Abnormal patient rehabilitation detection

Abstract

A system includes: a memory storing instructions; and a processor that executes the instructions. The instructions, when executed by the processor, cause the system to: repeatedly obtain a plurality of types of data derived from a rehabilitated patient; applying the plurality of types of data to the trained model; repeatedly generating rehabilitation scores, wherein the rehabilitation scores predict the rehabilitation condition of the rehabilitation patient; and outputting the rehabilitation score.

Description

Background Art

[0001] After some medical procedures (e.g., surgery), events (e.g., cardiac arrest), or illnesses (e.g., Covid-19), patients may recover through different pathways. Recovery is a complex process, and there is no universally accepted definition. Recovery can be defined as the time when a patient's condition has decreased to or exceeded their pre-illness health state, and therefore, recovery is a period when a patient's health status changes, and most of the time improves. Recovery can also be defined as the patient's ability to return to activities of daily living. A common metric used by hospitals to describe recovery is length of stay (LOS), but LOS only reflects a single point in time, not a path toward discharge. Furthermore, discharge decisions are often highly subjective and influenced by factors outside of the patient's health, such as the ability to go home and the logistics of the hospital. The end of the recovery period can be defined as the time when the patient's health level is the same as or even better than before the surgery / illness, or when the patient's health reaches a stable state / no longer improves. The end of the recovery period can also be defined as when the patient's health status reaches a sufficient level, and there is enough confidence in further recovery, so that the patient can be discharged.

[0002] Recovery pathways may vary, such as being fast or slow, and / or progressing in a regular or fluctuating manner. Recovery pathways may vary based on comorbidities, demographic characteristics such as age, and compliance with the treatments administered. Understanding the patient's recovery can help clinicians assess the patient's stability, readiness for discharge, and deterioration. To measure recovery over time, functional assessments or questionnaires can be used. However, functional assessments are time-consuming and may only cover a short period of time, while questionnaires provide a subjective assessment of recovery at discrete points in time. To support clinicians in making recovery-based decisions, such as assessing a patient's stability, readiness for discharge, and deterioration, early warning score (EWS) systems have been developed. Early warning systems utilize multiple parameters measured by spot checks during nursing rounds.

[0003] Functional assessments, questionnaires, and early warning scores have limitations, including lack of standardization, subjectivity, small amounts of data at discrete time points, and high workload. Reliance solely on spot checks results in a lack of relevant data for times outside of the spot check. Furthermore, spot checks are highly dependent on nurses’ time and are therefore time-consuming and labor-intensive. Furthermore, current high-level, universal discharge processes may result in the development of uniform protocols that direct each patient to be discharged along a similar, predefined route. Feedback is rarely incorporated to control pathways, and this may result in patients staying in the hospital longer than necessary, but may also result in delayed detection of abnormalities in the patient’s recovery process. Summary of the invention

[0004] According to one aspect of the present disclosure, a system includes: a memory storing instructions; and a processor executing the instructions. When the instructions are executed by the processor, the system: repeatedly obtains multiple types of data obtained from a rehabilitation patient; applies the multiple types of data to a trained model; repeatedly generates a rehabilitation score, the rehabilitation score predicts the rehabilitation status of the rehabilitation patient; and outputs the rehabilitation score.

[0005] According to another aspect of the present disclosure, a central system includes: a memory storing instructions; and a processor executing the instructions. The method performed by the central system includes: repeatedly obtaining multiple types of data obtained from a rehabilitation patient; applying the multiple types of data to a trained model by the processor executing the instructions from the memory; repeatedly generating a rehabilitation score that predicts the rehabilitation status of the rehabilitation patient; and outputting the rehabilitation score.

[0006] According to another aspect of the present disclosure, a tangible non-transitory computer-readable storage medium stores a computer program. When the computer program is executed by a processor, the system: repeatedly obtains multiple types of data obtained from a rehabilitation patient; applies the multiple types of data to a trained model; repeatedly generates a rehabilitation score that predicts the rehabilitation status of the rehabilitation patient; and outputs the rehabilitation score. BRIEF DESCRIPTION OF THE DRAWINGS

[0007] When read in conjunction with the accompanying drawings, the exemplary embodiments can be best understood from the following detailed description. It should be emphasized that the various features are not necessarily drawn to scale. In fact, for the clarity of discussion, the size can be increased or reduced at will. Where applicable and practical, the same reference numerals refer to the same elements.

[0008] Figure 1 A system for abnormal patient recovery detection according to a representative embodiment is illustrated.

[0009] Figure 2 Another system for abnormal patient recovery detection according to a representative embodiment is illustrated.

[0010] Figure 3 A method for abnormal patient recovery detection according to a representative embodiment is illustrated.

[0011] Figure 4 Another method for abnormal patient recovery detection according to a representative embodiment is illustrated.

[0012] Figure 5A , Figure 5B , Figure 5C , Figure 5D , Figure 5E and Fig. 5F A rehabilitation profile for a rehabilitation patient is illustrated according to a representative embodiment.

[0013] Figure 6 Correlations between recovery scores and features by day are illustrated according to a representative embodiment.

[0014] Fig. 7A The relationship of predicted recovery to expected recovery is illustrated according to a representative embodiment.

[0015] Figure 7B Graphs illustrate recovery scores over time according to representative embodiments.

[0016] Fig. 8A , Figure 8B and Figure 8C An individual prediction profile compared to an average profile is illustrated according to a representative embodiment.

[0017] Fig. 9 Predicted mean recovery curves according to representative embodiments are illustrated.

[0018] Fig.10 A computer system on which an abnormal patient recovery detection method is implemented is illustrated according to another representative embodiment. DETAILED DESCRIPTION

[0019] In the detailed description below, for the purpose of explanation, and not limitation, illustrative embodiments that disclose specific details are set forth to provide a thorough understanding of the embodiments according to the teachings of the present invention. The description of known systems, equipment, materials, operating methods, and manufacturing methods can be omitted to avoid obscuring the description of representative embodiments. Nevertheless, systems, equipment, materials, and methods within the capabilities of those of ordinary skill in the art are within the scope of this teaching and can be used according to representative embodiments. It should be understood that the terms used herein are only for the purpose of describing specific embodiments, and are not intended to be limited. The defined terms are the meanings outside the scientific and technological meanings of the defined terms that are generally understood and accepted in the technical field of this teaching.

[0020] It should be understood that although the terms first, second, third, etc. may be used herein to describe various elements or components, these elements or components should not be limited by these terms. These terms are only used to distinguish one element or component from another element or component. Therefore, without departing from the teachings of the inventive concept, the first element or component discussed below may also be referred to as the second element or component.

[0021] The terms used herein are only for the purpose of describing specific embodiments and are not intended to be limiting. The singular forms of the terms "one", "an" and "said" as used in the specification and claims are intended to include both the singular and plural forms, unless the context clearly dictates otherwise. In addition, when used in this specification, the terms "include" and / or "comprises" and / or similar terms specify the presence of recorded features, elements and / or parts, but do not exclude the presence or increase of one or more other features, elements, parts and / or their groups. The term "and / or" as used herein includes any combination and all combinations of one or more items in the associated listed items.

[0022] Unless otherwise specified, when an element or component is said to be "connected to," "coupled to," or "adjacent to" another element or component, it will be understood that the element or component can be directly connected or coupled to another element or component, or there may be intervening elements or components. That is, these and similar terms include situations where one or more intervening elements or components may be employed to connect two elements or components. However, when an element or component is referred to as being "directly connected" to another element or component, this only includes situations where two elements or components are connected to each other without any intervening or intermediate elements or components.

[0023] The present disclosure, through one or more of its various aspects, embodiments and / or specific features or sub-components, is intended to bring one or more of the advantages specifically noted below. For purposes of explanation and not limitation, example embodiments disclosing specific details are set forth to provide a thorough understanding of embodiments according to the present teachings. However, other embodiments consistent with the present disclosure that depart from the specific details disclosed herein are still within the scope of the claims.

