Abstract: Example implementations described herein can involve systems and methods involving, for receipt of input data from one or more assets, identifying and separating different event contexts from the input data; training a plurality of machine learning models for each of the different event contexts; selecting a best performing model from the plurality of machine learning models to form a compound model; selecting a best performing subset of the input data for the compound model based on maximizing a metric; and deploying the compound model for the selected subset.

Abstract: Systems and methods for tracking activities from a plurality of multimodal inputs are described. Activity tracking can include receiving a plurality of multimodal inputs, synchronizing the plurality of multimodal inputs, generating segments from the synchronized multimodal inputs, recognizing activities associated with each generated segment by performing a bagged formal concept analysis (BFCA), and recording the recognized activities in a storage. Tracking of activities can include the detection of moments (e.g., eating moments), during which an activity tracking application can prompt a user for information (e.g., a food journal).

Abstract: In some examples, a system may generate a plurality of care path patient profile models based on a plurality of care path patterns for a plurality of past patient admissions. For example, each care path patient profile model may include a trained classifier. Further, the system may receive information related to a new patient admission, and may generate features from the received information related to the new patient admission. The system may input the features generated from the received information related to the new patient admission into the plurality of care path patient profile models to obtain a respective probability of being classified in a respective care path based on an amount of similarity to the patients who have gone through each care path. In addition, the system may present, on a display, information related to at least one care path pattern in a graphical user interface.

Abstract: Example implementations described herein involve systems and methods for calculating the importance of each iteration and of each input feature for multi-label models that optimize a multi-label objective function in an iterative manner. The example implementations are based on the incremental improvement in the objective function rather than an application-specific metric such as accuracy.

Abstract: Systems and methods for tracking activities from a plurality of multimodal inputs are described. Activity tracking can include receiving a plurality of multimodal inputs, synchronizing the plurality of multimodal inputs, generating segments from the synchronized multimodal inputs, recognizing activities associated with each generated segment by performing a bagged formal concept analysis (BFCA), and recording the recognized activities in a storage. Tracking of activities can include the detection of moments (e.g., eating moments), during which an activity tracking application can prompt a user for information (e.g., a food journal).

Abstract: In some examples, a system may generate a plurality of care path patient profile models based on a plurality of care path patterns for a plurality of past patient admissions. For example, each care path patient profile model may include a trained classifier. Further, the system may receive information related to a new patient admission, and may generate features from the received information related to the new patient admission. The system may input the features generated from the received information related to the new patient admission into the plurality of care path patient profile models to obtain a respective probability of being classified in a respective care path based on an amount of similarity to the patients who have gone through each care path. In addition, the system may present, on a display, information related to at least one care path pattern in a graphical user interface.

Abstract: Systems and methods are provided for detecting the orientation and/or movement of a patient having an implantable cardiac stimulation device and evaluating whether a change in the patient's cardiac activity can be at least in part due to a change in the patient's orientation. In one particular embodiment, signals from an orientation sensor and/or a pressure sensor are evaluated to determine static positional orientation of the patient and determine based on the static orientation whether the patient's cardiac activity is abnormal.

Abstract: A method and system are provided for trending a coronary burden such as an ischemic burden or acute myocardial infarction (AMI) for a patient. Trending provides obtaining cardiac data over a period of time, identifying the onset and the termination of coronary episodes based on a ST segment variation within the cardiac data, recording coronary burden information, and presenting the coronary burden information to a user. The coronary burden information may include the number of coronary episodes occurring over a period of time, the time duration of the coronary episodes, and the maximum ST segment variations for the coronary episodes that occur over a period of time.

Abstract: A method and system are provided for trending variation in coronary burden across multiple heart rate ranges. The method and system include obtaining cardiac signals having a segment of interest over a period of time where each cardiac signal has an associated heart rate that falls within at least one heart rate range. Segment variations of the segment of interest are determined and grouped based on the associated heart rates to produce distributions of segment variations that are associated with the heart rate ranges. Trending information is produced by automatically comparing the distributions of segment variations between different heart rate ranges.

Abstract: Techniques are described for adaptively adjusting detection thresholds for use in detecting cardiac ischemia and other abnormal physiological conditions based on morphological parameters derived from intracardiac electrogram (IEGM) signals, impedance measurements, or other signals. In one example, where ST segment elevation is used to detect cardiac ischemia, default detection thresholds are determined in advance from an examination of variations in ST segment elevations occurring within a population of patients. Thereafter, an individual pacemaker or other implantable medical device uses the default thresholds during an initial learning period to detect ischemia within the patient in which the device is implanted. During the initial learning period, the pacemaker also collects data representative of the range of variation in ST segment elevations occurring within the patient.

Abstract: Techniques are described for efficiently detecting and distinguishing among cardiac ischemia, hypoglycemia or hyperglycemia based on intracardiac electrogram (IEGM) signals. In one example, a preliminary indication of an episode of cardiac ischemia is detected based on shifts in ST segment elevation within the IEGM. In response, the implanted device then records additional IEGM data for transmission to an external system. The external system analyzes the additional IEGM data to confirm the detection of cardiac ischemia using a more sophisticated analysis procedure exploiting additional detection parameters. In particular, the external system uses detection parameters capable of distinguishing hypoglycemia, hyperglycemia and hyperkalemia from cardiac ischemia, such as QTmax and QTend intervals. Alternatively, the more sophisticated analysis procedure may be performed by the device itself, if it is so equipped. Other examples described herein pertain instead to the detection of atrial fibrillation.

Abstract: A method and system are provided for tracking ST shift data. The system includes an implantable medical device having an input configured to receive cardiac signals. Each cardiac signal has an associated heart rate and includes a segment of interest. The implantable medical device further includes a processor configured to determine segment variations of the segment of interest in the cardiac signals. The processor determines a heart rate associated with each of the segment variations with each heart rate falling within a corresponding heart rate range. The implantable medical device also includes a memory configured to store a group of histograms for a corresponding group of heart rate ranges. The histograms store distributions for the segment variations within corresponding heart rate ranges.