CN113936172A - Disease classification method and device based on ensemble learning and multi-mode feature fusion - Google Patents
Disease classification method and device based on ensemble learning and multi-mode feature fusion Download PDFInfo
- Publication number
- CN113936172A CN113936172A CN202111160082.2A CN202111160082A CN113936172A CN 113936172 A CN113936172 A CN 113936172A CN 202111160082 A CN202111160082 A CN 202111160082A CN 113936172 A CN113936172 A CN 113936172A
- Authority
- CN
- China
- Prior art keywords
- feature
- image
- mode
- ensemble learning
- model
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
- G06F18/24155—Bayesian classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a disease classification method and equipment based on ensemble learning and multi-mode feature fusion, wherein the classification method comprises the following steps: obtaining a multi-modal magnetic resonance image and clinical text information of the same object on the same device, and performing data preprocessing; extracting a full brain morphological characteristic diagram under each mode from the preprocessed multi-mode magnetic resonance image, correspondingly calculating an image characteristic value in an interested brain region, and simultaneously extracting clinical text information with inter-group difference to form a text characteristic value; taking the image characteristic value and the text characteristic value of each mode as the input of a corresponding optimal base classifier to obtain a plurality of rough classification results; and fusing a plurality of coarse classification results to obtain a final classification result. Compared with the prior art, the method has the advantages of good accuracy, strong robustness and the like.
Description
Technical Field
The invention relates to the technical field of computer processing based on medical images, in particular to a disease classification method and equipment based on integrated learning and multi-mode feature fusion.
Background
With the development of medical Imaging technology, data detection in a variety of Imaging modes, including Magnetic Resonance (MR) Imaging, Diffusion Tensor Imaging (DTI), and the like, to more completely reveal the situation of a sampler has become a development trend to improve detection accuracy.
The existing mining and utilization of medical imaging data is not enough, and the following defects still exist in the research on a computer-aided classification model of diseases although the research has certain achievements: (1) the characteristic mode is single. Most of the existing models are based on independent mode analysis, only the single-mode image characteristics are considered, and the classification accuracy is insufficient. (2) Only a single classifier model is used, but no case that one algorithm is always superior to other algorithms exists, different optimal classifiers are provided based on different modal characteristics, and the construction of a stable and strong-robustness Parkinson disease classification model is still to be investigated.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, provides a disease classification method and equipment based on integrated learning and fusion multi-mode characteristics with good accuracy and strong robustness, and can be applied to early-stage Parkinson disease classification.
The purpose of the invention can be realized by the following technical scheme:
a disease classification method based on ensemble learning and multi-mode feature fusion comprises the following steps:
obtaining a multi-modal magnetic resonance image and clinical text information of the same object on the same device, and performing data preprocessing;
extracting a full brain morphological characteristic diagram under each mode from the preprocessed multi-mode magnetic resonance image, correspondingly calculating an image characteristic value in an interested brain region, and simultaneously extracting clinical text information with inter-group difference to form a text characteristic value;
taking the image characteristic value and the text characteristic value of each mode as the input of a corresponding optimal base classifier to obtain a plurality of rough classification results;
and fusing a plurality of coarse classification results to obtain a final classification result.
Further, the multi-modality magnetic resonance image includes a T1 weighted magnetic resonance image, a diffusion tensor image, and a rest-state functional magnetic resonance image.
Further, the image feature value is obtained by:
and carrying out morphological analysis based on voxels on the whole-brain morphological characteristic map to obtain voxel cluster blocks capable of reflecting differences, using the voxel cluster blocks as an interested brain area, and calculating the morphological characteristic mean value of all voxels in the interested brain area as the image characteristic value.
Further, the whole brain morphology feature map includes a gray matter volume feature map, a gray matter density feature map, a white matter volume feature map, a white matter density feature map, a fractional anisotropy map, a mean diffusivity map, and local consistency indicators for low frequency amplitudes and brain regions.
Further, the clinical text information with the difference between groups is extracted by using a double-sample t-test statistic.
Further, the optimal base classifier is obtained through training, and the training process comprises the following steps:
predefining an alternative base classifier model;
initializing a hyper-parameter search space of each base classifier model, and optimizing the parameters of each base classifier model by adopting a combined cross validation grid search method;
training each base classifier model under the optimal parameters, returning the average classification accuracy of each model by using cross validation, and obtaining the loss function value of each base model based on the average classification accuracy;
and selecting the model corresponding to the minimum loss function value as the optimal base classifier.
