CN-122029615-A - Detection of hemodynamic instability

CN122029615ACN 122029615 ACN122029615 ACN 122029615ACN-122029615-A

Abstract

Embodiments of the present disclosure relate to a method of detecting hemodynamic instability. The method includes obtaining a plurality of clinical parameters of the subject at a point in time. A plurality of prediction probabilities may then be determined based on the plurality of clinical parameters, wherein each of the plurality of prediction probabilities is determined by a corresponding sub-model of a plurality of sub-models of the machine learning model. The method also includes determining a combined prediction probability for the point in time based on the plurality of prediction probabilities. The determined combined predictive probability is used to detect hemodynamic instability. According to embodiments of the present disclosure, hemodynamic instability can be detected timely and accurately to alert a clinician, thereby greatly reducing mortality and medical costs.

Inventors

JIANG ZEYU
TIAN CONG
FU YONG
YUAN YANA

Assignees

皇家飞利浦有限公司

Dates

Publication Date: 20260512
Application Date: 20240919
Priority Date: 20230925

Claims (6)

1. A method of training a machine learning model for detecting hemodynamic instability, comprising: Obtaining a plurality of sample clinical parameters for a plurality of sample subjects at a point in time, the plurality of sample clinical parameters including at least two of vital sign data, laboratory measurement data, waveform characteristic data, and image data; determining a ratio of a number of positive samples in the plurality of sample objects to a number of negative samples in the plurality of sample objects; generating a plurality of sub-data sets, each sub-data set including a sample clinical parameter corresponding to the positive sample and a sample clinical parameter corresponding to a portion of the negative samples, wherein the negative samples are divided into a plurality of portions based on the ratio and the portion of the negative samples for each sub-data set are different from each other, and Training a plurality of sub-models of the machine learning model for detecting the hemodynamic instability based on the plurality of sub-data sets.
2. The method of claim 1, wherein the machine learning model is one of a plurality of machine learning models that form a detection model for detecting the hemodynamic instability, and the method further comprises: Obtaining a further plurality of sample clinical parameters of the plurality of sample objects at a further point in time different from the point in time, and Additional machine learning models of the plurality of machine learning models are trained based on the additional plurality of sample clinical parameters.
3. The method of claim 1, wherein obtaining a plurality of sample clinical parameters for a plurality of sample objects at a point in time comprises: Determining whether a sample clinical parameter of the plurality of sample clinical parameters is empty; Determining whether up-to-date data for the sample clinical parameters exists within a first screening time period in response to determining that the sample clinical parameters of the plurality of sample clinical parameters are empty, and The sample clinical parameters are updated based on the latest data in response to determining that the latest data for the sample clinical parameters exists within the first screening time limit.
4. The method of claim 3, wherein training a plurality of sub-models of the machine learning model for detecting the hemodynamic instability based on the plurality of sub-data sets comprises: obtaining a prediction probability of a sub-model of the plurality of sub-models by inputting a sub-data set of the plurality of sub-data sets into the sub-model, and Model parameters of the sub-model are adjusted based on the predicted probability and the true result of hemodynamic instability.
5. An electronic device, comprising: at least one processor, and At least one memory having stored thereon a plurality of instructions that, when executed by the at least one processor, cause the apparatus to perform the method of any of claims 1-4.
6. A computer readable medium having stored thereon computer instructions, which when executed by a processor, cause the processor to perform the method according to any of claims 1-4.

Description

Detection of hemodynamic instability Technical Field Embodiments of the present disclosure relate generally to the field of data processing, and more particularly to detection of hemodynamic instability. Background Hemodynamic instability describes unstable blood movement. It is a condition or state where a person's cardiovascular function becomes unreliable, inadequate or otherwise problematic. Thus, hemodynamic instability is a critical and common condition in the Intensive Care Unit (ICU). Because hemodynamic instability in post-traumatic patients cannot be detected in time, one third of ICU patients will develop shock with higher mortality. The hemodynamic instability warning system has the potential to improve timely detection and initiation of interventions. If hemodynamic instability in a patient after a trauma can be predicted in time, a prophylactic treatment plan can be provided for high risk patients, thereby greatly reducing mortality and medical costs. However, the warning system of hemodynamic instability still has some problems that need to be addressed to improve the prediction of hemodynamic instability. Potes et al, volume "A clinical prediction model to identify patients at high risk of hemodynamic instability in the pediatric intensive care unit"（Critical Care,, volume 21, stage 1, 12,1, 2017) proposes a model for predicting the need for hemodynamic intervention in a pediatric intensive care unit. Disclosure of Invention In an early warning system of hemodynamic instability, single parameter shock indicators (e.g., systolic Blood Pressure (SBP) and shock index (heart rate/SBP)) are reported to detect hemodynamic instability. However, the patient's condition will worsen at the later stages of shock and/or risk may be underestimated by merely addressing changes in the cardiovascular system. Low-precision pre-warning systems also do not facilitate clinicians to identify and interpret relevant information, and thus it is difficult for clinicians to continuously monitor or evaluate ICU patients. Low frequency and low accuracy warning systems are not suitable for patients with abrupt changes in condition. In the absence of predictive tools with sufficient accuracy, the clinician may take subjective judgment, which increases the risk that the clinician will take a wrong action. In view of this, embodiments of the present disclosure propose a method for detecting hemodynamic instability, a method of training a machine learning model for detecting hemodynamic instability, an electronic device, and a computer readable medium. The present disclosure enables timely and accurate detection of hemodynamic instability to alert a clinician, thereby greatly reducing mortality and medical costs. In a first aspect, embodiments of the present disclosure provide a method of detecting hemodynamic instability. The method includes obtaining a plurality of clinical parameters of the subject at a point in time, the plurality of clinical parameters including at least two of vital sign data, laboratory measurement data, waveform characteristic data, and image data. The method also includes determining, by the machine learning model, a plurality of prediction probabilities based on the plurality of clinical parameters, each of the plurality of prediction probabilities determined by a corresponding sub-model of a plurality of sub-models of the machine learning model. The method also includes determining a combined prediction probability for the point in time based on the plurality of prediction probabilities. The method further includes detecting hemodynamic instability based on the combined predictive probabilities. According to embodiments of the present disclosure, the method uses at least two types of clinical parameters. Thus, more parameters indicative of the underlying pathophysiology of the cardiovascular system will be used to improve the accuracy of the prediction. Further, the plurality of sub-models may determine a plurality of prediction probabilities using a plurality of clinical parameters and determine a combined prediction probability from the plurality of prediction probabilities. The combined prediction probability will be more accurate, so the method can accurately detect hemodynamic instability in time to alert the clinician, thereby greatly reducing mortality and medical costs. In some embodiments of the first aspect, the vital sign data comprises at least one of heart rate, temperature, respiration rate, and blood pressure, the laboratory measurement data comprises at least one of creatinine, blood glucose, glutamic acid, lactic acid, alanine aminotransferase, and aspartate aminotransferase, the waveform characteristic data comprises at least one of a peak Gu Xielv of a heart rate waveform, a peak spacing of a heart rate waveform, a peak Gu Xielv of a ventilator waveform, and a peak spacing of a ventilator waveform, and/or the image data comprises at least one of an X-ray image, an ultrasound im