[0024] As described herein, a patient's recovery over time can be objectively tracked so that abnormal recovery can be identified to help clinicians make educated decisions based on recovery.

[0025] Figure 1 A system for abnormal patient recovery detection according to a representative embodiment is illustrated.

[0026] Figure 1 The system 100 in is a system for abnormal patient rehabilitation detection, including components that can be provided together or can be distributed. The system 100 includes a central computer 101, a storage system 102, a network 130, a display 181, a patient monitor 187, an EMR 188, an activity sensor 189, and an AI training system 195.

[0027] Central computer 101 may include a single computer or may alternatively represent a central system including multiple single computers. For example, central computer 101 may represent a system including one or more servers provided on behalf of an organization providing abnormal patient rehabilitation detection services. Central computer 101 may represent a cloud-based system that provides multiple services in a data center. Fig.10 The computer that can be used to implement the central computer 101 is shown in FIG. 1 , but the central computer 101 may include Figure 1 or Fig.10 More or fewer components than those shown.

[0028] The central computer 101 includes at least a controller 150, and the controller 150 includes at least a memory 151 storing instructions and a processor executing the instructions. In an embodiment where multiple computers are used to provide the functions attributed to the central computer 101 herein, each of the multiple computers may include a controller (e.g., controller 150), and in turn includes a memory (e.g., memory 151) and a processor (e.g., processor 152). The controller 150 executes instructions to implement some or all of the features attributed to the central computer 101 herein. The controller 150 may directly perform some or all of the operations described herein. For example, the controller 150 may directly control operations based on data from a rehabilitation patient via a patient monitor 187 and an activity sensor 189, such as logical operations performed by the processor 152 executing instructions from the memory 151. When the processor 152 executes instructions from the memory 151, the process implemented by the controller 150 may also include steps that the controller 150 does not directly execute.

[0029] The storage system 102 stores data of the central computer 101 and may be provided together with the central computer 101. For example, the storage system 102 may store a trained prediction model for generating a rehabilitation score that predicts the rehabilitation status of a rehabilitation patient, and a type of baseline that may be dynamically established to compare with the rehabilitation score. The central computer 101 may retrieve and apply the trained prediction model to generate a rehabilitation score, and may dynamically generate or select and retrieve a baseline to compare with each rehabilitation score.

[0030] The network 130 may include a broadband wide area network (WAN), such as the Internet, and may include wired and wireless communication devices by which data may be transmitted directly or indirectly between a patient and the central computer 101 .

[0031] The display 181 may be a monitor, such as a computer monitor, a display on a mobile device, an augmented reality display, a television, an electronic whiteboard, or another screen configured to display an electronic image. The display 181 represents a medical monitor that displays a rehabilitation score received from the central computer 101. The display 181 may be positioned in close proximity to a patient monitor 187 and / or an activity sensor 189 around the patient. In some embodiments, the display 181 may be integrated with the patient monitor 187 and / or the activity sensor 189 so that the rehabilitation score and alarm from the central computer 101 are displayed on the user interface of the display 181 together with the data types obtained from the patient monitor 187 and the activity sensor 189. The display 181 may also be integrated with a computer or other type of electronic device, such as in an office other than a ward in a hospital. The display 181 may be or include a user interface of a patient monitor (e.g., a patient monitor 187) so that the rehabilitation score from the central computer 101 and the measurement results of blood pressure, SpO2, ECG, etc. may be displayed.

[0032] Patient monitors 187 and activity sensors 189 represent electronic equipment for collecting various types of data from rehabilitation patients. The various types of data may include data derived from the vital signs of rehabilitation patients, and / or may include data derived from activity metrics of rehabilitation patients. The various types of data may include data specifically identified as most relevant to the training prediction model executed by the central computer 101, but other types of data may also be obtained and provided to the central computer 101. Examples of patient monitors 187 include SpO2 monitors, blood pressure monitors, heart rate monitors, respiratory monitors, and other types of medical monitors that monitor the physiological condition of patients. Examples of activity sensors 189 include wearable fitness monitors, such as watches that monitor the physiological condition of users, and other types of electronic devices, including smartphones with fitness applications installed. Wearable devices with PPG and acceleration sensors can be used to collect input data for the central computer 101 for patients. Data from wearable sensors can also be used, including ECG, capnography, EMG, barometers, temperature sensors, gyroscopes, skin galvanic response, etc. Activity sensors 189 can also represent sensors that are not specifically measuring activity, such as measurements when patients are resting or unconscious. Furthermore, patient monitor 187 may represent more than one patient monitor, and activity sensor 189 may represent more than one activity sensor or other type of sensor.

[0033] EMR 188 represents an electronic medical record system of a medical institution (e.g., a hospital or hospital system). EMR 188 may store patient records for patients monitored by patient monitors 187 and using activity sensors 189. Electronic medical records stored in EMR 188 may include notes and measurements made by medical personnel such as physicians, nurses, and technicians. Patient records may include vital signs and activity measurements captured by patient monitors 187 and activity sensors 189, and automatically provided over a local network for storage in EMR 188. Patient records may also include vital signs and activity measurements captured by medical personnel, entered into an electronic user interface, and then provided over a local network for storage in EMR 188. Data from EMR 188 may include data from spot check measurements, including blood pressure or biomarker measurements (e.g., data from blood samples, weight, cardiac output, or questionnaires), as well as other types of EMR data (e.g., medication data).

[0034] As described above, display 181, patient monitor 187, EMR 188, and activity sensor 189 may be co-located in a medical facility such as a hospital. However, in some embodiments, even after a patient is discharged from such a medical facility, the patient may still be monitored to detect abnormal patient recovery, so patient monitor 187 and / or activity sensor 189 may not necessarily be co-located with display 181 and EMR 188.

[0035] exist Figure 1 In an embodiment of the present invention, a medical institution may include multiple patients, each of which is equipped with multiple patient monitors (e.g., patient monitor 187) and multiple activity sensors (e.g., activity sensor 189). For example, multiple patients at a single medical institution may simultaneously obtain the services of central computer 101 via network 130. Each such patient may also provide vital signs and activity metrics and data derived from vital signs and activity metrics from EMR 188 and from instances of patient monitor 187 and activity sensor 189 to central computer 101 via network 130.

[0036] The AI ​​training system 195 represents a system for training artificial intelligence, such as a trained model for repeatedly generating patient recovery scores. The AI ​​training system 195 can also be used to develop other types of predictive models and algorithms, not limited to training only artificial intelligence. The AI ​​training system 195 can apply multiple types of data inputs to machine learning algorithms to identify which data inputs have the best correlation with the expected results, and which machine learning algorithms can be used to train predictive models with the best performance compared to other machine learning algorithms. Figure 4An example process of using a machine learning algorithm to develop a predictive model as a trained model is shown and described in . The AI ​​training system 195 can be provided by an entity that provides the central computer 101 (e.g., a healthcare provider), or can be independently provided as a third-party service to multiple different entities (e.g., multiple different healthcare providers). For example, the AI ​​training system 195 can include one or more computers, including one or more servers, to develop a trained model applied by the central computer 101 to generate a rehabilitation score that predicts the rehabilitation status of a rehabilitation patient.

[0037] As described above, the central computer 101 can represent a system including a memory for storing instructions and a processor for executing instructions. Such a system can repeatedly obtain multiple types of data obtained from a rehabilitation patient and apply the multiple types of data to a trained model each time the multiple types of data are obtained. For example, the central computer 101 can periodically apply the trained model to multiple types of data obtained from a patient, or at least repeatedly apply the model when sufficient data is available. By applying the trained model to multiple types of data, such a system can repeatedly generate a rehabilitation score that predicts the rehabilitation of a rehabilitation patient and output the rehabilitation score. Figure 1 In the embodiment, the system including the central computer 101 is centralized and outputs rehabilitation scores of a plurality of rehabilitation patients.