Further, the base classifier model comprises a support vector machine algorithm, a random forest algorithm, a K neighbor algorithm, a multi-layer perceptron classifier or a naive Bayes model.
Further, integrating and learning by a meta-learner to fuse a plurality of the coarse classification results.
Further, the meta learner is a logistic regression classifier.
The present invention also provides an electronic device comprising:
one or more processors;
a memory; and
one or more programs stored in the memory, the one or more programs including instructions for performing a method of integrating multi-modal feature fusion based disease classification as described above.
Compared with the prior art, the invention has the following beneficial effects:
1. the invention provides a stacked two-stage integrated learning model based on multi-mode characteristics, which is characterized in that the characteristics of different modes are roughly classified, the automatic selection of an optimal classifier is realized according to different modal characteristics, and then the integrated learning of a meta-learner is carried out, so that the model has good self-learning capability and stable robustness, and the problems of low accuracy, poor model robustness and the like of the existing model are solved.
2. The invention can provide richer complementary features by fusing multi-mode information, can realize the high-efficiency fusion of the multi-mode features from different data structure angles and the space angle of an algorithm, and can obtain more accurate classification results.
3. The invention carries out the self-adaptive selection of the base classifier based on different modes, is beneficial to giving full play to the characteristic expression of each mode and avoids the prediction error caused by the shortage of the classifier in the use range.
4. According to the invention, through the integrated learning of the meta-learner, the classification performance complementation of the base classifier model can be realized, so that the whole model can fully learn complex and various multi-mode characteristics, and thus, the modal characteristics are efficiently fused, and the classification precision and the model generalization capability are improved.
Drawings
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a schematic diagram of an acquisition process of the optimal basis classifier according to the present invention.
Detailed Description
The invention is described in detail below with reference to the figures and specific embodiments. The present embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation manner and a specific operation process are given, but the scope of the present invention is not limited to the following embodiments.
Example 1
The embodiment provides a Disease classification method based on ensemble learning and multi-mode feature fusion, which is applied to classification of early Parkinson's Disease (PD), and comprises the following steps:
step 1: and acquiring a multi-mode magnetic resonance image and clinical text information of the same object on the same device, and preprocessing data.
The multi-modality magnetic resonance image comprises three modality image data of a T1 weighted magnetic resonance image (T1-weighted magnetic resonance Imaging, T1-WI), a Diffusion Tensor Image (DTI) and a resting state functional magnetic resonance image (rs-fMRI). The clinical text information selects the pre-clinical non-dyskinesia scoring data. And performing corresponding image preprocessing on the three-mode magnetic resonance image data, and performing data normalization preprocessing on clinical text information.
(1) The T1-WI image preprocessing comprises the following steps: manually adjusting the original points of all T1 images to a preposed joint point; dividing gray matter, white matter and cerebrospinal fluid; using DARTEL algorithm to establish group level brain template based on normal contrast group for the partitioned gray matter and white matter image; registering all gray matter and white matter images to a Montreal Neurological Institute (MNI) standard human brain template space by using a constructed template by adopting an affine transformation and bilinear interpolation algorithm, and resampling to be 1 x 1mm3(ii) a Modulating the registered image to obtain an image reflecting Gray Matter Volume (GMV) and White Matter Volume (WMV). The unmodulated images reflect Gray Matter Density (GMD) and White Matter Density (WMD) information;
(2) DTI image pre-processing includes: removing non-brain tissues; correcting eddy current to reduce the influence of blood flow on the image; and fitting a tensor model to respectively obtain a fractional anisotropy image (FA) and an average diffusivity image (MD) of the reaction water molecules, wherein the diffusion tensor information provides fine structure information in white matter tissues of the brain.
(3) rs-fMRI image preprocessing includes: removing the first 10 time points; correcting a time layer; correcting the head, and excluding the tested object with the head movement exceeding 1.5mm or the rotation angle exceeding 1.5 degrees; linear drift is eliminated, and interference influence such as breathing and heartbeat is reduced; normalizing the image to MNI space and re-sampling to 1X 1mm3(ii) a Filtering to minimize the influence of low-frequency drift and high-frequency physiological noise. And seventhly, respectively carrying out low-frequency amplitude analysis and local consistency analysis on the whole brain area to obtain an ALFF mapping map and a ReHo mapping map.
