a diagnosis 20. In still other embodiments, achieving or failing to achieve the definition of the classical syndrome 18 determines the computational algorithm 2 employed in the patient assessment 30, where multiple classical syndromes 18 selectively employ the computational algorithm 2 from multiple computational algorithms 2.

[0039] Now introduce Figures 3 - 10 , which shows another embodiment of the currently disclosed medical device 3, namely an artificial intelligence medical device 3 that employs multiple computational algorithms 2.

[0040] The artificial intelligence-based diagnostic medical device 3 (including "software as a medical device 3") may include a computational algorithm 2, which is configured as, for example, an artificial neural network, a support vector machine, and a Bayesian algorithm. A single computational algorithm 2 may be used, for example, to predict the diagnosis 20 of a disease 4 based on data inputs 14 such as symptoms 6, medical history elements 8, test results 10, and / or diagnostic test results 12.

[0041] According to another embodiment of the disclosed medical device 3, an artificial intelligence-based medical device 3, which will be further described below, has one or more computational algorithms 2 that coordinate the prediction of a diagnosis 20 of a disease 4 using methods such as consensus, majority, or other predefined thresholds. In some embodiments of the medical device 3, if all of the computational algorithms 2 employed do not reach a diagnosis 20, then the diagnosis 20 is not provided to the patient 22 or healthcare provider 28, or the diagnosis 20 provided is accompanied by an alert 32 or warning message, such as all computational algorithms 2 not reaching a diagnosis 20. And, in some embodiments, if all of the computational algorithms 2 employed do reach a diagnosis 20, then the diagnosis 20 is provided to the patient 22 or healthcare provider 28. In other embodiments, if at least a majority of the computational algorithms 2 do not reach a diagnosis 20, then the diagnosis 20 is not provided to the patient 22 or healthcare provider 28, or the diagnosis 20 is provided with an alert 32 or warning message. And in other embodiments, if at least a majority of the computational algorithms 2 do reach a diagnosis 20, then the diagnosis 20 is provided to the patient 22 or healthcare provider 28. In still other embodiments, if one or more computational algorithms 2 acting as a "gatekeeper" 34 provide a specific output 38, this will allow other computational algorithms 2 to determine a diagnosis 20, or will allow other computational algorithms 2 to provide a diagnosis 20, or in the case where a computational algorithm 2 acts as a "poison pill defense" 36, provide a specific output 38 that can prevent other computational algorithms 2 from determining a diagnosis 20 that might otherwise be determined. In such cases of "gatekeeper" 34 and "poison pill defense" 36, certain computational algorithms 2 or other computer processes may have a specific purpose in addition to identifying the target primary disease 4 (such as a stroke), for example, detecting a head injury, identifying seizure activity on an electroencephalogram, measuring increased intracranial pressure, or detecting a motor vehicle accident in an emergency. A similar seizure disorder 24 will cast doubt on the diagnosis 20 of a stroke, while in other cases, a diagnosis 20 of a stroke will be reached by different computational algorithms 2 in the artificial intelligence medical device 3. In additional embodiments, the computational algorithms 2 can provide two, three, four, five, or more diagnoses 20, each with its respective likelihood of being valid, and preferably, also with data inputs 14 in favor of each diagnosis 20 and data inputs 14 skeptical of each diagnosis 20, and preferably, also with data inputs 14 that can resolve any or all of the missing data for the diagnoses 20, especially data inputs 14 indicating whether a diagnosis 20 is correct or incorrect, such as the gatekeeper 34 and poison defense 36.

[0042] Each of the multiple computational algorithms 2 of the proposed artificial intelligence-based diagnostic medical device 3 can use their respective data inputs 14 for training and / or for reference. Alternatively, some or all of the computational algorithms 2 can perform calculations based on a common set of data inputs 14. In some embodiments of the medical device 3, the multiple computational algorithms 2 can have substantially the same original structure and / or code, but differ due to the different datasets they are trained on, and thus have different weights 16 for the same data input 14. For example, two initially identical computational algorithms 2 (such as artificial neural networks) trained on different patient databases 40 for a particular disease 4 will become different. These two algorithms will adjust the corresponding weights 16 assigned to the data input 14 based on the prediction weights 16 for a single data input 14 in a single patient database 40 for a particular disease 4.

[0043] In some embodiments of the proposed machine learning medical device 3, multiple computational algorithms 2 and / or analysis processes participate in a serial manner to improve the diagnosis 20, as Figure 5 shown. The improvement of the diagnosis 20 may continue, for example, by one computational algorithm 2 determining that a stroke is the cause of a neurological emergency in patient 22, then by a second computational algorithm 2 determining within a predetermined time limit that the stroke is acute / sudden onset, then by a third computational algorithm 2 determining that the acute onset is caused by cerebral ischemia, and then by a fourth computational algorithm 2 determining whether patient 22 is suitable for medical treatment or surgical / endovascular intervention treatment based on the indications and contraindications of the treatment. In some embodiments, the coordinated activities of the computational algorithms 2 are substantially like a decision tree or process Figure 1 with multiple independent or semi-independent decision points, where the decision calculations are performed at the decision point locations. If the multiple computational algorithms 2 of the medical device 3 cannot arrive at a high diagnostic value (e.g., high probability / high certainty) for determining the diagnosis 20 for patient 22, other embodiments of the medical device 3 will refer the assessment 30 of patient 22 to the physician 28 or other healthcare provider. In some embodiments of the medical device 3, the diagnostic accuracy / agreement of the diagnosis 20 achieved by the artificial intelligence-based diagnostic medical device 3 with the diagnosis 20 of the physician 28 is at least 85%, 90%, 92%, 94%, 95%, or 96% before the diagnosis 20 is considered to have a sufficiently high probability / certainty.

[0044] In other embodiments of the medical device 3, the intervention of the computational algorithm 2 of the individual 22 is flexible and can be adjusted such that those computational algorithms 2 with high diagnostic confidence will be used in patient assessment 30 and the diagnostic decision-making process, while, additionally, less efficient computational algorithms 2 will only be trained on the data of the patient 22 for possible future use and will not contribute to the diagnostic assessment. In such an embodiment, the use of a particular computational algorithm 2 as part of the plurality of computational algorithms 2 for diagnostic purposes can be changed or adjusted over time based on which computational algorithm 2 is most desired in terms of sensitivity, specificity, positive predictive value, negative prediction, and / or consistency / agreement rate with other computational algorithms 2. In one such embodiment of the medical device 3, a computational algorithm 2 with machine learning capabilities trained on retrospectively collected records of the patient 22 is replaced with a computational algorithm 2 trained on prospectively collected records of the patient 22. This embodiment can gradually replace the retrospectively trained computational algorithm 2 with the prospectively trained computational algorithm 2, for example, in a manner proportional to the number of records of the patient 22 used for training, or suddenly replace it with the prospectively trained computational algorithm when a predetermined threshold is reached. In some embodiments, different computational algorithms 2 prospectively collect different data inputs 14 to establish a prospective database 40 that they and / or other computational algorithms 2 can use in training and / or decision-making calculations.

[0045] Although the technical definition of "syndrome 18" only refers to the symptoms 6 reported by the patient 22, in this article, we use the term in a broad sense to refer to a collection of multiple symptoms 6, historical elements 8, examination results 10, and / or diagnostic test results 12 (collectively referred to as "syndrome elements 42"). A broader definition than the technical definition is more suitable for medical practice and is more closely related to medical practice.

