In the extraction of fetal electrocardiogram (ECG) signal, due to the unicity of the scale of the U-Net same-level convolution encoder, the size and shape difference of the ECG characteristic wave between mother and fetus are ignored, and the time information of ECG signals is not used in the threshold learning process of the encoder’s residual shrinkage module. In this paper, a method of extracting fetal ECG signal based on multi-scale residual shrinkage U-Net model is proposed. First, the Inception and time domain attention were introduced into the residual shrinkage module to enhance the multi-scale feature extraction ability of the same level convolution encoder and the utilization of the time domain information of fetal ECG signal. In order to maintain more local details of ECG waveform, the maximum pooling in U-Net was replaced by Softpool. Finally, the decoder composed of the residual module and up-sampling gradually generated fetal ECG signals. In this paper, clinical ECG signals were used for experiments. The final results showed that compared with other fetal ECG extraction algorithms, the method proposed in this paper could extract clearer fetal ECG signals. The sensitivity, positive predictive value, and F1 scores in the 2013 competition data set reached 93.33%, 99.36%, and 96.09%, respectively, indicating that this method can effectively extract fetal ECG signals and has certain application values for perinatal fetal health monitoring.
The synergistic effect of drug combinations can solve the problem of acquired resistance to single drug therapy and has great potential for the treatment of complex diseases such as cancer. In this study, to explore the impact of interactions between different drug molecules on the effect of anticancer drugs, we proposed a Transformer-based deep learning prediction model—SMILESynergy. First, the drug text data—simplified molecular input line entry system (SMILES) were used to represent the drug molecules, and drug molecule isomers were generated through SMILES Enumeration for data augmentation. Then, the attention mechanism in the Transformer was used to encode and decode the drug molecules after data augmentation, and finally, a multi-layer perceptron (MLP) was connected to obtain the synergy value of the drugs. Experimental results showed that our model had a mean squared error of 51.34 in regression analysis, an accuracy of 0.97 in classification analysis, and better predictive performance than the DeepSynergy and MulinputSynergy models. SMILESynergy offers improved predictive performance to assist researchers in rapidly screening optimal drug combinations to improve cancer treatment outcomes.
With the development of artificial intelligence, machine learning has been widely used in diagnosis of diseases. It is crucial to conduct diagnostic test accuracy studies and evaluate the performance of models reasonably to improve the accuracy of diagnosis. For machine learning-based diagnostic test accuracy studies, this paper introduces the principles of study design in the aspects of target conditions, selection of participants, diagnostic tests, reference standards and ethics.
Pathological images of gastric cancer serve as the gold standard for diagnosing this malignancy. However, the recurrence prediction task often encounters challenges such as insignificant morphological features of the lesions, insufficient fusion of multi-resolution features, and inability to leverage contextual information effectively. To address these issues, a three-stage recurrence prediction method based on pathological images of gastric cancer is proposed. In the first stage, the self-supervised learning framework SimCLR was adopted to train low-resolution patch images, aiming to diminish the interdependence among diverse tissue images and yield decoupled enhanced features. In the second stage, the obtained low-resolution enhanced features were fused with the corresponding high-resolution unenhanced features to achieve feature complementation across multiple resolutions. In the third stage, to address the position encoding difficulty caused by the large difference in the number of patch images, we performed position encoding based on multi-scale local neighborhoods and employed self-attention mechanism to obtain features with contextual information. The resulting contextual features were further combined with the local features extracted by the convolutional neural network. The evaluation results on clinically collected data showed that, compared with the best performance of traditional methods, the proposed network provided the best accuracy and area under curve (AUC), which were improved by 7.63% and 4.51%, respectively. These results have effectively validated the usefulness of this method in predicting gastric cancer recurrence.