[0038] Figure 2 Another system for abnormal patient recovery detection according to a representative embodiment is illustrated.

[0039] Figure 1 The system 200 in the embodiment is a system for abnormal patient rehabilitation detection, including components that can be provided together or can be distributed. The system 200 includes a central computer 201, a storage system 202, a network 230, a facility 270, a display 271, a facility 280, a display 281 and an AI training system 295. Figure 2 Each of the facilities 270 and 280 in may correspond to Figure 1 An example of a facility that includes a display 181, an EMR 188, a patient monitor 187, and an activity sensor 189. Figure 2 The display 271 and the display 281 in the embodiment may correspond to Figure 1 An example of display 181 in FIG. Figure 2 The AI ​​training system 295 in may correspond to Figure 1 AI training system in 195.

[0040] exist Figure 2In an embodiment of the present invention, multiple medical institutions may each include multiple patients, each of which is provided with multiple patient monitors and multiple activity sensors. For example, multiple patients at multiple different medical institutions may simultaneously obtain the services of central computer 201 through network 230. Additionally, each patient in facility 270 and facility 280 may send multiple types of data to central computer 201, and central computer 201 may apply multiple types of data to the trained model each time the multiple types of data are obtained.

[0041] As described above, the central computer 201 can represent a system including a memory for storing instructions and a processor for executing instructions. Such a system can repeatedly obtain multiple types of data obtained from rehabilitation patients and apply the multiple types of data to the trained model each time the multiple types of data are obtained. For example, the central computer 201 can periodically apply the trained model to the multiple types of data obtained from the patient, or at least repeatedly apply the model when sufficient data is available. By applying the trained model to multiple types of data, such a system can repeatedly generate a rehabilitation score that predicts the rehabilitation of the th rehabilitation patient, and output the rehabilitation score. Figure 2 In the embodiment, the system including the central computer 201 is centralized and outputs rehabilitation scores of a plurality of rehabilitation patients. Figure 2 In the embodiment, the system including the central computer 201 also generates and outputs rehabilitation scores of rehabilitation patients of multiple medical institutions.

[0042] Using system 100 and system 200, the data and metadata of a patient can be compared with the data and metadata of other patients with similar conditions and other conditions from the same hospital and other hospitals. Anomalies in patients or patient groups can be detected so that abnormal patient recovery can be detected individually, and abnormal patient recovery trends can be detected for patient groups (e.g., in a specific hospital or other care facility). The data provided to central computer 101 and central computer 201 may include health status information of patients over time. The collected health status information may include vital signs such as heart rate, respiratory rate, blood pressure, etc. The collected health status information may also include activity metrics such as number of steps, activity type, posture, walking speed and distance, activity level, activity count, etc. The data provided to central computer 101 and central computer 201 may also include demographic data such as age, height, weight, and gender, as well as medical record information such as disease classification, surgery type, surgery date, medication, and laboratory tests. Data collection can be performed through wearable devices, patient monitors, EMR systems, spot checks, etc. The collected data combination can be converted into a trend set for each patient. Each patient's trend set can be compared to a reference trend set of similar patients, such as patients undergoing the same surgery, suffering from the same disease, and of the same age. Central computer 101 and central computer 201 can detect abnormal recovery and can guide healthcare professionals in making recovery-based decisions.

[0043] Figure 3 A method for abnormal patient recovery detection according to a representative embodiment is illustrated.

[0044] Figure 3 The method can be Figure 1 Central computer 101 or Figure 2 Executed by the central computer 201 in. Figure 3 The method starts from S305, obtaining the data type of the needle-in-the-haystack self-rehabilitation patient. Multiple types of data can be collected continuously or semi-continuously in an automated manner to be used in Figure 1 Central computer 101 or Figure 2 The trained model implemented at the central computer 201 in is processed.

[0045] For example, the data type obtained from a recovering patient may include measurements of the patient's recovery from a treatment such as a cardiac intervention or an event such as a myocardial infarction. The data type may include data derived from measurements of the patient's vital signs, and / or may include data derived from activity metrics of the patient. Other forms of data may also be relevant to the trained models described herein. For example, the data type obtained at S305 may also include metrics such as biomarkers, demographic information, answers to questionnaires (e.g., questions such as pain levels), and the like.

[0046] At S310, the data obtained at S305 is applied to the trained model. The trained model can be a prediction model trained by the AI ​​training system 195 or the AI ​​training system 295. For example, the trained model can include a model based on extreme gradient enhancement. Other types of trained models can also be used as the trained models implemented by the central computer 101 or the central computer 201. For example, Lasso linear regression, k nearest neighbor, tree-based methods (such as decision trees or random forests), and support vector regression can be used to develop a trained model implemented by the central computer 101 or the central computer 201. The trained model can apply a data set to objectively determine whether the patient is recovering. As explained elsewhere in this article, the trained model can use a subset of data derived from vital sign measurements and activity metrics, because the trained model can be trained to identify a specific subset of data that is most relevant to determining an objective health state.

[0047] At S315, a rehabilitation score is generated for the patient based on applying the trained model to the data at S310. For example, the rehabilitation score may contain a value between 1.0 and 2.0, with 1.8 being a threshold for whether the patient is ready to be discharged. Rehabilitation can be pre-defined so that the rehabilitation score can be an objective measure relative to the definition. For example, the end of the rehabilitation period can be defined as when the patient's health level is the same or even better than before the surgery / illness, or when the patient's health reaches a stable state and no longer improves. The end of the rehabilitation period can also be defined as when the patient's health status reaches a sufficient level and there is enough confidence in further rehabilitation so that the patient can be discharged.

[0048] As an alternative to predicting the rehabilitation score in S315, discharge readiness can be predicted. Discharge readiness can be defined, for example, using a rehabilitation score of 1 or below as 0% discharge readiness and a rehabilitation score of 1.8 and above as 100% discharge readiness. In addition, similar predictions can be used to predict discharge date, such as by extrapolating the rehabilitation score to 1.8, or by comparing to different length of stay groups, such as those most likely to fit into the 5-7 day length of stay group.

[0049] The process from S305 to S315 is repeated. For example, each time an appropriate set of data is obtained in S305, the set of data may be applied to the trained model at S310 to generate a rehabilitation score at S315. The process from S305 to S315 may be performed periodically rather than continuously, such as every 5 minutes, every hour, every 6 hours, every 12 hours, or every 24 hours. In addition, the process from S305 to S315 may be performed intermittently, for example, once an appropriate data set is obtained when an appropriate data set is not otherwise continuously obtained at S305.

[0050] At S320, a baseline is established for the rehabilitation score generated at S315. The baseline is not necessarily fixed, but can be selected from a set of baseline types stored in the memory system 102 or the memory system 202. A set of baseline categories can be generated, and the set of baseline categories can vary based on any number of criteria, including but not limited to the type of treatment or event, the patient's condition before the treatment or event, the rehabilitation scores of previous similar patients who have recovered from the same or similar treatment or event, the patient's demographic characteristics (e.g., age and / or weight), and the time since the treatment or event. For example, the baseline used for comparative assessment of whether a patient has successfully recovered may increase over time, which is fundamentally based on the expectation that the patient will get better over time if he or she recovers. Therefore, establishing a baseline at S320 may include dynamically generating a baseline based on a set of variable values ​​input to an equation, or may include selecting a baseline class from a predetermined set of baseline classes.

[0051] In S325, the recovery score generated in S315 is compared to the baseline established in S320. For example, the recovery score and the comparable values ​​of the baseline curve can be directly compared. As long as the baseline curve increases over time, the comparable values ​​of the baseline curve can reflect the amount of time that has passed since the treatment or event.