The clinical text information includes the prodromal phase characteristics common to early PD, specifically including the odor identification test (UPSIT) data of pennsylvania university to assess olfactory dysfunction, the RBD screening questionnaire (RBDSQ) reflecting RBD characteristics, the parkinsonism autonomic nerve (SCOPA-AUT) score to assess autonomic dysfunction and the depression score (GDS) non-motor characteristics common to early PD, and the like.
The method is characterized in that the clinical text information with different dimensions needs to be subjected to characteristic normalization, and the normalization processing method comprises the following steps:
x'=(x-xmin)/(xmax-xmin)
wherein x is characteristic data of different dimensions, x' is normalized data, and xmaxIs the maximum value, x, in the type of feature dataminIs the minimum value in the type of characteristic data.
Step 2: extracting a full brain morphological characteristic diagram under each mode from the preprocessed multi-mode magnetic resonance image, correspondingly calculating an image characteristic value in an interested brain region, and simultaneously extracting clinical text information with inter-group difference to form a text characteristic value.
In this embodiment, the whole brain morphology feature map includes a Gray Matter Volume (GMV) feature map, a Gray Matter Density (GMD) feature map, a White Matter Volume (WMV) feature map, and a White Matter Density (WMD) feature map obtained based on T1-WI, a fractional anisotropy map (FA) and an average diffusivity Map (MD) based on a DTI image, and a low frequency Amplitude (ALFF) and a local homogeneity index (local homogeneity, ReHo) based on an rs-fMRI image.
The image characteristic value is obtained by the following method: smoothing the whole brain morphological characteristic diagram, wherein the smoothing kernel size adopts a Gaussian filter with 8mm FWHM to reduce image noise; performing voxel-based morphological analysis to obtain voxel cluster blocks capable of reflecting differences, taking the voxel cluster blocks as ROIs of a brain region of interest, calculating a morphological characteristic mean value of all voxels in the brain region of interest as an image characteristic value, and specifically: firstly, a GLM statistical analysis method based on the whole brain voxel level is adopted to mine a voxel cluster, p, with significant difference in each morphological measurement map<0.001, number of voxels>The voxel cluster of 30 is regarded as ROIs of interest; secondly, binarization processing is carried out to obtain corresponding mask images I of ROIsmask(ii) a Finally, the morphological feature mean of all voxels in the mask is calculated as the image feature value of the ROI, and the present embodiment can obtain the morphological feature values of 8 types of each object including a plurality of ROIs, that is, the image feature values.
In this embodiment, clinical text information with a difference between groups (P <0.05) is counted by using a double-sample t-test, and a text feature value is formed and incorporated into the feature set.
And step 3: and taking the image characteristic value and the text characteristic value of each mode as the input of the corresponding optimal base classifier to obtain a plurality of rough classification results.
The optimal base classifier is a model for a specific modal feature that is automatically trained and screened out. As shown in fig. 2, the adaptive selection process of the optimal base classifier model suitable for different modal features is as follows, taking M modal features as an example:
1) data set partitioning: representing raw data observations as xi∈RmN, where M represents the number of features extracted based on M-mode, n represents the number of samples used for training, and the class label is yiE {0,1}, where "0" and "1" represent the category of healthy population NC and early PD, respectively.
2) And predefining the alternative base classifier model, and selecting a support vector machine algorithm, a random forest algorithm, a K neighbor algorithm, a multi-layer perceptron classifier, a naive Bayes model and the like.
3) Initializing a hyper-parameter search space of each base classifier model, and optimizing parameters of each base classifier model by adopting a combined cross validation grid search method, specifically: let the jth hyperparameter of the ith model be sij,sijHas a value range of [ u ]ij,wij]If the ith model has h hyper-parameters, the parameter vector space of the ith model is: si=[si1,si2,...,sih]。
4) Under the optimal parameters, training each base classifier model, returning the average classification accuracy ACC of each model by using cross validation, and obtaining the loss function value of each base model based on the average classification accuracy, wherein the loss function is expressed as: and selecting the model corresponding to the minimum loss function value as the optimal base classifier.
And 4, step 4: and fusing a plurality of coarse classification results to obtain a final classification result.
The coarse classification result obtained in step 3 is recorded as:the implementation steps of the meta learner integrated learning are as follows: firstly, the coarse classification results of the optimal base classifier in each mode of the previous layer are fused in a linear splicing mode to obtain the input characteristics of the next-level meta classifier, and the input characteristics are recorded asAnd then, performing ensemble learning on the classification result of each upper-level base model by meta-learning to finish secondary classification and obtain a final classification result. The meta learner may be selected as a logistic regression classifier.
