Abstract: An image pre-processing method and an image processing apparatus for a fundoscopic image are provided. A region of interest (ROI) is obtained from a fundoscopic image to generate a first image. The ROI is focused on an eyeball in the fundoscopic image. A smoothing process is performed on the first image to generate a second image. A value difference between neighboring pixels in the second image is increased to generate a third image.

Abstract: An electronic device and a method for diagnosing heart state based on electrocardiogram (ECG) are provided. An ECG file is obtained, and the ECG file is in a first file format and includes a plurality of potential traces of a plurality of leads. The ECG file is converted to a second file format to obtain electrocardiogram data corresponding to multiple leads. Each potential trace relative to time in the ECG file is converted to the ECG data of each lead. Integrated ECG data associated with the leads is generated based on the ECG data of the plurality of leads through the zero-padding operation and the stacking operation. A diagnostic result of heart status is generated based on the integrated ECG data and a deep learning model.

Abstract: The disclosure provides a disease classification method and a disease classification device. The disease classification method includes: inputting samples into a first stage model and obtaining a first stage determination result; inputting first samples determined positive by the first stage model into a second stage high specificity model to obtain second samples determined to be positive and third samples determined to be negative and rule in the second samples; inputting fourth samples determined negative by the first stage model into a second stage high sensitivity model to obtain fifth samples determined to be positive and sixth samples determined to be negative and rule out the sixth samples; obtaining a second stage determination result of the second and sixth samples; and inputting the third and fifth samples not ruled in or ruled out into a third stage model and obtaining a third stage determination result of the third and fifth samples.

Abstract: A construction method of ribosomal RNA database is provided, including the following steps: selecting a source of nucleic acid sequence database; performing normalization and homogenization on species classification rules; using AI technology for normalized classification and correction; selecting the kingdom to which the sequence species belongs; filtering out redundant sequences and sequences with inconsistent lengths; setting a threshold for unknown bases other than A, T, C or G, and excluding unknown bases that exceed the threshold; and excluding sequences with insufficient classification information.

Abstract: A feature selection method is provided, including: inputting a plurality of pieces of training data into a plurality of training models to perform selection in a plurality of features through each training model for obtaining multiple feature pools; sorting the features based on the number of times each feature is selected by the feature pools to obtain a feature ranking; and extracting a plurality of designated features from the features based on the feature ranking.

Abstract: A method and an apparatus for compensating image segmentation lines are provided. In the method, image segmentation is applied to a medical image captured by using a segmentation model to obtain a segmentation image including at least one segmentation line between multiple layers in the medical image. Convolution computation is then performed on the segmentation image by using a kernel of a trained classification model to predict a location of a next pixel connected to a current pixel in the respective segmentation line within the segmentation image, in which the pixel to be predicted is limited to a neighboring pixel of the current pixel in a prediction direction. The predicted pixels are connected to form a compensated segmentation line for each segmentation line.

Abstract: An image processing method is provided. In the image processing method, a plurality of image quality assessment methods are received, a total execution time of a cloud server for executing image quality assessment methods is calculated, and individual execution times of an image processing device for executing image quality assessment are respectively calculated. The image quality assessment methods are classified into first type of image quality assessment methods and second type of image quality assessment methods according to the individual execution times and the total execution time. The first type of image quality assessment method are performed by the image processing device, and the second type of image quality assessment method are performed by the cloud server. Accordingly, an efficiency of image recognition may be improved by the image processing method classifying the image quality assessment methods.

Abstract: A remote diagnosis and treatment system and a remote diagnosis and treatment method are provided. The remote diagnosis and treatment method includes: establishing an online clinic room and adding a first terminal device and a second terminal device to the online clinic room by a server; in response to receiving a first end signal from the first terminal device, removing the first terminal device from the online clinic room, adding a patient corresponding to the second terminal device to a service-waiting queue, and transmitting the service-waiting queue to a third terminal device by the server; transmitting a first join signal corresponding to the second terminal device to the server according to the service-waiting queue by the third terminal device; and adding the third terminal device to the online clinic room according to the first join signal by the server.

Abstract: A physiological status evaluation method and a physiological status evaluation apparatus are provided. The method includes the following: obtaining original electrocardiogram data of a user by an electrocardiogram detection apparatus; converting the original electrocardiogram data into digital integration data; obtaining a plurality of physiological characteristic parameters according to the digital integration data; filtering the physiological characteristic parameters for at least one notable characteristic parameter through at least one filter model, where decision importance of the at least one notable characteristic parameter in a decision process of the at least one filter model is greater than a threshold; building a prediction model according to the at least one notable characteristic parameter; and evaluating a physiological status of the user through the prediction model.

Abstract: An electronic device and a method of training a classification model for age-related macular degeneration (AMD) are provided. The method includes the following steps. Training data is obtained. A loss function vector corresponding to the training data is calculated based on a machine learning algorithm, in which the loss function vector includes a first loss function value corresponding to a first stage of AMD and a second loss function value corresponding to a second stage of AMD. A first penalty weight is generated according to a stage difference between the first stage and the second stage. The first loss function value is updated according to the second loss function value and the first penalty weight, so as to generate an updated loss function vector. The classification model is trained according to the updated loss function vector.

Abstract: An electronic device and a method of training a classification model for age-related macular degeneration (AMD) are provided. The method includes the following steps. Training data is obtained. A loss function vector corresponding to the training data is calculated based on a machine learning algorithm, in which the loss function vector includes a first loss function value corresponding to a first classification of AMD and a second loss function value corresponding to a second classification of AMD, the first classification corresponds to a first group, and the second classification corresponds to one of the first group and a second group. The first loss function value is updated according to the second loss function value and a group penalty weight in response to the second classification corresponding to the second group to generate an updated loss function vector. The classification model is trained according to the updated loss function vector.

Abstract: An image analyzation method and an image analyzation device are disclosed. The method includes: obtaining a first image which presents at least a first object and a second object; analyzing the first image to detect a first central point between a first endpoint of the first object and a second endpoint of the second object; determining a target region based on the first central point as a center of the target region; capturing a second image located in the target region from the first image; and analyzing the second image to generate status information which reflects a gap status between the first object and the second object.

Abstract: A classification method and a classification device for classifying a level of an age-related macular degeneration are provided. The classification method includes the following. An object detection model and a first classification model are pre-stored. A fundus image is obtained. A bounding box is generated in the fundus image according to a macula in the fundus image detected by the object detection model. An intersection over union between a predetermined area and the bounding box in the fundus image is calculated. A classification of the fundus image is generated according to the first classification model in response to the intersection over union being greater than a threshold.

Abstract: Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for computer aided diagnosis of a medical image. One of the methods includes processing the medical image through a machine learning (ML) model to provide a first feature representation of the medical image, wherein the ML model includes an input layer and a pooling layer, wherein the first feature representation is an output from the pooling layer, generating, by the ML model, a sequence of second feature representations of the medical image from the first feature representation of the medical image, wherein each second feature representation in the sequence of the second feature representations having a lower dimension than the first feature representation, and generating, by the ML model, an output as a last second feature representation in the sequence of the second feature representations.

Abstract: An image pre-processing method and an image processing apparatus for a fundoscopic image are provided. A region of interest (ROI) is obtained from a fundoscopic image to generate a first image. The ROI is focused on an eyeball in the fundoscopic image. A smoothing process is performed on the first image to generate a second image. A value difference between neighboring pixels in the second image is increased to generate a third image.