[0052] At S330, it is determined whether the rehabilitation score generated at S315 is higher than the baseline established at S320. The judgment at S330 can be a simple "yes" or "no", or can include a relative value reflecting the distance of the rehabilitation score from the baseline, particularly when the rehabilitation score is lower than the baseline. For example, the baseline may be a class or group of values. The judgment at S330 can also be whether the rehabilitation score is within the baseline range, or is a member of a baseline family, class or set, without specifically requiring it to be higher than the baseline. The comparison can also reflect the tolerable variance in the time series of the measured values, so that a tolerance can be provided within the definition or definable range of the rehabilitation score, so that the rehabilitation score may be greater than or less than the actual threshold.

[0053] If the rehabilitation score is not above the baseline (S330=No), an alarm is generated at S335. The nature of the alarm may vary depending on the relative distance of the rehabilitation score from the baseline. For example, a rehabilitation score that is far below the baseline may result in a notification to the attending physician, while a rehabilitation score that is close to the baseline but still below the baseline may result in a notification to the display 181 in the patient's room in the hospital. The alarm may be displayed continuously on the display 181, for example until the rehabilitation score is above the baseline in a subsequent comparison or until the alarm is turned off by an authorized user. In some embodiments, a patient whose recovery is better than expected may also result in an alarm, for example to remind a physician or nurse to check on the rehabilitation patient to ensure that the rehabilitation patient is recovering better than expected, or to enable a discharge plan to be made in a timely manner.

[0054] If the recovery score is above the baseline (S330=Yes), then a trend of the recovery score is determined at S340. The trend may include an overview that includes the recovery score generated at S315 and previous recovery scores generated for the same patient. For example, if the trend is not improving or is particularly worsening, it may indicate that there is a problem. The trend may include recovery scores for two or more days. In addition, the trend may include a comparison with a relevant baseline established at S320, such as when the same baseline is used for the patient throughout the patient's recovery process.

[0055] In some embodiments, a trend in the rehabilitation score may be determined at S340 even if the rehabilitation score is not above the baseline. For example, the trend in the rehabilitation score may be investigated at all times, regardless of whether the rehabilitation score is above the baseline at any particular point in time. The trend in the rehabilitation score may be investigated through high rehabilitation scores and low rehabilitation scores. A downward trend or a trend that is plateauing prematurely may be identified and an alert may be issued in the manner described herein.

[0056] At S345, a determination is made as to whether the trend is alarming. For example, if the trend is worsening or not improving at all, it may be considered alarming. The determination at S345 may reflect consideration of only the last two recovery scores, or may reflect consideration of more than two recent recovery scores, such as when the recovery score has failed to improve for two or more days.

[0057] If the trend causes an alarm at S345 (S345=Yes), an alarm is generated at S350. The trend determined at S340 is, for example, a trend of rehabilitation scores generated each day. However, in other embodiments, estimates of future rehabilitation score values ​​may also or alternatively be determined so that future trends themselves can be predicted. The trend prediction may be based on data collected in the previous few days and rehabilitation trends of other patients. The predicted trend may also be associated with a confidence interval. For example, the upper limit of the confidence interval may be calculated using the trend of patients who recover quickly, while the lower limit of the confidence interval may be calculated using the trend of patients who recover slowly.

[0058] If the trend does not raise an alarm at S345 (S345=No), or after an alarm is generated at S335 or after an alarm is generated at S350, the recovery score and any alarms are output to the patient monitor at S355. The associated baseline may also be output at S355 so that the correlation of the recovery score with the baseline can be visually checked. For example, the recovery score, alarms, and baselines may be output from the central computer 101 to the patient monitor. Figure 1 or output from the central computer 101 to the display 181 in Figure 2 In addition, the rehabilitation score and / or alarm is not necessarily output to the medical monitor, or is output only to the medical monitor. For example, the rehabilitation score and / or alarm can be output to a mobile device (e.g., a (hospital-issued) smartphone), a central monitoring station (e.g., a hospital nurse station), or other types of networked communication devices.

[0059] The output at S355 can be to an output device such as a patient monitor in a medical institution. The output at S355 can also or alternatively be to a central monitor, to an application on a mobile device, directly to a mobile device (e.g., via a text message), or in another electronic transmission form.

[0060] The output of S355 may result in a visualization of the profiles of individual patients superimposed in the short hospital stay group or the long hospital stay group, and this may help the healthcare provider identify patients with abnormal recovery. However, the healthcare provider may not have the time to analyze such charts for all patients every day. Therefore, an alert may be provided as an output at S355 based on a threshold or trend. For example, the group profile boundary may be set so that an alert is sent to the healthcare provider when the recovery score exceeds the group profile boundary. In the example use case, if the patient exceeds the lower limit of the confidence interval for the short hospital stay group, a flag may be output to indicate that the patient is likely to be hospitalized for a longer time. In general, if the patient deviates from the range of a certain group, such deviation may be notified to the caregiver. If the patient's hospital stay also exceeds the lower limit of the long-term hospital stay confidence interval, a warning may be issued to ask the patient to carefully review the recovery score and model feature values ​​(activity, vital signs, etc.) because the recovery speed is abnormally slow. In addition, the confidence of such a mark can also be reported, such as when the mark may be affected by data completeness / incompleteness and quality. For example, when data is missing, less than all features may be calculated, which may affect model accuracy. Additionally, the number of days a patient is below or above a certain threshold and the distance from that threshold can also be provided as output to the healthcare provider.

[0061] Another metric that can be observed and output at S355 is the change in the patient's recovery curve. For example, a reference curve can be determined as a smoothed version of the average curve in the patient group band. The deviation of the actual patient's data from this reference curve can then be calculated, and the absolute values ​​of all values ​​can be added and divided by the number of points included. Alternatively, a smooth curve can be fit to the patient's data, and then the deviation from the fit can be calculated. The deviation may be the sum of the absolute differences or the sum of the squared differences, both divided by the number of points, and the square root of the sum of the squares is calculated. Other methods of obtaining variance measures are known in the art and can be used.

[0062] In addition to thresholds, trends in predicted recovery scores can also be used to decide whether to alert healthcare providers. For example, multiple consecutive days / predicted declines in condition may warrant an alert even if the patient's condition has not yet fallen below a predefined threshold. Furthermore, instead of differentiating between short and long hospital stays, one could differentiate between those with complications and those without, or those with long-term disability and those without, or different levels of long-term independence, etc. It is not necessary to differentiate between two groups either. For example, a distinction can be made between short, medium, and long stays. Furthermore, different subgroups can be created. For example, a 90-year-old patient undergoing a specific type of surgery may recover more slowly than a 50-year-old patient undergoing the same type of surgery, so patients with this type of surgery can be compared to the overall group as well as to age-matched subgroups. Differentiation can also be used to provide contextualized alerts to healthcare providers. For example, an alert to a healthcare provider might state "Patient X has a slower recovery (expected length of stay of 11-13 days) compared to the average after surgery Y, but is within the expected range for his age group (80+), with an average length of stay of 13 days."

[0063] Compared with functional assessments, questionnaires, and / or early warning scores alone, Figure 3 The results of the method can provide a reliable and objective representation of the patient's recovery with a relatively low workload. Figure 3 The method is not limited to postoperative rehabilitation. For example, it can also be used with Figure 3 Similar or identical approaches to the ones shown are used to predict recovery after illnesses such as Covid-19 to help identify patients who are slow to recover early, even if they recover outside of a hospital, or after events such as a stroke or cardiac arrest.

[0064] Figure 4 A method for abnormal patient recovery detection according to a representative embodiment is illustrated.

[0065] Figure 4The method begins at S455 by applying data input to a machine learning algorithm. The data input may include physiological characteristic measurements obtained from the patient, and data derived from these measurements. The data input may also include activity measurements obtained from the patient, and data derived from such activity measurements. The machine learning algorithm may be one of a plurality of machine learning algorithms being tested to see which performs best in developing an optimal trained model for abnormal patient rehabilitation detection as described herein.