[Abstract]Automatic and accurate segmentation of lung parenchyma is essential for assisted diagnosis of lung cancer. In recent years, researchers in the field of deep learning have proposed a number of improved lung parenchyma segmentation methods based on U-Net. However, the existing segmentation methods ignore the complementary fusion of semantic information in the feature map between different layers and fail to distinguish the importance of different spaces and channels in the feature map. To solve this problem, this paper proposes the double scale parallel attention (DSPA) network (DSPA-Net) architecture, and introduces the DSPA module and the atrous spatial pyramid pooling (ASPP) module in the “encoder-decoder” structure. Among them, the DSPA module aggregates the semantic information of feature maps of different levels while obtaining accurate space and channel information of feature map with the help of cooperative attention (CA). The ASPP module uses multiple parallel convolution kernels with different void rates to obtain feature maps containing multi-scale information under different receptive fields. The two modules address multi-scale information processing in feature maps of different levels and in feature maps of the same level, respectively. We conducted experimental verification on the Kaggle competition dataset. The experimental results prove that the network architecture has obvious advantages compared with the current mainstream segmentation network. The values of dice similarity coefficient (DSC) and intersection on union (IoU) reached 0.972 ± 0.002 and 0.945 ± 0.004, respectively. This paper achieves automatic and accurate segmentation of lung parenchyma and provides a reference for the application of attentional mechanisms and multi-scale information in the field of lung parenchyma segmentation.
Among numerous medical imaging modalities, diffusion weighted imaging (DWI) is extremely sensitive to acute ischemic stroke lesions, especially small infarcts. However, magnetic resonance imaging is time-consuming and expensive, and it is also prone to interference from metal implants. Therefore, the aim of this study is to design a medical image synthesis method based on generative adversarial network, Stroke-p2pHD, for synthesizing DWI images from computed tomography (CT). Stroke-p2pHD consisted of a generator that effectively fused local image features and global context information (Global_to_Local) and a multi-scale discriminator (M2Dis). Specifically, in the Global_to_Local generator, a fully convolutional Transformer (FCT) and a local attention module (LAM) were integrated to achieve the synthesis of detailed information such as textures and lesions in DWI images. In the M2Dis discriminator, a multi-scale convolutional network was adopted to perform the discrimination function of the input images. Meanwhile, an optimization balance with the Global_to_Local generator was ensured and the consistency of features in each layer of the M2Dis discriminator was constrained. In this study, the public Acute Ischemic Stroke Dataset (AISD) and the acute cerebral infarction dataset from Yantaishan Hospital were used to verify the performance of the Stroke-p2pHD model in synthesizing DWI based on CT. Compared with other methods, the Stroke-p2pHD model showed excellent quantitative results (mean-square error = 0.008, peak signal-to-noise ratio = 23.766, structural similarity = 0.743). At the same time, relevant experimental analyses such as computational efficiency verify that the Stroke-p2pHD model has great potential for clinical applications.
Temporomandibular joint disorder (TMD) is a common oral and maxillofacial disease, which is difficult to detect due to its subtle early symptoms. In this study, a TMD intelligent diagnostic system implemented on edge computing devices was proposed, which can achieve rapid detection of TMD in clinical diagnosis and facilitate its early-stage clinical intervention. The proposed system first automatically segments the important components of the temporomandibular joint, followed by quantitative measurement of the joint gap area, and finally predicts the existence of TMD according to the measurements. In terms of segmentation, this study employs semi-supervised learning to achieve the accurate segmentation of temporomandibular joint, with an average Dice coefficient (DC) of 0.846. A 3D region extraction algorithm for the temporomandibular joint gap area is also developed, based on which an automatic TMD diagnosis model is proposed, with an accuracy of 83.87%. In summary, the intelligent TMD diagnosis system developed in this paper can be deployed at edge computing devices within a local area network, which is able to achieve rapid detecting and intelligent diagnosis of TMD with privacy guarantee.
Objective To develop a deep learning system for CT images to assist in the diagnosis of thoracolumbar fractures and analyze the feasibility of its clinical application. Methods Collected from West China Hospital of Sichuan University from January 2019 to March 2020, a total of 1256 CT images of thoracolumbar fractures were annotated with a unified standard through the Imaging LabelImg system. All CT images were classified according to the AO Spine thoracolumbar spine injury classification. The deep learning system in diagnosing ABC fracture types was optimized using 1039 CT images for training and validation, of which 1004 were used as the training set and 35 as the validation set; the rest 217 CT images were used as the test set to compare the deep learning system with the clinician’s diagnosis. The deep learning system in subtyping A was optimized using 581 CT images for training and validation, of which 556 were used as the training set and 25 as the validation set; the rest 104 CT images were used as the test set to compare the deep learning system with the clinician’s diagnosis. Results The accuracy and Kappa coefficient of the deep learning system in diagnosing ABC fracture types were 89.4% and 0.849 (P<0.001), respectively. The accuracy and Kappa coefficient of subtyping A were 87.5% and 0.817 (P<0.001), respectively. Conclusions The classification accuracy of the deep learning system for thoracolumbar fractures is high. This approach can be used to assist in the intelligent diagnosis of CT images of thoracolumbar fractures and improve the current manual and complex diagnostic process.
ObjectivesTo explore the effect of the deep learning algorithm convolutional neural network (CNN) in screening of randomized controlled trials (RCTs) in Chinese medical literatures.MethodsLiterature with the topic " oral science” published in 2014 were retrieved from CNKI and exported citations containing title and abstract. RCTs screening was conducted by double independent screening, checking and peer discussion. The final results of the citations were used for CNN algorithm model training. After completing the algorithm model training, a prospective comparative trial was organized by searching all literature with the topic "oral science" published in CNKI from January to March 2018 to compare the sensitivity (SEN) and specificity (SPE) of algorithm with manual screening. The initial results of a single screener represented the performance of manual screening, and the final results after peer discussion were used as the gold standard. The best thresholds of algorithm were determined with the receptor operative characteristic (ROC) curve.ResultsA total of 1 246 RCTs and 4 754 non-RCTs were eventually included for training and testing of CNN algorithm model. 249 RCTs and 949 non-RCTs were included in the prospective trial. The SEN and SPE of manual screening were 98.01% and 98.82%. For the algorithm model, the SEN of RCTs screening decreased with the increase of threshold value while the SPE increased with the increase of threshold value. After 27 changes of threshold value, ROC curve were obtained. The area under the ROC curve was 0.9977, unveiling the optimal accuracy threshold (Threshold=0.4, SEN=98.39%, SPE=98.84%) and high sensitivity threshold (Threshold=0.06, SEN=99.60%, SPE=94.10%).ConclusionsA CNN algorithm model is trained with Chinese RCTs classification database established in this study and shows an excellent classification performance in screening RCTs of Chinese medical literature, which is proved to be comparable to the manual screening performance in the prospective controlled trial.
Objective To construct and evaluate a screening and diagnostic system based on color fundus images and artificial intelligence (AI)-assisted screening for optic neuritis (ON) and non-arteritic anterior ischemic optic neuropathy (NAION). MethodsA diagnostic test study. From 2016 to 2020, 178 cases 267 eyes of NAION patients (NAION group) and 204 cases 346 eyes of ON patients (ON group) were examined and diagnosed in Zhongshan Ophthalmic Center of Sun Yat-sen University; 513 healthy individuals of 1 160 eyes (the normal control group) with normal fundus by visual acuity, intraocular pressure and optical coherence tomography examination were collected from 2018 to 2020. All 2 909 color fundus images were as the data set of the screening and diagnosis system, including 730, 805, and 1 374 images for the NAION group, ON group, and normal control group, respectively. The correctly labeled color fundus images were used as input data, and the EfficientNet-B0 algorithm was selected for model training and validation. Finally, three systems for screening abnormal optic discs, ON, and NAION were constructed. The subject operating characteristic (ROC) curve, area under the ROC (AUC), accuracy, sensitivity, specificity, and heat map were used as indicators of diagnostic efficacy. ResultsIn the test data set, the AUC for diagnosing the presence of an abnormal optic disc, the presence of ON, and the presence of NAION were 0.967 [95% confidence interval (CI) 0.947-0.980], 0.964 (95%CI 0.938-0.979), and 0.979 (95%CI 0.958-0.989), respectively. The activation area of the systems were mainly located in the optic disc area in the decision-making process. ConclusionAbnormal optic disc, ON and NAION, and screening diagnostic systems based on color fundus images have shown accurate and efficient diagnostic performance.