[0046] Certain diseases are particularly easy to identify because of the presence of syndrome 18 and can even be pathologically identified by syndrome 18 (“classical syndrome 18”). In this way, the diagnosis 20 of certain medical diseases 4 in the field of the nervous system is particularly easy to identify. In clinical neuroscience, focal brain injury / dysfunction typically involves multiple discrete neural structures that participate in important aspects of anatomically distributed functional networks, where these neural structures are physically close (if not overlapping), but have no or little functional relationship. Taking the brainstem as an example, the brainstem is a part of the brain that directly connects the larger forebrain to the body through cranial nerves and indirectly serves this connection by projecting to the spinal cord. Most non-cognitive neural functions can be located in the brainstem. Therefore, even a small focal injury to the brainstem will produce many neurological abnormalities in a way unique to the nature of the damaged part of the brainstem and the pathophysiological mechanism that caused the injury. An example is that a unilateral injury to the lateral part of the lower brainstem (medulla oblongata) causes a set of syndrome elements 42 called the lateral medullary syndrome 18, such as Figure 1 as shown. The lateral medullary syndrome 18 is closely related to the occlusion of the vertebral artery or the posterior inferior cerebellar artery. Therefore, it represents an ischemic stroke. Also for this reason, clinicians will regard it as a “classical stroke syndrome 18” or “classical ischemic stroke syndrome 18”, and it is very likely an ischemic stroke that requires no further diagnostic evaluation 30.

[0047] Given the high predictive value of classical syndrome 18, its presence or absence can be used as a single data input 14 for one or more computational algorithms 2 in an artificial intelligence diagnostic medical device 3. However, not all syndrome elements 42 may appear in every typical patient 22 considered to be a typical syndrome 18. In order to weight 16 the single data input 14 of the typical symptoms 18 for computational analysis or to evaluate the likelihood that a patient 22 has typical symptoms 18, the inventors disclose a calculation based on the number and prevalence of syndrome elements 42 present in a given patient 22 relative to the population average. For some embodiments of the medical device 3, the inventors disclose a calculation in which: 1) syndrome elements 42 recorded, for example, in medical literature are assigned a score that is proportional to the prevalence in the affected population 22 known / recorded to have classical syndrome 18; 2) it is determined whether the patient 22 being evaluated has or does not have syndrome elements 42; 3) then, the weight 16 of the data input or the probability of the diagnosis 20 of classical syndrome 18 is calculated as the score representing the syndrome elements 42 identified in the evaluated patient 22 divided by the total score of all elements of this syndrome 18 included in the historical definition of this syndrome 18, resulting in a percentage.

[0048] In Figures 4A - 4DA case example of a score-based assessment of classical syndrome 18 is shown. Other classical neurological syndromes 18 are more closely associated with seizures, demyelinating diseases (such as multiple sclerosis), brain tumors, or brain trauma, similar to seizure disorders 24 or counterexamples to these diseases in the case of stroke 4. In one embodiment of an artificial intelligence-based medical device 3, the aim is to obtain a diagnosis 20 of stroke 4, and it may be necessary to consider, evaluate, or identify and thus rule out seizure-like syndromes 24 (the classical syndromes 18 of these non-stroke diseases) or stroke-like seizures 24. The seizure-like syndromes 24 may be negative factors that reduce the likelihood of arriving at a diagnosis 20 of having a specific disease 4 (such as ischemic stroke). The presence of seizure-like syndromes 24 may also be a full stop that halts subsequent calculations, thus completely preventing the obtaining of a diagnosis 20 of having a specific disease 4. In addition, or alternatively, if the likelihood of other diagnoses 20 (such as seizure-like diseases 24) is higher than a certain level, such as 15%, 25%, 50%, 75%, or 90%, the medical device 3 may arrive at a diagnosis 20 of having disease 4, but it includes an alarm 32.

[0049] In some embodiments of the medical device 3, a patient 22 is only evaluated as having classical syndrome 18 when a predetermined number or proportion of syndrome elements 42 are identified during an initial screening of the patient 22. For example, one, two, or three of the most common syndrome elements 42 are found in a patient 22 diagnosed with syndrome 18, or the syndrome elements 42 found in a patient 22 diagnosed with syndrome 18 reach one-quarter or even one-half. Figure 4A The score-based system shown in -D can similarly be used to trigger a more comprehensive assessment of whether a patient 22 has a classical syndrome 18, where the assessment 30 only begins when the score representing the syndrome elements 42 identified in the patient 22 reaches a certain proportion or is in the majority.

[0050] Referring to Figure 6, showing the hierarchical computing algorithm 2 system of the medical device 3. As shown in the figure, in various databases 40 composed of records of patients 22 suffering from the same disease 4, there may be inconsistent data inputs 14. In the figure, "Y" indicates that there is data in the displayed box, while "N" indicates that there is no data in the displayed box. Taking ischemic stroke 4 as an example, many data inputs 14 may exist in all databases 40 (e.g., age, atrial fibrillation) because it is a known fact that they can strongly predict having a given disease 4. Some other data inputs 14 may exist in some but not other databases 40, such as alcohol consumption or family history of stroke. This can be a challenge, namely, how to serially train the disclosed medical device 3 on databases 40 that do not all have the same data inputs 14. Estimation of missing data can be problematic. One way to train the computing algorithm 2 of the medical device 3 to improve the accuracy of diagnosis 20 is to repeatedly cycle through the patient databases 40, including looking up databases 40, accessing databases 40, training using databases 40, and then searching other databases 40. One benefit of this embodiment of the training plan is that it maximizes the machine learning ability, while a potential drawback is that it may lose / dilute / overwhelm the existing training in successive training cycles.

[0051] The disclosed medical device 3 can utilize various databases 40. The FABS database 40 (related to the FABS scoring system), the FAST-MAG (Stroke Treatment - Field Administration of Magnesium) database 40, and the GWTG (Get With The Guidelines) database 40 are shown only as examples. Additional and / or other databases 40 can be used based on their availability and applicability for a specific disease.

[0052] The inventors disclose multiple embodiments for training the medical device 3 and obtaining a diagnosis 20 by utilizing various databases 40. The first embodiment is serial training using a database 40 that uses all data elements and allows dilution of uncommon data inputs 14. One advantage of this embodiment is that its implementation is very simple. A potential weakness is that it may dilute previous training efforts, may ignore the value of data inputs 14 that are difficult to collect, and may become infeasible due to the need to continuously estimate and delete data inputs 14. The second embodiment is to train only on common data inputs 14 that are available in all databases 40. The advantage of this embodiment is that its implementation is very simple. A potential weakness is that this method may discard potentially useful data inputs 14 even if the data inputs 14 have strong or only weak predictive power. The third embodiment is serial transfer learning training with or without training from the largest to the smallest databases 40. The advantage of this embodiment is that it protects previous training, and it presumably starts with the most accurate single training estimate (i.e., the largest database 40) first, and it limits variance. The potential weakness of this embodiment is that it may depend on having the largest database 40 at the beginning and the uncertainty of the partially frozen computational algorithm 2 during training may mean trial and error. The fourth embodiment is to estimate values for data inputs 14 that are missing in all databases 40. The advantage of this embodiment is that it is not complex for certain data inputs 14. The potential weakness of this embodiment is that it may produce confounding interference, such as some data inputs 14 cannot be estimated, such as it may dilute the value of previous training, and such as there may be a large number of missing data inputs 14 resulting in potentially greater unreliability. The fifth embodiment is to estimate missing data inputs 14 for smaller databases 40 based on the largest database 40, then fuse all databases 40 and train on the combined database 40. The advantage of this embodiment is that it creates the largest-scale database 40 and potentially has less unreliability compared to the fourth embodiment. The potential disadvantage is that it assumes that the missing data inputs 14 have a low diagnostic value and that the missing data inputs 14 can be estimated. The sixth embodiment sets a threshold for data inputs including data input 14, for example, the data input 14 must be present in multiple databases 40 and / or must be a risk factor for an acceptable disease 4 (such as stroke). The advantage of this embodiment is that it is based on knowledge of established risk factors. The potential weakness is that it may eliminate data elements that are overlooked or whose current value is unknown. The seventh embodiment is to train on the largest database 40 and then validate on smaller databases 40. The advantage of this embodiment is that it does not exclude other possible designs. The disadvantage is that it may be necessary to estimate missing data inputs 14 in the validation database 40, thus limiting the value of the validation process.The eighth embodiment is to group together rare data inputs 14 (e.g., classifying heart rate and body temperature as "non-blood pressure vital signs"). The advantage of this embodiment is that it is simple. The potential weakness of this embodiment is that the unique predictive value of the rare data inputs 14 may be lost.