[0066] At S460, the prediction model is trained based on applying the data input to the machine learning algorithm at S455. The prediction model can be trained by repeatedly inputting multiple types of data to optimize the decisions of the prediction model nodes using real-life data. The real-life data used in the training at S460 can include real-life rehabilitation scores, but in some embodiments, alternative real-life data can include a proxy for rehabilitation scores, such as a functional assessment or questionnaire score.

[0067] If the real rehabilitation scores of each patient at different time points are known, these real rehabilitation scores can be used to build and train models using machine learning techniques to predict rehabilitation scores based on input parameters such as vital signs and activity metrics. In order to obtain a reference rehabilitation score to train the model, a continuous rehabilitation reference profile was designed for the first set of embodiments. The continuous rehabilitation reference profile is designed based on events such as discharge, complications, and readmissions, as well as literature and intuition. The functional rehabilitation profile shows an exponential trend. That is, there will be a great improvement in the first few days, after which the rehabilitation will gradually improve slowly until it reaches a plateau. Patients with different levels of rehabilitation will reach the plateau in different days. In addition, all patients may receive the same starting rehabilitation score on the day of surgery, except for those who experience complications in the first few days, because these patients may drop below the baseline level after surgery.

[0068] At S465, it is determined whether the current machine learning algorithm is the last machine learning algorithm. For example, when AI training system 195 or AI training system 295 is training a trained model used by central computer 101 or central computer 201, the training may involve training multiple different types of machine learning algorithms of different types to determine which algorithm provides the best results.

[0069] If the current machine learning algorithm is not the last machine learning algorithm (S465=No), the process returns to S455 to begin applying another (next) machine learning algorithm to the data input.

[0070] If the current machine learning algorithm is the last machine learning algorithm (S465=Yes), then at S470, Figure 4The method includes identifying the prediction model from the machine learning algorithms with the best performance. In order to find the prediction model from the best performing machine learning algorithms, the machine learning algorithms can be compared based on the prediction performance and interpretability. Regarding the prediction performance, the mean absolute percentage error can be used to judge according to how well the machine learning algorithm fits the reference recovery curve. Next, the Spearman rank correlation coefficient can be calculated to ensure that the trend of the predicted recovery curve is correct. The Spearman rank correlation can reflect whether there is an upward trend and whether the recovery rate decreases when complications occur. Finally, the error at discharge can be investigated. The error at discharge may reflect how well the trained model predicts the same recovery score on the day of discharge. In terms of the level of interpretability, models with greater interpretability may be preferred because this may lead to higher clinical adoption rates. Although the interpretability of the XGBoost model may not be as good as other algorithms, the use of only six features increases the ability of the XGBoost model to interpret the results.

[0071] Rehabilitation testing can be framed as a classification problem, e.g., predicting whether a patient belongs to a long-stay or short-stay group. If rehabilitation testing is framed as a classification problem, the predictive ability can be expressed using metrics such as precision and recall. Alternatively, sensitivity and specificity can be used instead. These pairs of metrics can be summarized as scalar metrics such as F-scores and area under the curve (AUC). In addition, when comparing designs, metrics such as correlation and kappa scores are often used. Methods such as Bland-Altman plots can be used alternatively or additionally.

[0072] At S475, the identified prediction model is developed into a trained model. The trained model can be provided to the central computer 101 or the central computer 201 as Figure 3 The trained model in the method.

[0073] based on Figure 4 Embodiments of model development may also be directed toward detecting alternatives to abnormal recovery, such as predicting length of hospital stay.

[0074] Although it can be Figure 4The trained model obtained by the method is designed to have a per-patient recovery score, but other methods can also be used to obtain reference scores for the trained model, such as asking healthcare providers to provide recovery scores for the training set or by taking a recovery proxy. Data from healthcare providers may be provided by multiple raters to overcome the limitation of subjectivity. Proxy data may be taken from a combination of rehabilitation proxies, such as functional assessments or questionnaire scores as the real situation, although these have limitations. In addition, it may be challenging to obtain a sufficient number of good reference scores that are representative and cover the target patient population. Recovery trajectories may vary greatly for different conditions and interventions. For example, in the case of planned surgery, data can be collected before surgery to provide a baseline for each patient, which can also be used as an input for the recovery score. In addition, the option of selecting patients to calculate different recovery trend groups can also be implemented. The selection can be done automatically based on the collected data or manually based on the input of healthcare providers. Automatic selection can be based on, for example, data quality, amount of data per patient, and patient characteristics, such as age, gender, BMI, etc. Manual selection of data may include asking healthcare providers to only include patients with specific characteristics.

[0075] Reference below Figure 5A , Figure 5B , Figure 5C , Figure 5D , Figure 5E , Fig. 5F , Figure 6 , Fig. 7A , Figure 7B , Fig. 8A , Figure 8B and Figure 8C A first set of example embodiments is described.

[0076] To illustrate the first set of example embodiments, health status parameters can be collected from a group of tumor patients undergoing abdominal surgery to remove tumors. Health status parameters can be continuously collected in the hospital and at home from the day of surgery until three weeks after surgery. Health status parameters can include vital sign parameters from accelerometer, electrocardiogram (ECG) and photoplethysmography (PPG) sensor data, such as heart rate, respiratory rate, heart rate variability index, etc. Health status parameters can also include activity parameters derived from accelerometer sensor data, such as number of steps, activity level, walking speed and duration, posture, etc. Combinations of vital sign parameters and activity parameters can also be continuously collected. Examples of such combinations include rehabilitation of heart rate after activity, resting heart rate, heart rate while walking, and the like.

[0077] Figure 5A , Figure 5B , Figure 5C , Figure 5D , Figure 5E and Fig. 5F A rehabilitation profile for a rehabilitation patient is illustrated according to a representative embodiment.

[0078] Figure 5A , Figure 5B , Figure 5C , Figure 5D , Figure 5E and Fig. 5F The design of the rehabilitation profile in six patients undergoing oncological surgery shows the impact of discharge date, complications of different severity expressed as Clavien-Dindo (CD) score, and readmission. Figure 5A , Figure 5B , Figure 5C , Figure 5D , Figure 5E and Fig. 5F It can be seen that the final recovery curve is smooth, interrupted only by complications or readmissions, while the actual curve may fluctuate more. Figure 4 This is not a problem for the model in .

[0079] The recovery curve of the first set of embodiments is designed to increase from 1.0 to 2.0. On the day of discharge, the score is 1.8, and assuming a recovery rate of 80%, the patient is expected to recover further at home. Using the patient's discharge date, the reference recovery status of patients with short and long hospital stays is different, such as Figure 5A and Figure 5B As shown. If a complication occurs, the recovery score decreases, and because there may be lead time, it starts to decrease one or two days before the actual complication occurs. The magnitude of the decrease depends on the severity of the complication, which is expressed on a scale of 1 to 5 according to the Clavien-Dindo classification, where 5 means the patient has died. If a complication occurs and the Clavien-Dindo classification is 1 or 2, there will be a small decrease. If a complication occurs and the Clavien-Dindo grade is higher, the decrease will be greater. For example, a Clavien-Dindo grade 4 complication that requires the patient to be admitted to the ICU will result in a large decrease in recovery. Readmissions are designed in the same way as complications according to the Clavien-Dindo classification. However, there are differences in the recovery scores at discharge. The discharge score for the first discharge is changed to 1.6 because the patient may not be ready to be discharged since the readmission. Therefore, the discharge recovery score after the readmission may be 1.8. Fig. 5F An example of a readmission with a Clavien-Dindo score of 3 is shown.

[0080] Machine learning methods can be used to predict recovery scores in patients undergoing abdominal surgery for tumors. The goal may be to develop trained models that take observations and time series as input and output recovery scores or expectations. The inputs to the machine learning algorithm during the training phase may include features of the training set and an engineering reference profile. Figure 5A , Figure 5B , Figure 5C , Figure 5D , Figure 5E and Fig. 5F It can be seen that the engineering reference is a continuous profile that can be used to predict rehabilitation scores at different time scales, but in the first set of embodiments is used for rehabilitation scores by day.