Example 2
The embodiment also provides a disease classification device based on integrated learning and fusion of multi-mode features, which comprises a data screening and preprocessing module, a feature extraction module, an adaptive base classifier selection module and a meta-learner integrated learning module, wherein the data screening and preprocessing module is used for acquiring a T1 weighted magnetic resonance image, a diffusion tensor image, a resting state functional magnetic resonance image and a clinical early-stage non-dyskinesia score, respectively performing corresponding image preprocessing on three-mode magnetic resonance image data, and performing data normalization preprocessing on clinical text information; the feature extraction module is used for extracting a full brain morphological feature map of each modal image, calculating a morphological feature value in an interested brain region, and counting clinical text information with inter-group difference by adopting double-sample t test to be included in a feature set; the self-adaptive base classifier selection module inputs each modal characteristic into a self-adaptive optimal base model respectively for classification to obtain a coarse classification result; and the meta-learner integrated learning module is used for integrally learning the rough prediction results of the base models and outputting the final classification results. The rest is the same as example 1.
Example 3
The present invention also provides an electronic device comprising one or more processors, memory, and one or more programs stored in the memory, the one or more programs including instructions for performing the method for integrating multi-modal feature fusion based disease classification as described in embodiment 1.
The foregoing detailed description of the preferred embodiments of the invention has been presented. It should be understood that numerous modifications and variations could be devised by those skilled in the art in light of the present teachings without departing from the inventive concepts. Therefore, the technical solutions available to those skilled in the art through logic analysis, reasoning and limited experiments based on the prior art according to the concept of the present invention should be within the scope of protection defined by the claims.
Claims (10)
1. A disease classification method based on ensemble learning and multi-mode feature fusion is characterized by comprising the following steps:
obtaining a multi-modal magnetic resonance image and clinical text information of the same object on the same device, and performing data preprocessing;
extracting a full brain morphological characteristic diagram under each mode from the preprocessed multi-mode magnetic resonance image, correspondingly calculating an image characteristic value in an interested brain region, and simultaneously extracting clinical text information with inter-group difference to form a text characteristic value;
taking the image characteristic value and the text characteristic value of each mode as the input of a corresponding optimal base classifier to obtain a plurality of rough classification results;
and fusing a plurality of coarse classification results to obtain a final classification result.
2. The ensemble learning fusion multi-modal feature based disease classification method according to claim 1, wherein the multi-modal magnetic resonance images include T1 weighted magnetic resonance images, diffusion tensor images, and rest state functional magnetic resonance images.
3. The method according to claim 1, wherein the image feature value is obtained by:
and carrying out morphological analysis based on voxels on the whole-brain morphological characteristic map to obtain voxel cluster blocks capable of reflecting differences, using the voxel cluster blocks as an interested brain area, and calculating the morphological characteristic mean value of all voxels in the interested brain area as the image characteristic value.
4. The ensemble learning fusion multimodal feature based disease classification method according to claim 1 or 3, characterized in that the whole brain morphology feature map comprises gray matter volume feature map, gray matter density feature map, white matter volume feature map, white matter density feature map, fractional anisotropy map, mean diffusivity map and local consistency index of low frequency amplitude and brain region.
5. The ensemble learning fused multimodal feature based disease classification method according to claim 1, wherein the clinical text information with the difference between groups is extracted using a two-sample t-test statistic.
6. The method for classifying diseases based on ensemble learning and multi-modal feature fusion according to claim 1, wherein the optimal basis classifier is obtained by training, and the training process comprises:
predefining an alternative base classifier model;
initializing a hyper-parameter search space of each base classifier model, and optimizing the parameters of each base classifier model by adopting a combined cross validation grid search method;
training each base classifier model under the optimal parameters, returning the average classification accuracy of each model by using cross validation, and obtaining the loss function value of each base model based on the average classification accuracy;
and selecting the model corresponding to the minimum loss function value as the optimal base classifier.
7. The ensemble learning fusion multi-modal feature based disease classification method according to claim 6, wherein the base classifier model comprises a support vector machine algorithm, a random forest algorithm, a K-nearest neighbor algorithm, a multi-layer perceptron classifier, or a naive bayes model.
8. The ensemble learning-fused multi-modal feature-based disease classification method according to claim 1, wherein a plurality of the coarse classification results are fused by ensemble learning through a meta-learner.
9. The ensemble learning-fused-multimodal-feature-based disease classification method according to claim 8, wherein the meta-learner is a logistic regression classifier.