[0053] The grouping computational algorithm 2 can include one or more ANNs, support vector machines (SVMs) (including NuSVM and linear SVM), and Naive Bayes (NB) algorithms (including Gaussian NB) and polynomial NB computational algorithms. The grouping computational algorithm 2 may include multi-class SVM, directed acyclic graph SVM (DAGSVM), structured SVM, least squares support vector machine (LS-SVM), Bayesian SVM, transductive support vector machine, support vector clustering (SVC), classification SVM type 1 (also known as C-SVM classification), classification SVM type 2 (also known as nu-SVM classification), regression SVM type 1 (also known as ε-SVM regression), regression SVM type 2 (also known as nu-SVM regression). In one embodiment of the device, the NB algorithm and the SVM algorithm must reach a consistent diagnosis before a diagnosis can be established. In other embodiments of the device, different combinations of 2, 3, 4, 5 or more computational algorithms must reach a consistent diagnosis before a diagnosis can be established. During the first stage, the device can use the summary database 40 to train the main computational algorithm group 44, and the summary database 40 only includes the data inputs 14 shared by all databases 40, see Figure 6 , that is, the data inputs 14 that are representative in each database 40. In the second stage or second domain, the medical device 3 can use all the data inputs 14 from each database 40 to train a group of individual computational algorithms 2, where each database 40 trains its respective computational algorithm group (secondary computational algorithm group 46). The main and secondary computational algorithms 2 can be the same or different types of algorithms.

[0054] Refer to Figure 7, in this case, the main computational algorithm group 44 first attempts to obtain a diagnosis 20 based on the common and highly predictive data inputs 14 of known disease 4 risk factors. If it is uncertain, or if the diagnostic score 48 does not reach the first value 50, then the case of patient 22 can be referred to the secondary computational algorithm group 46. Then, other data inputs 14 can be obtained through a front-end patient interface 50 of the device, a third party 74, and / or by retrieving the records of patient 22 as required by the secondary computational algorithm group 46. Then, for example, using the consistent results of the secondary computational algorithm group 2 as a tiebreaker or by evaluating the superimposed probabilities based on the calculations of the main computational algorithm group 44, the diagnosis 20 is re-evaluated. If the process re-evaluation 30 cannot conclusively give a diagnosis 20 of the presence of disease 3 and / or cannot conclusively give a diagnosis 20 of the absence of disease 3, then the patient 22 may later be referred to a physician 28 (a neurologist in this embodiment) for evaluation 30.

[0055] Now introduce Figures 11 - 14 , further discuss embodiments of the disclosed medical device 3. In some embodiments, the device inputs 56 obtained by the sensors 62 (e.g., microphones and cameras) and the direct interface 64 can be passed through a feature extraction module (also referred to as a feature extractor), which converts the device inputs 56 into "features" 58 that are a valid digital representation of the device inputs for training the computational algorithm 2. The interface can be, for example, a keyboard or a touch screen. In addition to the features 58 of the device inputs 56, "labels" 60 can be provided for the symptoms. The "labels" can include, for example, the degree of slurring of a speech sample or the degree of drooping of the front half of the face relative to the rear half when smiling. In addition to data 14 such as user input, other types of markers of visual or auditory images / videos or recorded diagnostic symptoms 6 from patient 22 can also be used, including binary or scaled or ranged values directly input into the medical device 3. This includes "yes" or "no" binary answers to questions posed to the medical device 3, such as answers from "0 - 10" or analog answers (e.g., real or virtual sliders), to ranged answers to questions posed to the medical device 3. The neural network 66 can be trained by receiving, processing, and learning from multiple device inputs 56 and their associated labels 60 or groups of labels 60 to allow the device to estimate the diagnosis 20 of patient 22.

[0056] In some embodiments, the construction of a neural network architecture can have a sufficient number of layers 52 and nodes 60 within each layer 52 such that when trained with the input data 14 obtained by the sensors 62 (e.g., cameras and microphones) and the user interface (UI) 64, it can model the diagnosis 20 with sufficient accuracy. Figure 11Shows an example of the architecture of neural network 66, where the features extracted therefrom are provided as neural inputs 68 (f1, f2, ..., fL) to one or more lower layers 52, one or more long short-term memory (LSTM) layers 52, and one or more deep neural network (DNN) layers 52 to estimate a diagnostic score 48. Various types of neural network layers 52 can be implemented within the scope of the present disclosure. For example, as an alternative or supplement to the DNN layer 52, one or more convolutional neural network (CNN) layers 52 or LSTM layers 52 can be implemented. In some cases, in addition to or as part of one or more neural network layers, various types of filters can be implemented, such as infinite impulse response (IIR) filters, linear prediction filters, Kalman filters, and the like.

[0057] Figure 12 Shows a specific example of an embodiment of neural network 66, which has four neural network layers 52, namely layers 1, 2, 3, and 4, for processing features 58 extracted from device input 56. In Figure 12 it, two graphs are shown, namely graph n and graph n+L. It should be understood that within a given time length, the device input 56 can be represented as multiple graphs, and the size of the graph can represent the diagnostic score 48. For graph n, the first layer 52, i.e., layer 1, includes device inputs for multiple data inputs 14 for each diagnosis, as shown in the first diagnostic score 48. Similarly, for graph n+L, layer 1 includes device inputs 56 for multiple data inputs 14 from each diagnosis, as shown in the second diagnostic score 48. Other information about patient 22 may be included in layer 1. To vary the data inputs 14, layer 1 of graph n and graph n+L may include device inputs 56 representing symptoms 6 that change over time. In one embodiment, when there are a sufficient number of nodes 54 or units in layers 2 and 3, neural network 66 will be able to acquire knowledge or diagnostic accuracy and predict diagnosis 20.

[0058] At least one layer 52 of neural network 66 may be required to process complex numbers. In one example, complex numbers may be processed in the second layer of neural network 66. Complex numbers can be in the form of real and imaginary parts, or alternatively in the form of magnitude and phase. For example, in the second layer of the neural network, each unit or node 54 can receive complex inputs and produce complex outputs. In this example, a neural unit with complex inputs and complex outputs may be a relatively straightforward setup for the second layer. In one example, the net result U within the complex unit is given by: U = ∑ i W i X i+V, where Wi is a complex-valued weight 16 that connects complex-valued inputs, and V is a complex-valued threshold. To obtain a complex-valued output signal, the net result U is converted to real and imaginary parts, and these real and imaginary parts are passed through an activation function f R (x) to obtain the output f out , with the formula where f r (x) = 1 / (1 + e^-x); for example, x ∈ R. Various other complex-valued computations can also be implemented within the scope of the present disclosure.