[0081] Thus, for each measured parameter, a value per day may be calculated as an input feature. The measured parameters may include total steps per day, average heart rate, total energy expenditure, upright time, etc. These types of data may include the type of data repeatedly obtained by central computer 101 and central computer 201 and applied to the trained model.

[0082] Figure 6 Correlations between reference recovery scores and day-by-day characteristics are illustrated according to a representative embodiment.

[0083] exist Figure 6 , the Pearson correlation coefficient between the recovery score and the features created by day is shown for the oncology training set. Vital signs are shown as the first group, activity measures are shown as the second group, and combined features are shown as the third group. The boundaries used for feature selection can also be displayed. Figure 6 The Pearson correlation coefficient of the features to the engineering reference score is shown, which gives an indication of which parameters may be linked to rehabilitation in the tumor population. In the first set of embodiments, activity metrics have high correlation scores, but there are also high correlation scores for the circadian rhythm of heart rate, some heart rate variability parameters in the frequency domain, and certain vital sign / activity combination metrics (such as heart rate rehabilitation). In addition to these features calculated each day, the change in feature values ​​can also be used to add historical records and trends of features to the trained model. The change in these feature values ​​can be calculated by subtracting the feature values ​​of the previous day from the values ​​of the current day. Since the change in feature values ​​on the first day cannot be calculated, no rehabilitation predictions will be made for that day.

[0084] You can use Figure 4 Various machine learning algorithms shown in the figure are combined with various feature selection methods and data interpolation methods to build prediction models. Figure 4 The method tests different supervised machine learning algorithms, including linear regression using Lasso, k-nearest neighbors, tree-based methods (such as decision trees, random forests, and extreme gradient boosting), and support vector regression.

[0085] For the first set of embodiments, an extreme gradient boosting method (XGBoost) can be selected to provide the best performance in the training phase. The extreme gradient boosting method can also be used to train a prediction model to predict the score of the test set. However, there are many different machine learning methods that may also successfully predict the rehabilitation score. In addition, other methods can be used in addition to machine learning. For example, a rule-based solution or an analytical expression can be used. The analytical expression can be or can include a function that has input values ​​as parameters, processes the input values ​​according to certain formulas, and returns a predicted score.

[0086] Alternatively, rather than predicting a score, the features themselves can be used directly to cluster patients. For example, one can plot trends in heart rate rehabilitation over time or steps by day, etc., and identify abnormal patterns by comparing to the profiles of a subgroup of patients. While a single parameter may not be as accurate in detecting, for example, readiness for discharge, this approach can be used to identify specific parameters of abnormal patient recovery and targeted countermeasures can be taken. For example, if there is a problem with circadian rhythms, one can investigate whether sleep disturbances can be overcome, while low activity levels can be offset with physical therapy or activity goals. In addition, combinations of parameter profiles can also be used. For example, a specific heart rate rehabilitation may be within the expected confidence interval of normal recovery for a representative group of patients, but not necessarily for a subgroup with a similar resting heart rate profile to that patient, so a combination of parameters can still be used to flag patients to healthcare practitioners.

[0087] Fig. 7A The relationship of predicted recovery to expected recovery is illustrated according to a representative embodiment.

[0088] exist Fig. 7A , the relationship between predicted recovery and expected recovery is shown, where each point represents a day. The predicted recovery rate is the result of the trained model applied by the central computer 101 or the central computer 201. The expected recovery rate may be a baseline, median or mean value of recovery for a group or subgroup to which the patient is compared. In a setting such as Fig. 7A In the visualization, different patients can be displayed with different colors, different symbols, or different labels. The expected recovery rate reflects the engineering reference recovery rate score.

[0089] Figure 7B Graphs illustrate recovery scores over time according to representative embodiments.

[0090] exist Figure 7B , the predicted recovery scores versus the expected recovery scores over time for two different patients are shown. Figure 7B It is shown that for the training set, the trained model is able to predict the rehabilitation score well, even though the resulting curve is not as smooth as the engineering reference curve.

[0091] Fig. 8A , Figure 8B and Figure 8C An individual prediction profile compared to an average profile is illustrated according to a representative embodiment.

[0092] exist Fig. 8A , Figure 8B and Figure 8C In , a single prediction curve can be compared to the average curves for the long-stay (e.g., greater than eight days) and short-stay (e.g., less than or equal to eight days) groups for the entire oncology dataset. Error bands are shown as 10th and 90th percentiles, and events are represented by vertical lines. On the test set, the trained model may also be able to predict the recovery score of a patient based on the expected profile. Individual profiles can be compared to previous clusters of patient profiles, such as in Fig. 8A , Figure 8B and Figure 8C As shown in the figure. For example, oncology patients can be divided into two groups: those with short hospital stays (less than or equal to eight days) and those with long hospital stays (more than eight days). The patients can then be grouped accordingly. Although logically, the profiles of the groups may overlap, this division can be used to identify individual patients who are likely to have a shorter hospital stay (e.g. Fig. 8A ) or individual patients who may have a longer hospital stay (e.g. Figure 8B and Figure 8C ). Fig. 8A The predicted recovery score in the model shows that in the first few days after surgery, the patient is likely to have a shorter hospital stay, which can help with discharge planning, such as by scheduling the patient for early discharge. According to the model, a recovery score of 1.8 indicates readiness for discharge. However, the recovery prediction should not replace current care, and other factors play a role that are not reflected in the recovery score. But the recovery score can serve as an objective indicator of discharge readiness assessment. A new / refined prediction is provided every day. Communication with patients and healthcare providers should make it clear that the prediction is an expectation, not a guarantee. Figure 8B and Figure 8C The recovery of patients whose recovery was lower than the mean of the long-stay group is shown. Figure 8B The patient was discharged on the ninth day, but the predicted recovery score was still well below 1.8, and he was readmitted on the thirteenth day, indicating that the original discharge may have been too early. Figure 8CThe patient in also showed slow recovery compared to the reference curve, with three consecutive days of decline after the seventh day and complications on the eleventh day. Three consecutive days of declining scores to the lower end of the long-stay group may prompt healthcare providers to identify this patient as needing additional care and potentially prevent complications. Additionally, in the final model for this example use case for predicting recovery in oncology patients, only six features were used to enhance interpretability, namely activity, circadian rhythm, heart rate recovery, and a combination of differential features. By looking at these features, it is possible to understand why the recovery score is low, such as due to fewer steps or upright time, or poor circadian rhythm or heart rate recovery, which may point to different underlying issues and help explain the patient's condition.

[0093] Fig. 9 Predicted mean recovery curves according to representative embodiments are illustrated.

[0094] Fig. 9 The predicted mean recovery for the oncology and bariatric test set patient groups is shown. Figure 5A , Figure 5B , Figure 5C , Figure 5D , Figure 5E , Fig. 5F , Figure 6 , Fig. 7A , Figure 7B , Fig. 8A , Figure 8B and Figure 8C The first set of example embodiments explained involved predictions for oncology patients following abdominal surgery. Similar methods can be used for different surgical patient groups, and testing of the model trained on the oncology group on a separate group of bariatric surgery patients showed that the recovery of that group could also be predicted over time. In fact, the bariatric surgery group was predicted to recover faster than the oncology group, although this might have been expected based on the shorter total hospital stay. The predicted recovery for the bariatric surgery group is shown in Figure 2. Fig. 9 shown.

[0095] like Fig. 9 As shown, for most patients, the time to reach a score of 1.8 (defined as readiness for discharge) was later than the actual time of discharge, indicating that the trained model may need to be retrained for bariatric patients. Additionally, or alternatively, different features or models may be used to make predictions for the bariatric surgery group. Furthermore, a different reference recovery score may be more appropriate for training a model for the bariatric patient group because patients may tend to have similar short hospital stays, fewer complications, and readmissions, which are key differentiators for creating a reference profile in the oncology patient group.