10. An electronic device, comprising:
one or more processors;
a memory; and
one or more programs stored in the memory, the one or more programs including instructions for performing the method for integrating learning and fusing multi-modal feature-based disease classification of any of claims 1-9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111160082.2A CN113936172A (en) | 2021-09-30 | 2021-09-30 | Disease classification method and device based on ensemble learning and multi-mode feature fusion |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111160082.2A CN113936172A (en) | 2021-09-30 | 2021-09-30 | Disease classification method and device based on ensemble learning and multi-mode feature fusion |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113936172A true CN113936172A (en) | 2022-01-14 |
Family
ID=79277581
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111160082.2A Pending CN113936172A (en) | 2021-09-30 | 2021-09-30 | Disease classification method and device based on ensemble learning and multi-mode feature fusion |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113936172A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117393167A (en) * | 2023-12-11 | 2024-01-12 | 中国人民解放军军事科学院军事医学研究院 | Brain health assessment method based on artificial intelligence model |
CN117789988A (en) * | 2024-02-27 | 2024-03-29 | 首都医科大学附属北京友谊医院 | Method for training predictive model for predicting parkinsonism and related products |
-
2021
- 2021-09-30 CN CN202111160082.2A patent/CN113936172A/en active Pending
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117393167A (en) * | 2023-12-11 | 2024-01-12 | 中国人民解放军军事科学院军事医学研究院 | Brain health assessment method based on artificial intelligence model |
CN117393167B (en) * | 2023-12-11 | 2024-03-12 | 中国人民解放军军事科学院军事医学研究院 | Brain health assessment method based on artificial intelligence model |
CN117789988A (en) * | 2024-02-27 | 2024-03-29 | 首都医科大学附属北京友谊医院 | Method for training predictive model for predicting parkinsonism and related products |
CN117789988B (en) * | 2024-02-27 | 2024-06-11 | 首都医科大学附属北京友谊医院 | Method for training predictive model for predicting parkinsonism and related products |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Ortiz et al. | Parkinson's disease detection using isosurfaces-based features and convolutional neural networks | |
Ashtari-Majlan et al. | A multi-stream convolutional neural network for classification of progressive MCI in Alzheimer’s disease using structural MRI images | |
Duraisamy et al. | Alzheimer disease detection from structural MR images using FCM based weighted probabilistic neural network | |
Cover et al. | Computational methods for corpus callosum segmentation on MRI: a systematic literature review | |
Attallah | CoMB-deep: composite deep learning-based pipeline for classifying childhood medulloblastoma and its classes | |
Zhang et al. | Structured learning for 3-D perivascular space segmentation using vascular features | |
Baskar et al. | An efficient classification approach for detection of Alzheimer’s disease from biomedical imaging modalities | |
CN113936172A (en) | Disease classification method and device based on ensemble learning and multi-mode feature fusion | |
Zhu et al. | Discovering dense and consistent landmarks in the brain | |
Algorri et al. | Classification of anatomical structures in MR brain images using fuzzy parameters | |
Amiri et al. | 3D cerebral MR image segmentation using multiple-classifier system | |
CN115484864A (en) | Brain function connection correlation value clustering device, brain function connection correlation value clustering system, brain function connection correlation value clustering method, brain function connection correlation value classifier program, brain activity marker classification system, and brain function connection correlation value clustering classifier model | |
Martins et al. | An adaptive probabilistic atlas for anomalous brain segmentation in MR images | |
CN115457334A (en) | Resting state functional magnetic resonance image data classification method and system and related products | |
CN108805181B (en) | Image classification device and method based on multi-classification model | |
S U et al. | A Review on state-of-the-art techniques for image segmentation and classification for brain MR images | |
Pallawi et al. | Study of Alzheimer’s disease brain impairment and methods for its early diagnosis: a comprehensive survey | |
Rudhra et al. | A Novel deep learning approach for the automated diagnosis of normal pressure hydrocephalus | |
Baik | Class imbalance learning–driven Alzheimer’s detection using hybrid features | |
CN114418966A (en) | Alzheimer disease risk assessment method and device | |
Young et al. | Classification of Alzheimer's disease patients with hippocampal shape wrapper-based feature selection and support vector machine | |
Ghani | On forecasting lung cancer patients’ survival rates using 3D feature engineering | |
CN114926477B (en) | Brain tumor multi-mode MRI image segmentation method based on deep learning | |
CN117934981B (en) | Neurology symptom identification method and system based on multi-mode information | |
Avci et al. | Case study: deep convolutional networks in healthcare |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination |