US-12622664-B2 - Acute respiratory distress syndrome evaluating method and system thereof

US12622664B2US 12622664 B2US12622664 B2US 12622664B2US-12622664-B2

Abstract

An acute respiratory distress syndrome (ARDS) evaluating method includes an image preprocessing step, a lung infiltration determining step, a probability generating step and an evaluating step. The image preprocessing step includes inputting an X-ray image data to a first model to generate an X-ray partial image. The lung infiltration determining step includes inputting the X-ray partial image to a second model to generate a lung infiltration probability and a lung infiltration region image. The probability generating step includes inputting the lung infiltration probability, a blood inspection data, a vital signs data and a respiratory data to a third model to generate an ARDS suffering probability. The evaluating step includes calculating an evaluating result according to the lung infiltration region image and the ARDS suffering probability.

Inventors

Kai-Cheng Hsu
Wei-Cheng Chen
Wei-Yang Yu
Yu-Chao Lin
How-Yang Tseng
Chieh-Lung CHEN
Xin-Jie Liang
Bo-Hao Yang

Assignees

CHINA MEDICAL UNIVERSITY

Dates

Publication Date: 20260512
Application Date: 20241004
Priority Date: 20240131

Claims (14)

1 . An acute respiratory distress syndrome (ARDS) evaluating method, comprising: an image preprocessing step comprising driving a processor to input an X-ray image data to a first model so as to generate an X-ray partial image; a lung infiltration determining step comprising driving the processor to input the X-ray partial image to a second model so as to generate a lung infiltration probability and a lung infiltration region image; a probability generating step comprising driving the processor to input the lung infiltration probability, a blood inspection data, a vital signs data and a respiratory data to a third model so as to generate an ARDS suffering probability; and an evaluating step comprising driving the processor to calculate the lung infiltration region image and the ARDS suffering probability so as to obtain an evaluating result; wherein the first model is an Efficient-UNet model, the second model is an Efficient-Net model and comprises a Convolutional Block Attention Module (CBAM), and the third model is an extreme Gradient Boosting model.
2 . The acute respiratory distress syndrome evaluating method of claim 1 , wherein the X-ray image data is cut by the first model so as to obtain the X-ray partial image, and the X-ray partial image consists of a lung region image.
3 . The acute respiratory distress syndrome evaluating method of claim 1 , further comprising: a model training step, comprising: a first training step comprising driving the processor to train the first model based on an X-ray image training set and a lung X-ray image training set; a second training step comprising driving the processor to train the second model based on a lung infiltration X-ray image training set and the lung X-ray image training set; and a third training step comprising driving the processor to train the third model based on a plurality of medical datasets, wherein each of the plurality of medical datasets comprises the lung infiltration probability, the blood inspection data, the vital signs data and the respiratory data.
4 . The acute respiratory distress syndrome evaluating method of claim 3 , wherein the lung X-ray image training set comprises a plurality of labelled chest X-ray images.
5 . The acute respiratory distress syndrome evaluating method of claim 3 , wherein the third training step further comprises: driving the processor to calculate the plurality of medical datasets so as to obtain an average value, a standard deviation, a skewness value and a kurtosis value of at least one of the blood inspection data, the vital signs data and the respiratory data of each of the plurality of medical datasets, and then to train the plurality of medical datasets and the average value, the standard deviation, the skewness value and the kurtosis value of the at least one of the blood inspection data, the vital signs data and the respiratory data of each of the plurality of medical datasets so as to establish the third model.
6 . The acute respiratory distress syndrome evaluating method of claim 3 , wherein each of the plurality of medical datasets comprises a plurality of feature information, and the acute respiratory distress syndrome evaluating method further comprises: a data preprocessing step comprising driving the processor to assess whether a number of the plurality of feature information of each of the plurality of medical datasets is equal to a present number or not; when the number of the plurality of feature information of each of the plurality of medical datasets is smaller than the present number, the processor is driven to fill in an average value corresponding to one of the plurality of feature information being missing based on an imputation procedure; and when the number of the plurality of feature information of each of the plurality of medical datasets is larger than the present number, and one of the plurality of feature information comprises at least two values, the processor is driven to replace the at least two values with another average value of the at least two values.
7 . The acute respiratory distress syndrome evaluating method of claim 6 , wherein the imputation procedure is a K-Nearest Neighbor (KNN) algorithm.
8 . An acute respiratory distress syndrome evaluating system, comprising: a database for accessing a first model, a second model, a third model, an X-ray image data, a blood inspection data, a vital signs data and a respiratory data; and a processor signally connected to the database and configured to perform an acute respiratory distress syndrome evaluating method, wherein the acute respiratory distress syndrome evaluating method comprises: an image preprocessing step comprising inputting the X-ray image data to the first model so as to generate an X-ray partial image; a lung infiltration determining step comprising inputting the X-ray partial image to the second model so as to generate a lung infiltration probability and a lung infiltration region image; a probability generating step comprising inputting the lung infiltration probability, the blood inspection data, the vital signs data and the respiratory data to the third model so as to generate an ARDS suffering probability; and an evaluating step comprising calculating the lung infiltration region image and the ARDS suffering probability so as to obtain an evaluating result; wherein the first model is an Efficient-UNet model, the second model is an Efficient-Net model and comprises a Convolutional Block Attention Module (CBAM), and the third model is an extreme Gradient Boosting model.
9 . The acute respiratory distress syndrome evaluating system of claim 8 , wherein the X-ray image data is cut by the first model so as to obtain the X-ray partial image, and the X-ray partial image consists of a lung region image.
10 . The acute respiratory distress syndrome evaluating system of claim 8 , wherein the acute respiratory distress syndrome evaluating method further comprises: a model training step, comprising: a first training step comprising training the first model based on an X-ray image training set and a lung X-ray image training set; a second training step comprising training the second model based on a lung infiltration X-ray image training set and the lung X-ray image training set; and a third training step comprising training the third model based on a plurality of medical datasets, wherein each of the plurality of medical datasets comprises the lung infiltration probability, the blood inspection data, the vital signs data and the respiratory data.
11 . The acute respiratory distress syndrome evaluating system of claim 10 , wherein the lung X-ray image training set comprises a plurality of labelled chest X-ray images.
12 . The acute respiratory distress syndrome evaluating system of claim 10 , wherein the third training step further comprises: calculating the plurality of medical datasets so as to obtain an average value, a standard deviation, a skewness value and a kurtosis value of at least one of the blood inspection data, the vital signs data and the respiratory data of each of the plurality of medical datasets, and then training the plurality of medical datasets and the average value, the standard deviation, the skewness value and the kurtosis value of the at least one of the blood inspection data, the vital signs data and the respiratory data of each of the plurality of medical datasets so as to establish the third model.
13 . The acute respiratory distress syndrome evaluating system of claim 10 , wherein each of the medical datasets comprises a plurality of feature information and the acute respiratory distress syndrome evaluating method further comprises: a data preprocessing step comprising assessing whether a number of the plurality of feature information of each of the plurality of medical datasets is equal to a present number or not; when the number of the plurality of feature information of each of the plurality of medical datasets is smaller than the present number, an average value corresponding to one of the plurality of feature information being missing is filled therein based on an imputation procedure; and when the number of the plurality of feature information of each of the plurality of medical datasets is larger than the present number, and one of the plurality of feature information comprises at least two values, the at least two values are replaced with another average value of the at least two values.
14 . The acute respiratory distress syndrome evaluating system of claim 13 , wherein the imputation procedure is a K-Nearest Neighbor (KNN) algorithm.

Description

RELATED APPLICATIONS This application claims priority to Taiwan Application Serial Number 113103702, filed Jan. 31, 2024, which is herein incorporated by reference. BACKGROUND Technical Field The present disclosure relates to an evaluating method and a system thereof. More particularly, the present disclosure relates to an acute respiratory distress syndrome (ARDS) evaluating method and a system thereof. Description of Related Art Acute respiratory distress syndrome (ARDS) is a severe disease with high mortality rate, the rapid clinical course progress thereof is the reason for the high mortality rate, and it's difficult to implement by early treatment. According to current research statistics, only 50% of the mild patients and 75% of the severe patients can be successfully diagnosed, and less than two-thirds of the patients can receive the treatment of the lung-protective ventilation. Therefore, there is still a lack of an acute respiratory distress syndrome evaluating method and a system thereof that can be used to early diagnose, immediately treatment and dynamically adjust the acute respiratory distress syndrome according to the course thereof, so that it is indeed the expectation of people, and it's also the goal and direction of work for related industry. SUMMARY The present disclosure provides an acute respiratory distress syndrome (ARDS) evaluating method includes an image preprocessing step, a lung infiltration determining step, a probability generating step and an evaluating step. The image preprocessing step includes driving a processor to input an X-ray image data to a first model so as to generate an X-ray partial image. The lung infiltration determining step includes driving the processor to input the X-ray partial image to a second model so as to generate a lung infiltration probability and a lung infiltration region image. The probability generating step includes driving the processor to input the lung infiltration probability, a blood inspection data, a vital signs data and a respiratory data to a third model so as to generate an ARDS suffering probability. The evaluating step includes driving the processor to calculate the lung infiltration region image and the ARDS suffering probability so as to obtain an evaluating result. The present disclosure provides an acute respiratory distress syndrome evaluating system includes a database and a processor. The database is use for accessing a first model, a second model, a third model, an X-ray image data, a blood inspection data, a vital signs data and a respiratory data. The processor is signally connected to the database and configured to perform an acute respiratory distress syndrome evaluating method, wherein the acute respiratory distress syndrome evaluating method includes an image preprocessing step, a lung infiltration determining step, a probability generating step and an evaluating step. The image preprocessing step includes inputting the X-ray image data to the first model so as to generate an X-ray partial image. The lung infiltration determining step includes inputting the X-ray partial image to the second model so as to generate a lung infiltration probability and a lung infiltration region image. The probability generating step includes inputting the lung infiltration probability, the blood inspection data, the vital signs data and the respiratory data to the third model so as to generate an ARDS suffering probability. The evaluating step includes calculating the lung infiltration region image and the ARDS suffering probability so as to obtain an evaluating result. BRIEF DESCRIPTION OF THE DRAWINGS The present disclosure can be more fully understood by reading the following detailed description of the embodiment, with reference made to the accompanying drawings as follows: FIG. 1 is a block diagram of an acute respiratory distress syndrome evaluating system according to the first example of the present disclosure. FIG. 2 is a flow chart of an acute respiratory distress syndrome evaluating method according to the second example of the present disclosure. FIG. 3 is a schematic view of an evaluating result of the acute respiratory distress syndrome evaluating method of FIG. 2. FIG. 4 is a flow chart of an acute respiratory distress syndrome evaluating method according to the third example of the present disclosure. DETAILED DESCRIPTION The embodiment will be described with the drawings. For clarity, some practical details will be described below. However, it should be noted that the present disclosure should not be limited by the practical details, that is, in some embodiment, the practical details is unnecessary. In addition, for simplifying the drawings, some conventional structures and elements will be simply illustrated, and repeated elements may be represented by the same labels. It will be understood that when an element (or device) is referred to as be “connected to” another element, it can be directly connected to other element