[0059] In another embodiment, the first and second layers of the neural network 66 may involve complex number computations, while the upper layers 52, such as the third and fourth layers, may involve real number computations. For example, each unit or node 54 in the second layer of the neural network 66 can receive complex inputs and produce real outputs. Various schemes can be implemented to generate real outputs based on complex inputs. For example, one method is to implement a complex input-complex output formula and make the complex output real by simply taking the magnitude of the complex output: Alternatively, another method is to apply the activation function to the absolute value of the complex sum, i.e., f out = fR(|U|). In another alternative method, each complex input feature is decomposed into magnitude and phase or real and imaginary parts. These components can be regarded as real input functions. In other words, each complex number can be regarded as two separate real numbers to represent the real and imaginary parts of the complex number, or, two separate real numbers to represent the magnitude and phase of the complex number.

[0060] Embodiments of the presently disclosed subject matter can be implemented in and used with various components and network architectures. For example, as Figure 14 shown, the medical device 3 neural network 66 can include one or more computing devices 70 for implementing embodiments of the above-described subject matter. Figure 13An example of a computing device 70 suitable for implementing embodiments of the present disclosure is shown. The computing device 70 can be, for example, a desktop or laptop computer, or a mobile computing device, such as a smart phone, a tablet computer, a video conferencing / telemedicine system, and so on. The computing device 70 can include a bus that interconnects the main components of the computer, such as a central processing unit, memories such as main memory (RAM), read-only memory (ROM), flash RAM, and so on, a user display (such as a display screen), a user input interface, which can include one or more controllers and associated user input devices, such as a keyboard, a mouse, a touch screen (which can be considered part of interface 64), and so on, a storage device such as a hard disk drive, a flash memory, a removable media component operable to control and receive optical discs, flash drives, and so on, and a network interface operable to communicate with one or more remote devices via a suitable network connection.

[0061] As previously mentioned, the bus allows data communication between the central processing unit and one or more memory components, which can include RAM, ROM, and other memories. Generally, RAM is the main memory in which the operating system and applications are loaded. The ROM or flash memory component can contain, among other codes, a basic input / output system (BIOS) that controls basic hardware operations, such as interactions with peripheral components. Applications residing in the computer are typically stored on and accessed via a computer-readable medium, such as a hard disk drive (such as fixed memory), an optical disc drive, a floppy disk, or other storage media.

[0062] The fixed memory can be integrated with the computer or can be stand-alone and can be accessed via other interfaces. The network interface can provide a direct connection to a remote server via a wired or wireless connection. The network interface can use any suitable technologies and protocols readily understood by those skilled in the art to provide such a connection, including digital cellular phones, Wi-Fi, near field, and so on. For example, the network interface can allow the computer to communicate with other computers via one or more local area networks, wide area networks, or other communication networks, which will be described in further detail below.

[0063] Many other devices or components (not shown) can be connected in a similar manner (such as a document scanner, a digital camera, etc.). Conversely, to practice the present disclosure, Figure 13 not all of the components shown need to be present. The components can be interconnected in a manner different from that shown. Such as Figure 13The operation of the computer shown is well known in the art and will not be discussed in detail in this application. The code for implementing the present disclosure can be stored in a computer-readable storage medium, such as one or more memories, fixed memories, removable media, or at a remote storage location.

[0064] In a more general sense, various embodiments of the presently disclosed subject matter may include or be embodied in the form of computer-implemented processes and apparatuses for practicing those processes. The embodiments can also be embodied in the form of a computer program product having computer program code embodied therein, the instructions being embodied in a non-transitory or tangible medium (such as a floppy disk, CD-ROM, hard drive, USB (Universal Serial Bus) drive, or any other machine-readable storage medium), such that when the computer program code is loaded into and executed by a computer, the computer becomes an apparatus for practicing the embodiments of the disclosed subject matter. The embodiments can also be embodied in the form of computer program code, for example, whether stored in a storage medium, loaded into and executed by a computer, or transmitted over some transmission medium (such as via wire or cable, via fiber optic, or via electromagnetic radiation), such that when the computer program code is loaded into and executed by a computer, the computer becomes an apparatus for practicing the embodiments of the disclosed subject matter. When implemented on a general-purpose microprocessor, the computer program code segments configure the microprocessor to create specific logic circuits.

[0065] In some configurations, a set of computer-readable instructions stored on a computer-readable storage medium can be implemented by a general-purpose processor, which can transform the general-purpose processor or a device incorporating the general-purpose processor into a special-purpose device configured to implement or execute the instructions. The embodiments can be implemented using hardware, which can include a processor such as a general-purpose microprocessor or a special application integrated circuit (ASIC) that embodies all or part of the techniques of the embodiments of the disclosed subject matter in hardware or firmware. The processor can be coupled to a memory such as RAM, ROM, flash memory, a hard disk, or any other device capable of storing electronic information. The memory can store instructions suitable for execution by the processor to perform the techniques of the embodiments of the disclosed subject matter.

[0066] In some embodiments, as Figure 14The illustrated microphone and camera can be implemented as part of a network of sensors 62. For example, these sensors 62 can include microphones for sound detection and cameras for visual detection, and can also include other types of sensors 62. Generally, a "sensor 62" refers to any device that can acquire information about its environment. Sensors 62 can be described by the type of information they collect. For example, the types of sensors 62 disclosed herein can include waveforms, chemical emissions, motion, smoke, carbon monoxide, proximity, temperature, time, physical orientation, acceleration, position, entry, presence, pressure, light, sound, and so on. Sensors 62 can also be described according to the specific physical device for obtaining environmental information. For example, an accelerometer can acquire acceleration information and thus can be used as a general motion sensor 62 or an acceleration sensor 62. Sensors 62 can also be described according to the specific hardware components used to implement the sensors 62. For example, a temperature sensor 62 can include a thermistor, a thermocouple, a resistance temperature detector, an integrated circuit temperature detector, or a combination thereof. Sensors 62 can also be described according to the functions performed within an integrated sensor 62 network such as a smart home environment. For example, when a sensor 62 is used to determine a security event such as unauthorized entry, it can be used as a security sensor 62. Sensors 62 can operate with different functions at different times. For example, a motion sensor 62 is used to control lighting in a smart home environment in the presence of an authorized user, and is used to warn of unauthorized or unexpected motion when there is no authorized user or when the alarm system is in an "armed" state, etc. In some cases, sensors 62 can operate as multiple types of sensors 62 sequentially or simultaneously. For example, in the case where a temperature sensor 62 is used to detect temperature changes and the presence or absence of a person or an animal. Sensors 62 can also operate in different modes at the same or different times. For example, a sensor 62 can be configured to operate in one mode during the day and in another mode at night. Another example is that a sensor 62 can operate in different modes based on the state of a home security system or a smart home environment, or under other indications of such systems.

[0067] Generally, the "sensors 62" disclosed herein can include multiple sensors 62 or sub-sensors 62. For example, a position sensor 62 includes both a global positioning sensor 62 (GPS) and a wireless network sensor 62, which provides data that can be associated with a known wireless network to obtain position information. Multiple sensors 62 can be arranged in a single physical space, such as where a single device includes motion, temperature, magnetic, or other sensors 62. Such a space can also be referred to as a sensor 62 or a sensor 62 device. For clarity, when the specific function performed by a sensor 62 or the specific physical hardware used is necessary for understanding the embodiments disclosed herein, such descriptions will be provided when describing the sensor 62.

[0068] In addition to the specific physical sensors 62 for obtaining environmental information, the sensor 62 may also include hardware. The sensor 62 may include environmental sensors 62, such as temperature sensors 62, smoke sensors 62, carbon monoxide sensors 62, motion sensors 62, accelerometers, distance sensors 62, passive infrared (PIR) sensors 62, magnetic field sensors 62, radio frequency (RF) sensors 62, optical sensors 62, humidity sensors 62, pressure sensors 62, microphones, weighing scales, or any other suitable environmental sensors 62. Such sensors obtain the corresponding types of information about the environment in which the sensor 62 is located. The processor can receive and analyze the data obtained by the sensor 62, control the operation of other components of the sensor 62, and handle the communication between the sensor 62 and other devices. The processor can execute instructions stored on a computer-readable memory. The memory in the sensor 62 or another memory can also store the environmental data obtained by the sensor 62. A communication interface, such as Wi-Fi or other wireless interfaces, Ethernet or other local network interfaces, etc., can allow the sensor 62 to communicate with other devices. The user interface can provide information to the user or receive input from the user or the sensor 62. The user interface 64 can include, for example, a speaker to emit an audible alarm, i.e., output 38, when the sensor 62 detects an event. Alternatively, or in addition, the user interface 64 can also include a light that will be activated when the sensor 62 detects an event. The user interface can be relatively very small, such as a display with limited output 38, or can be a full-function interface, such as a touch screen. It will be readily understood by those skilled in the art that the components within the sensor 62 can send and receive information to and from each other via an internal bus or other mechanism. In addition, the sensor 62 can include one or more microphones to detect sounds in the environment. One or more components can be implemented in a single physical arrangement, such as in the case where multiple components are implemented on a single integrated circuit. The sensor 62 as disclosed herein can include other components or may not include all of the illustrative components shown.

[0069] The sensor 62 as disclosed herein may operate within a communication network such as a conventional wireless network or within a network specific to the sensor 62, through which the sensors 62 may communicate with each other or with other dedicated devices. In some configurations, one or more sensors 62 may provide information to one or more other sensors 62, to a central controller, or to any other device capable of communicating with one or more sensors 62 over a network. A central controller may be of general or special purpose. For example, one type of central controller is a home automation network that can collect and analyze data from one or more sensors 62 in a home. Another example of a central controller is a dedicated controller for a subset of functions, such as a security controller, which primarily or specifically collects and analyzes data from sensors 62 as it relates to various security considerations at a location. The central controller may be located locally with respect to the sensors 62 with which it communicates, and the central controller obtains data of the sensors 62 from the sensors 62, such as by arranging the central controller within a house containing a home automation or sensor 62 network. Alternatively, or in addition, the central controller disclosed herein may be remote from the sensors 62, such as by arranging the central controller as a cloud-based system that communicates with a plurality of sensors 62, which may be located at multiple locations and may all be local or remote from each other.

[0070] In addition, the smart home environment may infer which individuals 22 are present in the home and are thus users, and which electronic devices are associated with those individuals 22. In this way, the smart home environment can "learn" who the users are (e.g., an authorized user) and allow the electronic devices associated with those individuals 22 to control the networked smart devices of the smart home environment, including, in some embodiments, the sensors 62 used by or within the smart home environment. Various types of notifications and other information may be provided to the user via messages sent to one or more user electronic devices. For example, messages may be sent via email, short message service (SMS), multimedia message service (MMS), unstructured supplementary service data (USSD), and any other type of messaging service or communication protocol.

[0071] The smart home environment may include communication with devices outside the smart home environment but within the adjacent geographical area of the home. For example, the smart home environment may convey information about the detected movement or presence or absence of people, animals, and any other objects to a central server or cloud computing system via a communication network or directly, and receive in return commands for controlling the environmental lighting accordingly.

[0072] In some embodiments of the medical device 3, the medical device 3 periodically evaluates the disease or abnormal symptoms of the patient 22. If symptoms are detected, a more comprehensive evaluation 30 of the patient 22's device may be automatically triggered, resulting in a diagnosis 20. In one embodiment of the medical device 3, the periodic evaluation 30 involves an evaluation of gait, speech, and limb movement to look for symptoms of neurological abnormalities, such as limping, slurred speech, weakness in one arm, or drooping of one part of the face, respectively. In this embodiment of the medical device 3, the detection of such neurological abnormalities will trigger a comprehensive evaluation of the symptoms 6 and physical examination abnormalities consistent with focal brain injury, such as a stroke. In another embodiment of the device, the change in respiratory rate or pauses in speech of a patient 22 known to have congestive heart failure will be evaluated. These symptoms can indicate a worsening of congestive heart failure. At this time, the condition of the patient 22 will be specifically evaluated, which may include weighing the patient 22. In this embodiment, the device can guide the adjustment of the patient 22's medication based on the diagnosis 20 of worsening congestive heart failure, such as the use of diuretics. In another embodiment of the device, the device will identify shortness of breath or difficulty breathing and regard it as evidence of an impending or ongoing asthma attack. At this time, the patient 22 will be notified to use any available respiratory treatment measures, including inhalers, and / or if the patient 22's condition is severe enough, the device will notify the nursing and / or emergency medical services 28 to assist the patient 22. In another embodiment of the device, the device will identify the following behaviors of the monitored target person 22, such as grasping the chest, a facial expression indicating pain, rapid breathing, flushing, and / or sweating. These behaviors will suggest that the patient has a myocardial infarction. At this time, the device will confirm other symptoms 6 and test results 10 consistent with a myocardial infarction, thereby instructing the patient 22 to take emergency treatment for the myocardial infarction and notifying the ambulance 72 to receive the patient 22 based on the possible diagnosis 20 of myocardial infarction. In another embodiment of the device, the patient 22's routine thermal scan or spot temperature measurement can be used to identify whether there is an abnormal body temperature, which will trigger the medical device 3 to evaluate the patient 22 for symptoms 6 consistent with an infection; based on the diagnosis 20 of the medical device 3, presumptive antibiotic treatment can be administered by the patient 22 himself / herself or provided to the patient 22 by a third party 74 before the evaluation 30 of the patient 22 is made in the physician 28's office or the hospital where the patient is located. In additional embodiments, when the medical device 3 determines that the patient 22 needs medication, the medical device 3 can also directly dispense the medication. In another embodiment, the same medical device 3 can evaluate whether the patient 22 has any or all of the above diseases 4.

[0073] In some embodiments of the medical device 3, any electronic device equipped with a sufficient combination of input devices 76 (such as microphones, thermometers, and one or more cameras), output devices 38 (such as speakers and lights), processors, and / or memory can be used, either alone or in concert, to monitor the patient 22. In some embodiments of the medical device 3, various electronic devices are scattered throughout the patient 22's home to monitor the patient 22.

[0074] In some embodiments of the medical device 3, the medical device 3 passively assesses whether the patient 22 has evidence of having certain diseases 4. When used in an ambulance 72, the medical device 3 can listen to reports provided by an ambulance / emergency medical service (EMS) dispatcher and interpret the reports to indicate whether the next patient 22 to be seen by the ambulance crew has a certain target disease 4, such as a stroke. In that case, for example, the medical device 3 can have the ability to activate itself and perform a comprehensive assessment 30 of the patient 22. In other cases, the medical device 3 can request an opportunity to assess the patient 22 (e.g., in person or by phone), in a manner that involves a conversation between the patient 22 and an emergency medical technician or healthcare provider, and / or a physical examination assessment 30 of the patient 22 by a healthcare provider in such an ambulance 72.

[0075] A first function of one embodiment of the medical device 3 is to diagnose ischemic stroke in a prehospital setting. Based on this diagnosis 20, treatment options can be immediately obtained. A preferred embodiment of the medical device 3 is to determine that the ischemic stroke symptoms 6 and test results 10 have resolved, indicating that the ischemic stroke has subsided and the patient 22 has had a transient ischemic attack. In such a case, the medical device 3 will instruct the patient 22 to take aspirin or other antiplatelet medications before further diagnostic evaluation 30 or arrival at the hospital. Administration of the medication can be completed under the guidance of the medical device 3 before the arrival of any healthcare provider or any professional medical service, including nurses, healthcare providers, or ambulances. In another embodiment, treatment of ischemic stroke, such as a facial nerve stimulator, is safe enough in the event of a hemorrhagic stroke that it can be used on undifferentiated stroke patients 22 without prior neuroimaging evaluation. The medical device 3 can diagnose that the patient 22 has a stroke and then instruct the use of a transcranial magnetic stimulation (TMS) facial nerve stimulator for treatment. In another embodiment, the medical device 3 itself can provide TMS to the patient 22 after diagnosing that the patient 22 has a stroke. In other embodiments, after an initial therapeutic TMS stimulation has been administered to the patient, the medical device 3 will evaluate whether the patient's symptoms 6 and test abnormalities have improved and determine whether the symptoms 6 and test abnormalities that were appropriately treated / benefited from repeated therapeutic TMS stimulation have recurred or new episodes have occurred.

[0076] The diagnosis of stroke 20 is established in the ambulance 72 or elsewhere before arrival at the hospital, so that upon arrival at the hospital, the patient 22 can be immediately subjected to neuroimaging to identify whether there is intracranial hemorrhage, which will establish the diagnosis of hemorrhagic stroke 20, thus avoiding treatment with established therapies for ischemic stroke (such as intravenous tissue plasminogen activator ((rtPA) or surgery with endovascular intubation). Such a system would bypass the emergency department assessment 30, thus facilitating the treatment of patients 22 with ischemic stroke.

[0077] Now introduce Figure 15 , showing another embodiment, namely determining the diagnosis of the classical syndrome 18 based on the cumulative probability of the symptoms 6 and signs of the patient 22. In this embodiment, the proportions of patients 22 with the typical syndrome 18, i.e., the appearance of individual symptoms 6, examination abnormalities 10, or diagnostic test results 12 ("components of the syndrome 18") are compiled into a database 40 or repository. A bias factor can also be assigned to each syndrome element 42 equally or unequally, where the unequal bias factor can be determined by a survey informed by the patient 22, the impact on quality of life, or the decision of one or more healthcare providers. Assuming that all syndrome elements 42 of the classical syndrome 18 require a bias factor of 1.0, then the weight 16 of the data input for the classical syndrome 18 can be adjusted according to the proportions of the syndrome elements 42 of the patient 22. As Figure 15 shown in the example, in the target patient 22 being evaluated, there are ipsilateral limb and gait ataxia, ipsilateral facial numbness, and ipsilateral Horner's syndrome 18, without contralateral hemianesthesia, dysphagia, and dysarthria, and without nausea, vomiting, vertigo, and nystagmus. The relative proportions of the various components of this syndrome 18 are 90%, 50%, 50%, 90%, 20%, and 10%. The device will calculate the potency of the symptoms 6 and examination results 10, as well as the percentage of affected individuals 22 with all components of the syndrome 18 of the patient 22, by subtracting the product of the probabilities from 1. For example, if the patient 22 only presents with ipsilateral limb and gait ataxia and contralateral hemianesthesia, and the bias factor 42 for each syndrome element = 1, then the diagnostic score 48 is 1 - (0.9 * -0.9) = 1 - (0.81) = 0.19. If Figure 15If all of the symptoms 6 and test results 10 therein are present, the diagnostic score 48 according to this embodiment will be 0.996. When a higher diagnostic value 78 of the diagnostic score 48 is reached, a diagnosis 20 of having a specific disease 4 can be made. For example, the diagnostic value can be a value greater than 0.80, greater than 0.85, greater than 0.90, or greater than 0.95. For example, if these higher diagnostic values 78 are between 80% and less than 100% (e.g., for an embodiment with a higher diagnostic value of 0.80, between 0.64 and 0.80, and for an embodiment with a higher diagnostic value of 0.90, between 0.72 and 0.90), the diagnostic score 48 can be marked as unable to make a diagnosis 20 of a specific disease due to lack of specificity or the diagnosis 20 of the specific disease is not entirely accurate, and / or trigger a referral process to refer the patient to a medical professional for a diagnosis 20. Such medium diagnostic scores 48 will be referred to as having a medium diagnostic value 80 and are indeterminate as to whether the patient has a specific disease 4. Other lower limit values of the medium diagnostic value can be 75%, 85%, 90%, and 95% of the value of the higher diagnostic value 78. A lower diagnostic score 48, such as a value less than four-fifths or 80% of the higher diagnostic value 78 (e.g., less than 0.64 for an embodiment with a higher diagnostic value of 0.80, and less than 0.72 and 0.90 for an embodiment with a higher diagnostic value of 0.90), can be diagnosed by the medical device 3 as having a lower diagnostic value 82 and being negative for having a specific disease 4. Other upper limit values of the lower diagnostic value 82 can be 75%, 85%, 90%, and 95% of the value of the higher diagnostic value 78.

[0078] Some embodiments of the medical device 3 aggregate or compile the syndrome elements 42 of each classical syndrome 18 into a database 40 or one or more libraries. For example, multiple libraries can be used to distinguish the classical stroke syndrome 18 from the classical stroke-like seizure syndrome 24. In such embodiments of the medical device 3, the front-end patient interface 64 of the device can collect the initial symptoms 6 informed by the patient 22, and then the medical device 3 can use these initial symptoms 6 to identify the classical syndrome 18 from the library containing these initial symptoms 6. Then, the medical device 3 can evaluate the distribution of other syndrome elements 42 from the selected classical syndrome 18 and sort the syndrome elements 42 according to the number of times they are found in the selected classical syndrome 18. Next, the medical device 3 can then ask the patient 22 whether the syndrome elements 42 are most frequently encountered. Based on the patient 22's answer, the subset of the classical syndrome 18 selected including the initial symptoms 6 is reduced, and only the classical syndrome 18 that also contains the most common syndrome elements 42 is retained. This process is repeated until a single classical syndrome 18 remains, or until there may still be a small group of classical syndromes 18, and all the remaining classical syndromes 18 belong to a single type, such as the classical stroke syndrome 18. At this time, the diagnosis 20 can be output by the medical device or communicated to the patient 22 and / or the healthcare provider through the network. In other embodiments, the patient 22 can be asked whether they have uncommon symptom elements 42 that exist in only a few or one of the initially selected classical syndromes 18, thereby reducing the number of possible classical syndromes 18. In other embodiments, the test results 10 or other data inputs 14 can be used to select among various classical syndromes and to rule out various classical syndromes.

[0079] Now introduce Figure 16 and 17 FIGs. show other embodiments of the computational algorithm 2 for selectively or specifically using the database 40 and training with certain databases 40. In this figure, the lighter gray cells indicate that the value / data input 14 exists in the database 40, while the darker gray cells indicate that there is no value for different elements of different databases 40. For example, in the illustrated embodiment, the FABS database 40 is used and there is a value representing the glucose level, but there is no value representing the medication. The database 40 to be used, which is particularly relevant to the patient 22 in the assessment 30 made by the device, can be selected based on demographic characteristics (age, gender, race, medical history, geographical environment, etc.). Two examples are given in this figure. In Figure 16 FIG., non-geographical / personal demographic feature similarity is used to select the database 40 and its associated computational algorithm most relevant to the patient. In Figure 17In [the system], geographical similarity (e.g., geographical proximity) is used to select database 40 and its associated computational algorithm most relevant to the patient. The demographic similarity or relevance of database 40 to the target patient 22 being evaluated 30 may cause the medical device 3 to adopt a computational process adjusted or adapted according to a certain database 40, even if the data values of that particular database 40 are less than those of another database 40.