[0096] In a first set of embodiments, parameters related to vital signs, activity and combinations thereof are derived. However, for example, sleep metrics or metrics related to pain, mental and social health may also be incorporated.

[0097] In a first set of embodiments, the features can be transformed into a single value per day. However, different time windows can also be used. Furthermore, the individual features can be combined using, for example, principal component analysis. Furthermore, stronger changes in the feature value characteristics can be used. For example, features with multiple time points can be used, rather than just the change relative to the previous time point, as this can provide additional information about the trend.

[0098] In some embodiments, raw data (e.g., accelerometer signals, ECG data, or PPG data) may be used to provide a rehabilitation score. Physiological and movement information important for estimating a rehabilitation score may be extracted directly from the raw data, for example, by using end-to-end deep learning.

[0099] In some embodiments, information about previous recovery trends of the same patient can be used to improve the accuracy of the latest estimated recovery trend. Such information may be useful for patients who are frequently hospitalized.

[0100] The patient grouping described herein is not limited to two groups of recovery trends. On the contrary, in some embodiments, the output of S355 may indicate the possibility that the patient belongs to more than two recovery trends of slow recovery or fast recovery. For example, the number of patient groups may exceed two. In addition, after the patient is initially assigned to a recovery trend group, the recovery trend group can be further divided into subgroups to more accurately classify the patients. The splitting of the recovery trend groups can be iteratively performed until the trend group reaches the desired accuracy or specificity, or until the number of patients in each subgroup is no longer above a predetermined threshold.

[0101] The teachings herein are primarily directed to predicting the recovery of individual patients. However, the trained models described herein can also be used to make predictions for entire hospitals or wards. For example, the recovery curve averaged across the entire population of patients at one hospital can be compared to a set of similar curves from different hospitals. In this way, problems such as infections in a hospital can be detected to reflect whether the curve for that hospital is different from that of other hospitals. Geographic density can also be determined in a similar manner.

[0102] Fig.10 A computer system on which an abnormal patient recovery detection method is implemented is illustrated according to another representative embodiment.

[0103] refer to Fig.10, the computer system 1000 includes a set of software instructions that can be executed to cause the computer system 1000 to perform any method or computer-based function disclosed herein. The computer system 1000 can operate as a standalone device, or can be connected to other computer systems or peripheral devices, for example using a network 1001. In an embodiment, the computer system 1000 performs logical processing based on digital signals received through an analog-to-digital converter.

[0104] In a networked deployment, the computer system 1000 can operate as a server or client user computer in a server-client user network environment, or as a peer computer system in a peer (or distributed) network environment. The computer system 1000 can also be implemented as or incorporated into various devices, such as a workstation, a fixed computer, a mobile computer, a personal computer (PC), a laptop, a tablet computer, or any other machine capable of executing a set (sequential or otherwise) of software instructions including a controller, the set of software instructions specifying the operations to be taken by the machine. The computer system 1000 can be incorporated as a device or incorporated into a device, which is then included in an integrated system containing additional devices. In an embodiment, the computer system 1000 can be implemented using an electronic device that provides voice, video or data communication. In addition, the computer system 1000 is shown as a monomer, but the term "system" should also be considered to include a collection of any system or subsystem that executes one or more sets of software instructions individually or jointly to perform one or more computer functions.

[0105] like Fig.10As shown in , computer system 1000 includes processor 1010. Processor 1010 can be regarded as a representative example of a processor of a controller and executes instructions to implement some or all aspects of the methods and processes described herein. Processor 1010 is tangible and non-transient. As used herein, the term "non-transient" should not be interpreted as a permanent feature of a state, but rather as a feature of a state that will last for a period of time. The term "non-transient" specifically negates fleeting features, such as features of a specific propagation carrier or signal or other forms that exist only transiently anywhere at any time. Processor 1010 is an article and / or machine component. Processor 222 is configured to execute software instructions to perform functions as described in various embodiments herein. Processor 1010 may be a general-purpose processor, or may be part of an application-specific integrated circuit (ASIC). Processor 1010 may also be a microprocessor, a microcomputer, a processor chip, a controller, a microcontroller, a digital signal processor (DSP), a state machine, or a programmable logic device. Processor 1010 may also be a logic circuit including a programmable gate array (PGA) such as a field programmable gate array (FPGA), or another type of circuit including discrete gates and / or transistor logic. Processor 1010 may be a central processing unit (CPU), a graphics processing unit (GPU), or both. In addition, any processor described herein may include multiple processors, parallel processors, or both. Multiple processors may be included in or coupled to a single device or multiple devices.

[0106] The term "processor" as used herein encompasses an electronic component capable of running a program or machine executable instructions. References to a computing device that includes a "processor" should be interpreted as including more than one processor or processing core, such as a multi-core processor. A processor may also refer to a collection of processors within a single computer system or distributed between multiple computer systems. The term computing device should also be interpreted as including a collection or network of computing devices, each of which includes a processor or multiple processors. A program has software instructions executed by one or more processors, which may be within the same computing device or may be distributed across multiple computing devices.

[0107] The computer system 1000 also includes a main memory 1020 and a static memory 1030, wherein the memories in the computer system 1000 communicate with each other and with the processor 1010 via a bus 1008. Either or both of the main memory 1020 and the static memory 1030 can be considered as representative examples of the memory of the controller, and store instructions for implementing some or all aspects of the methods and processes described herein. The memory 307 described herein is a tangible storage medium for storing data and executable software instructions, and is non-transient during the time when the software instructions are stored. As used herein, the term "non-transient" should not be interpreted as a permanent feature of a state, but rather as a feature of a state that will last for a period of time. The term "non-transient" specifically negates fleeting features, such as features of a specific propagating carrier or signal or other forms that exist only transiently anywhere at any time. The main memory 1020 and the static memory 1030 are manufactured articles and / or machine components. The main memory 1020 and the static memory 1030 are computer-readable media from which a computer (e.g., the processor 1010) can read data and executable software instructions. Both the memory 1020 and the static memory 1030 can be implemented as random access memory (RAM), read-only memory (ROM), flash memory, electrically programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), registers, one or more disks in a hard disk, a removable disk, a magnetic tape, a compact disk read-only memory (CD-ROM), a digital versatile disk (DVD), a floppy disk, a Blu-ray disc, or any other form of storage medium known in the art. The memory can be volatile or non-volatile, secure and / or encrypted, unsecure and / or unencrypted.

[0108] "Memory" is an example of a computer-readable storage medium. Computer memory is any memory that a processor can directly access. Examples of computer memory include, but are not limited to, RAM memory, registers, and register files. References to "computer memory" or "memory" should be interpreted as potentially being multiple memories. The memory may be, for example, multiple memories within the same computer system. The memory may also be multiple memories distributed among multiple computer systems or computing devices.

[0109] As shown, the computer system 1000 also includes a video display unit 1050, such as a liquid crystal display (LCD), an organic light emitting diode (OLED), a flat panel display, a solid state display, or a cathode ray tube (CRT). In addition, the computer system 1000 includes an input device 1060, such as a keyboard / virtual keyboard or a touch input screen or a voice input with voice recognition, and a cursor control device 1070, such as a mouse or a touch input screen or a touch pad. The computer system 1000 also optionally includes a disk drive unit 1080, a signal generating device 1090 (such as a speaker or a remote controller) and / or a network interface device 1040.

[0110] In one embodiment, Fig.10 As shown in FIG. 1 , the disk drive unit 1080 includes a computer readable medium 1082 in which one or more sets of software instructions 1084 (software) are embedded. The set of software instructions 1084 is an example of a computer product or a computer program product, which when executed can be used to directly and / or indirectly implement one or more aspects of the methods described herein, such as the methods attributed to the controller. The set of software instructions 1084 is read from the computer readable medium 1082 and executed by the processor 1010. In addition, when the processor 1010 executes the software instructions 1084, one or more steps of the methods and processes described herein are performed. In one embodiment, the software instructions 1084 reside in the main memory 1020, the static memory 1030 and / or the processor 1010 during execution by the computer system 1000. In addition, the computer readable medium 1082 can include the software instructions 1084 or receive and execute the software instructions 1084 in response to a propagation signal, so that a device connected to the network 1001 transmits voice, video or data through the network 1001. Software instructions 1084 may be transmitted or received over network 1001 via network interface device 1040 .

[0111] In one embodiment, a dedicated hardware implementation is constructed, such as an application specific integrated circuit (ASIC), a field programmable gate array (FPGA), a programmable logic array, and other hardware components to implement one or more methods described herein. One or more embodiments described herein may use two or more specific interconnected hardware modules or devices to implement functions, and the hardware modules or devices have related controls that can communicate between modules and through modules. Therefore, the present disclosure includes software, firmware, and hardware implementations. Nothing in this application should be interpreted as being implemented or implemented only with software rather than hardware such as tangible non-transient processors and / or memories.

[0112] According to various embodiments of the present disclosure, a hardware computer system executing a software program can be used to implement the methods described herein. In addition, in an example non-limiting embodiment, the implementation can include distributed processing, component / object distributed processing, and parallel processing. Virtual computer system processing can be used to implement one or more methods or functions as described herein, and the processor 205 described herein can be used to support a virtual processing environment.

[0113] Therefore, abnormal patient rehabilitation detection can improve the accuracy of patient care plans, physical therapy plans, and improve resource utilization.

[0114] Although the abnormal patient recovery test has been described with reference to several exemplary embodiments, it should be understood that the words used are words of description and illustration, rather than words of limitation. Changes may be made within the scope of the appended claims, as presently stated and amended, without departing from the scope and spirit of the abnormal patient recovery test. Although the abnormal patient recovery test has been described with reference to specific methods, materials, and embodiments, the abnormal patient recovery test is not intended to be limited to the details disclosed. Rather, the abnormal patient recovery test extends to all functionally equivalent structures, methods, and uses, such as are within the scope of the appended claims.

[0115] The illustrations of the embodiments described herein are intended to provide a general understanding of the structures of various embodiments. These illustrations are not intended to be used as a complete description of all elements and features of the disclosure described herein. After reviewing the disclosure, many other embodiments will be apparent to those skilled in the art. Other embodiments may be utilized and derived from the disclosure so that structural and logical replacements and changes may be made without departing from the scope of the disclosure. In addition, these illustrations are only representative and may not be drawn to scale. Certain ratios in the illustrations may be amplified, while other ratios may be minimized. Therefore, the disclosure and the accompanying drawings should be considered to be illustrative rather than restrictive.

[0116] One or more embodiments disclosed herein may be referred to as the term "invention" independently and / or collectively for convenience purposes only, but this does not mean that the scope of the present application is limited to any particular invention or inventive concept. In addition, although specific embodiments have been illustrated and described herein, it should be understood that any subsequent arrangement designed to achieve the same or similar purpose can replace the specific embodiments shown. The present disclosure is intended to cover any and all subsequent modifications or variations of the various embodiments. By reviewing the specification, the combination of the above embodiments and other embodiments not specifically described herein will be apparent to those skilled in the art.

[0117] The Abstract of the present disclosure is provided to comply with 37 CFR §1.72(b) and is submitted with the understanding that it is not intended to interpret or limit the scope or meaning of the claims. In addition, in the foregoing Detailed Description, various features may be grouped together or described in a single embodiment in order to simplify the disclosure. This disclosure should not be interpreted as reflecting an intention that the claimed embodiments require more features than are expressly recited in each claim. On the contrary, as reflected in the following claims, an inventive indication may be directed to less than all features of any one of the disclosed embodiments. Therefore, the following claims are incorporated into the Detailed Description, with each claim independently defining the claimed indication.

[0118] The foregoing description of the disclosed embodiments is provided to enable any person skilled in the art to practice the concepts described in the present disclosure. As such, the subject matter disclosed above should be considered illustrative rather than restrictive, and the claims are intended to cover all such modifications, enhancements, and other embodiments that fall within the true spirit and scope of the present disclosure. Therefore, to the maximum extent permitted by law, the scope of the present disclosure will be determined by the broadest permissible interpretation of the following claims and their equivalents, and shall not be limited or restricted to the foregoing detailed description.

Claims

1. A system comprising: a memory storing instructions; as well as a processor that executes the instructions, wherein the instructions, when executed by the processor, cause the system to: Repeatedly obtain multiple types of data from rehabilitation patients; applying the plurality of types of data to the trained model; repeatedly generating a rehabilitation score that predicts a rehabilitation status of the rehabilitation patient; and The rehabilitation score is output.

2. The system according to claim 1, wherein: The instructions, when executed by the processor, further cause the system to: Establish a baseline; The recovery score was compared to the baseline.

3. A system according to any one of claims 1 or 2, wherein: The instructions, when executed by the processor, further cause the system to: An alert is generated based on a comparison of the recovery score to a baseline.

4. A system according to any one of claims 1, 2 or 3, wherein: The instructions, when executed by the processor, further cause the system to: An alert is generated based on a recovery score trend including the recovery score.

5. A system according to any one of claims 1, 2, 3 or 4, wherein: The system is centralized and generates and outputs rehabilitation scores for a plurality of rehabilitation patients.

6. A system according to any one of claims 1, 2, 3, 4 or 5, wherein: The system generates and outputs rehabilitation scores for rehabilitation patients at multiple medical institutions.

7. The medical system of any one of claims 1, 2, 3, 4, 5 or 6, wherein: The trained models include extreme gradient boosting based models.

8. A system according to any one of claims 1, 2, 3, 4, 5, 6 or 7, in, The plurality of types of data include data derived from vital signs of the recovering patient.

9. A system according to any one of claims 1, 2, 3, 4, 5, 6, 7 or 8, in, The plurality of types of data include data derived from activity metrics of the rehabilitation patient.

10. The system according to any one of claims 1 or 9, in, The plurality of types of data also includes data derived from vital signs of the rehabilitating patient.

11. A method performed by a central system, comprising: Repeatedly obtain multiple types of data from rehabilitation patients; applying, by a processor executing instructions from a memory, the plurality of types of data to the trained model; repeatedly generating a rehabilitation score that predicts a rehabilitation status of the rehabilitation patient; and The rehabilitation score is output.

12. The method according to claim 11, further comprising: Establish a baseline; and The recovery score was compared to the baseline.

13. The method according to any one of claims 11 or 12, further comprising: An alert is generated based on a comparison of the recovery score to a baseline.

14. The method according to claim 11, 12 or 13, further comprising: An alert is generated based on a recovery score trend including the recovery score.

15. The method according to any one of claims 13 or 14, wherein: The central system generates and outputs a rehabilitation score for a plurality of rehabilitation patients.

16. The method according to any one of claims 11, 12, 13, 14 or 15, wherein: The central system generates and outputs rehabilitation scores for rehabilitation patients at multiple medical institutions.

17. The method according to any one of claims 11, 12, 13, 14, 15 or 16, wherein: The trained models include extreme gradient boosting based models.

18. The method according to any one of claims 11, 12, 13, 14, 15, 16 or 17, in, The multiple types of data are derived from the vital signs of the recovering patient.

19. The method according to any one of claims 11, 12, 13, 14, 15, 16, 17 or 18, in, The multiple types of data are derived from activity metrics of the rehabilitation patient.

20. The method according to any one of claims 11 or 19, in, The plurality of types of data also includes data derived from vital signs of the rehabilitating patient.

21. A tangible, non-transitory computer-readable storage medium storing a computer program, wherein: The computer program, when executed by a processor, causes the system to: Repeatedly obtain multiple types of data from rehabilitation patients; applying the plurality of types of data to the trained model; repeatedly generating a rehabilitation score that predicts a rehabilitation status of the rehabilitation patient; and The rehabilitation score is output.

Images