[0080] Figure 18 and 19 illustrates other embodiments of diagnosing a disease 4 using the medical device 3. In these embodiments, services (e.g., ) are used as examples of functions or roles, but these services are considered paradigms of other similar services that can be used. For example, there are other companies, such as and the services of other companies.

[0081] In some other embodiments, the medical device 3 can be integrated into one or more household devices 84 or household fixtures or appliances, such as mirrors, clocks, refrigerators, sinks, toilets, chairs, beds, televisions, microwaves, or electric lights, for example including ceiling-mounted light fixtures. Preferably, the household device 84 is a household device that the patient 22 will interact with regularly or frequently or be close to. The household device 84 is preferably connected or connectable to a network such as WIFI, Bluetooth, cellular, and / or the Internet of Things, and it includes sensors 62 capable of passively or interactively recording the patient's 22 condition, such as one or more microphones, cameras, thermal imagers or thermometers, electrocardiogram sensors, photoplethysmograph sensors (for measuring heart rate), and / or for example electromagnetic pulse monitors and other input devices 76 that may interact with the patient 22. After integrating the medical device 3 into the household device, it can have output elements 38, such as screens, lights, and speakers, to prompt the patient 22 to respond or otherwise interact with the patient 22, for example with light, sound, or voice questions. In embodiments where the medical device 3 monitors the stroke perception symptoms 6 or physical signs 10 reported by the patient, for example, a camera can capture video to determine if there is a drooping of the face or a part of the body compared to other parts. A speaker can capture and determine if there is slurring of speech or if the slurring has become more severe when the patient 22 speaks, or if the grammar, syntax, or content of the language is disrupted. For example, this can also be a response to the medical device, which greets the patient 22 with "Good morning" in spoken form through the speaker or in written text form through the interface 64. In addition, the medical device 3 can periodically or continuously monitor the patient's 22 speech, and if slurred speech is detected, the device can automatically trigger an assessment 30 and / or send an alert 32 to a caregiver, healthcare provider, or other first responders and / or a central server or other display screen. The camera can also detect if the patient 22 has a limp in their gait. If the patient 22 has a high risk factor for stroke, the medical device 3 can regularly screen for initial warning symptoms 6 or signs and automatically trigger an assessment 30 and / or send an alert 32 when an initial warning symptom 6 or sign 10 is detected. If an initial warning symptom 6 is detected, or if a diagnosis 20 of having a disease 4 is made, the medical device 3 can also instruct the patient 22 to take emergency treatment or both guide and provide the patient 22 with emergency treatment. In the case of cardiac ischemia or transient ischemic attack, the device can instruct the patient 22 to take aspirin while waiting for a medical professional to arrive. In additional embodiments, the medical device 3 can dispense aspirin or other appropriate emergency medications for a diagnosed patient having a specific disease 4. Medications can be dispensed from a reservoir containing medications for treating one or more different diseases.The medications can be contained in containers or pouches that are color-coded, numbered, named, or otherwise marked so that if the patient 22 needs to take multiple types of medications, it will be clearly indicated which medications are to be taken. For example, the medical device 3 might state, "You have been diagnosed as possibly having a transient ischemic attack. Take one aspirin from the blue pouch marked 'A' in the medication container." The medical device 3 can be self-activated. It is contemplated that the medical device 3 will be used, for example, in nursing and long-term care facilities, as well as in the residences of the elderly.

[0082] The medical device 3 preferably will collect information that directly or indirectly interacts with the patient 22, but also information from other sources such as a third party 74, for example, witnesses to an accident, nursing home staff, family members, medical staff, and the hospital / medical records of the patient 22. If there are inconsistencies in the input data, the medical device 3 can flag the inconsistent information and issue a warning 32 to the physician 28, and / or the medical device 3 can query the source of the inconsistent information as well as other information in an attempt to clarify. For example, if the patient 22 enters that symptom 6 started six hours ago and the nurse enters that symptom 6 started two days ago, this inconsistent information can be flagged by the medical device 3 and then it can be determined by the physician 28 or the medical device 3, for example, by further questioning the source of the inconsistent information, which is the true information. Alternatively or additionally, the accuracy of the answers given by each individual can be weighted 16 based on the truthfulness or accuracy of the other answers given by that individual to the given question or to all questions posed to the individual 22, or based on the truthfulness or accuracy of the answers given by a population of individuals such as nurses or paramedics, or individuals working in a particular hospital.

[0083] In some embodiments of the medical device 3, the medical device 3 will only talk to and visualize the patient 22 as a means of obtaining diagnostic information about the patient 22. In other, more preferred embodiments, the medical device 3 will be able to interact with multiple third parties 74 (individuals other than the patient 22) and other information sources related to the patient 22, such as multiple witnesses when the patient 22 was injured in an accident, the patient 22's family members or healthcare providers, and the sources of the patient 22's medical records. In some embodiments, the medical device 3 includes multiple device units, one of which may be arranged in an ambulance 72, while other devices may be small handheld devices capable of transmitting information such as voice, images, and input data from users (including third parties 74) to the data collection process of the medical device 3. Then, the ambulance paramedics can hand the handheld device to the personnel 74 at the emergency scene, allowing them to interact with the medical device system 3. An independent device can ask the personnel 74 for the necessary information related to the patient 22's condition, while the paramedics attend to the patient 22 or transport the patient 22 to the hospital. The advantage of having an independent medical device unit is that when the device is provided to a third-party user 74, the device may already be connected to a network together with the rest of the medical device system 3 and already loaded with the questions to be asked, which is a crucial advantage when time is of the essence. In other embodiments, the independent medical device unit is directly connected to a communication-enabled device already available to the third party 74, such as their personal mobile phone or other computing device 70, through which information related to the diagnosis 20 of the patient 22 can be queried and communicated. In a preferred embodiment of the present invention, the medical device 3 will collect information about the patient 22 from multiple sources including the patient, third parties, and / or medical records in a parallel, simultaneous, or overlapping manner.

[0084] Paramedics may leave one or more embodiments of the medical device 3, preferably small handheld devices, with third parties 74, such as witnesses or family members, especially when the patient 22 is unconscious when the paramedics arrive. The medical device unit will ask questions, input answers, and send the data wirelessly to the medical device system host for analysis, storage, and diagnostic decision-making. This can be forwarded to the physician 28 at the destination hospital and / or the paramedics transporting the patient 22 to the destination hospital. This can save time and improve the quality of the information. In one embodiment, the third parties 74 can be mailed out after they have completed inputting data on the small handheld medical device unit for return and reuse.

[0085] In another embodiment, a program for asking third-party information can be downloaded onto the smart phone or computing device 70 of the third party 74, or the third party 74 can be brought to a website that asks the third party about information regarding the patient 22. The information will be sent via a network to a medical device central server for compilation and analysis, and then sent to a medical professional 28 or a diagnostic machine. Alternatively, the information can be sent directly or indirectly to the medical professional 28 for compilation and analysis.

[0086] In another embodiment, each or one or more patients 22 diagnosed as having, not having, or including having a particular disease 4 can be tracked according to different data inputs 14 to obtain result data. If the patient 22 receives a definitive diagnosis 20 issued by a doctor 28 or medical professional at a hospital or other healthcare center, the diagnosis 20 can be used to create a post-use patient 22 database 40 to improve the accuracy of the computational algorithm 2 used by the medical device. That is, using, for example, backpropagation and other analysis and calculations, the accuracy of the diagnoses 20 of past users 22 of the medical device 3 can be used to train and / or update the computational algorithm 2 and determine the selection of weights 16 and data inputs 14 (including changes based on demographic characteristics) and the biases required for the activation / transfer functions of the nodes 54 when determining the diagnosis 20 of the current user 22 of the medical device 3. In some preferred embodiments, a portable computing and / or communication device is provided to the patient 22 during and after hospitalization, and the device is capable of providing information regarding the patient's chronic disease to the database 40. In other embodiments, the portable computing and / or communication device provided to the patient 22 can be used to monitor the patient 22's condition and / or summon emergency medical services.

[0087] In the absence of any elements not specifically disclosed herein, the presently claimed invention disclosed herein by way of example can be appropriately and explicitly implemented. Although various embodiments of the present invention have been described in detail, it is apparent that various modifications and changes to those embodiments will be obvious to those skilled in the art. However, it should be clearly understood that such modifications and changes fall within the scope and spirit of the presently claimed invention as set forth in the appended claims. Additionally, the invention described herein can be applied to other embodiments and can be practiced or carried out in various other related ways. Further, it should be understood that the wording and terminology used herein are for the purpose of description and should not be considered restrictive. The use of the terms "comprising", "including", or "having" and their synonyms herein is intended to cover the items listed thereafter and their equivalents as well as other items, while only the terms "consisting of" and "consisting only of" have a limiting meaning when interpreted.

Claims

1. A medical device, which is an artificial intelligence-based diagnostic medical device for diagnosing neurological diseases of patients, and the medical device is configured to be used in a pre-hospital environment. The medical device comprises: a memory; a camera configured to acquire image / video data of the patient; one or more sensors, including a microphone for acquiring natural language data of the patient; and a processor communicatively connected to the memory, the sensors, and the camera, wherein the processor is configured to: execute computer instructions to obtain patient data inputs, wherein the patient data inputs include one or more of demographic data, symptoms of the patient, elements of medical history, and physical examination results; recognition results obtained from the image / video data and / or the natural language data through at least one of visual recognition, computer vision, and natural language recognition; execute at least one primary calculation algorithm and / or analysis process to evaluate the patient data inputs and determine a primary diagnosis based on the patient input data; in response to the primary diagnosis indicating that further diagnosis is necessary, execute at least one secondary calculation algorithm and / or analysis process to evaluate the patient data inputs and determine one or more additional diagnoses based on the patient data inputs; and in response to the primary diagnosis and the one or more additional diagnoses, determine the final diagnosis and / or treatment plan of the patient.

2. The medical device according to claim 1, wherein the at least one primary calculation algorithm and / or analysis process includes at least one first calculation algorithm and / or analysis process configured to classify the patient to determine whether the patient has a neurological emergency based on the patient data inputs, and whether the determined neurological emergency is a stroke.

3. The medical device according to claim 1 or 2, wherein the at least one secondary calculation algorithm and / or analysis process includes at least one of the following: at least one algorithm and / or analysis process configured to determine a stroke subtype based on the patient data inputs; at least one algorithm and / or analysis process configured to determine whether the patient has a large artery occlusion based on the patient data inputs; at least one algorithm and / or analysis process configured to determine whether the patient has a transient ischemic attack based on the patient data inputs.

4. The medical device according to any one of claims 1-3, wherein the at least one primary calculation algorithm and / or analysis process and the at least one secondary calculation algorithm and / or analysis process are employed in a serial manner to improve their respective diagnoses, thereby determining the final diagnosis and / or treatment plan.

5. The medical device according to any one of claims 1-4, wherein the primary calculation algorithm and the secondary calculation algorithm are trained on at least one patient database based on the patient data inputs to generate their respective determination results.

6. The medical device according to any one of claims 1-5, wherein the primary analysis process and the secondary analysis process evaluate whether the classical syndrome definition is met or not based on the patient data inputs.

7. The medical device according to claim 6, Wherein, using the satisfaction of the classical syndrome definition to confirm the determination result of the calculation algorithm, using the non - satisfaction of the classical syndrome definition to invalidate the determination result of the calculation algorithm, or making the determination result of the calculation algorithm an indeterminate diagnosis.

8. The medical device according to any one of claims 1 - 7, wherein, the at least one auxiliary calculation algorithm and / or analysis process includes: a calculation algorithm and / or analysis process configured to determine a treatable stroke - like condition in response to determining a non - stroke condition by the at least one main calculation algorithm and / or analysis process.

9. The medical device according to any one of claims 1 - 8, wherein, performing the main and auxiliary calculation algorithms includes determining a diagnostic score, wherein: when the diagnostic score is higher than a first value, it indicates the presence of a neurological disorder; when the diagnostic score is lower than a second value, it indicates the absence of a neurological disorder; when the diagnostic score is between the first value and the second value, it indicates an indeterminate presence of a neurological disorder.

10. The medical device according to any one of claims 1 - 9, further comprising a strap to allow the patient to wear the medical device.

11. The medical device according to any one of claims 1 - 10, wherein, the patient data input is automatically obtained by the medical device and is received from the patient or a third party or both the patient and the third party.

12. The medical device according to any one of claims 1 - 11, wherein, the medical device is configured to initiate an interaction with the patient through voice prompts to provide the patient data input.

13. The medical device according to any one of claims 1 - 12, wherein, the final diagnosis includes acute stroke, transient ischemic attack, seizure, demyelinating disease, multiple sclerosis, brain trauma, or brain tumor.

14. The medical device according to any one of claims 1 - 13, wherein, the at least one main calculation algorithm and / or analysis process is configured to: automatically evaluate whether the patient has presented one or more initial signs of the neurological disorder, and automatically trigger a more comprehensive evaluation by the at least one auxiliary calculation algorithm and / or analysis process when the initial signs are detected.

15. The medical device according to claim 14, wherein, the at least one main calculation algorithm and / or analysis process is further configured to: evaluate a change in one of gait, speech, vision, movement of a body part, or a combination thereof as an initial physical examination result indicating the neurological disorder.

16. The medical device according to any one of claims 1 - 15, wherein, the treatment plan includes appropriate medications for the patient to take.

17. The medical device according to claim 8, wherein, the at least one auxiliary calculation algorithm and / or analysis process is further configured to: provide an alert for the other diagnosis in response to determining a treatable stroke - like condition when the probability of the other diagnosis is higher than one of 25%, 50%, 75%, and 90%.

18. The medical device according to any one of claims 1 - 17, wherein, The main calculation algorithm and the auxiliary calculation algorithm each include one or more of the following: artificial neural network, support vector machine SVM, Nu-SVM, linear SVM, naive Bayes NB algorithm, Gaussian NB, polynomial NB calculation algorithm, multi-class SVM, directed acyclic graph SVM, structured SVM, least squares support vector machine LS-SVM, Bayesian SVM, transductive support vector machine, support vector clustering SVC, C-SVM classification, nu-SVM classification, ε-SVM regression, and nu-SVM regression.

19. The medical device according to any one of claims 1-18, wherein, the processor is further configured to send the final diagnosis to a medical institution via a wired or wireless network.

20. The medical device according to any one of claims 1-19, wherein, the medical device is configured to be set in an ambulance.

21. The medical device according to any one of claims 1-20, wherein, the natural language data includes audio data obtained from the microphone, and the processor is further configured to evaluate the audio data related to the neurological disease.

22. The medical device according to any one of claims 1 to 21, wherein, in response to the at least one main calculation algorithm and / or analysis process determining a preliminary non-stroke diagnosis, the at least one auxiliary calculation algorithm and / or analysis process is further configured to evaluate the patient data input to evaluate at least one of the following: limb and gait ataxia, numbness, Horner syndrome, dysphagia, dysarthria, nausea, vomiting, vertigo, nystagmus, limping, slurred speech, weakness, partial facial droop, pain, tachypnea, flushing, sweating, facial expression, chest clutching, respiratory rate, speech interruption, shortness of breath, dyspnea, tachypnea, or other syndrome elements recorded in medical literature.

Images