RESUMEN
Skin cancer stands as one of the foremost challenges in oncology, with its early detection being crucial for successful treatment outcomes. Traditional diagnostic methods depend on dermatologist expertise, creating a need for more reliable, automated tools. This study explores deep learning, particularly Convolutional Neural Networks (CNNs), to enhance the accuracy and efficiency of skin cancer diagnosis. Leveraging the HAM10000 dataset, a comprehensive collection of dermatoscopic images encompassing a diverse range of skin lesions, this study introduces a sophisticated CNN model tailored for the nuanced task of skin lesion classification. The model's architecture is intricately designed with multiple convolutional, pooling, and dense layers, aimed at capturing the complex visual features of skin lesions. To address the challenge of class imbalance within the dataset, an innovative data augmentation strategy is employed, ensuring a balanced representation of each lesion category during training. Furthermore, this study introduces a CNN model with optimized layer configuration and data augmentation, significantly boosting diagnostic precision in skin cancer detection. The model's learning process is optimized using the Adam optimizer, with parameters fine-tuned over 50 epochs and a batch size of 128 to enhance the model's ability to discern subtle patterns in the image data. A Model Checkpoint callback ensures the preservation of the best model iteration for future use. The proposed model demonstrates an accuracy of 97.78% with a notable precision of 97.9%, recall of 97.9%, and an F2 score of 97.8%, underscoring its potential as a robust tool in the early detection and classification of skin cancer, thereby supporting clinical decision-making and contributing to improved patient outcomes in dermatology.
Asunto(s)
Aprendizaje Profundo , Dermoscopía , Redes Neurales de la Computación , Neoplasias Cutáneas , Humanos , Neoplasias Cutáneas/diagnóstico por imagen , Neoplasias Cutáneas/patología , Dermoscopía/métodos , Interpretación de Imagen Asistida por Computador/métodosRESUMEN
Event of the disastrous scenarios are actively discussed on microblogging platforms like Twitter which can lead to chaotic situations. In the era of machine learning and deep learning, these chaotic situations can be effectively controlled by developing efficient methods and models that can assist in classifying real and fake tweets. In this research article, an efficient method named BERT Embedding based CNN model with RMSProp Optimizer is proposed to effectively classify the tweets related disastrous scenario. Tweet classification is carried out via some of the popular the machine learning algorithms such as logistic regression and decision tree classifiers. Noting the low accuracy of machine learning models, Convolutional Neural Network (CNN) based deep learning model is selected as the primary classification method. CNNs performance is improved via optimization of the parameters with gradient based optimizers. To further elevate accuracy and to capture contextual semantics from the text data, BERT embeddings are included in the proposed model. The performance of proposed method - BERT Embedding based CNN model with RMSProp Optimizer achieved an F1 score of 0.80 and an Accuracy of 0.83. The methodology presented in this research article is comprised of the following key contributions:â¢Identification of suitable text classification model that can effectively capture complex patterns when dealing with large vocabularies or nuanced language structures in disaster management scenarios.â¢The method explores the gradient based optimization techniques such as Adam Optimizer, Stochastic Gradient Descent (SGD) Optimizer, AdaGrad, and RMSprop Optimizer to identify the most appropriate optimizer that meets the characteristics of the dataset and the CNN model architecture.â¢"BERT Embedding based CNN model with RMSProp Optimizer" - a method to classify the disaster tweets and capture semantic representations by leveraging BERT embeddings with appropriate feature selection is presented and models are validated with appropriate comparative analysis.
RESUMEN
Rare genetic diseases are difficult to diagnose and this translates in patient's diagnostic odyssey! This is particularly true for more than 900 rare diseases including orodental developmental anomalies such as missing teeth. However, if left untreated, their symptoms can become significant and disabling for the patient. Early detection and rapid management are therefore essential in this context. The i-Dent project aims to supply a pre-diagnostic tool to detect rare diseases with tooth agenesis of varying severity and pattern. To identify missing teeth, image segmentation models (Mask R-CNN, U-Net) have been trained for the automatic detection of teeth on patients' panoramic dental X-rays. Teeth segmentation enables the identification of teeth which are present or missing within the mouth. Furthermore, a dental age assessment is conducted to verify whether the absence of teeth is an anomaly or a characteristic of the patient's age. Due to the small size of our dataset, we developed a new dental age assessment technique based on the tooth eruption rate. Information about missing teeth is then used by a final algorithm based on the agenesis probabilities to propose a pre-diagnosis of a rare disease. The results obtained in detecting three types of genes (PAX9, WNT10A and EDA) by our system are very promising, providing a pre-diagnosis with an average accuracy of 72 %.
RESUMEN
Hippocampal atrophy (tissue loss) has become a fundamental outcome parameter in clinical trials on Alzheimer's disease. To accurately estimate hippocampus volume and track its volume loss, a robust and reliable segmentation is essential. Manual hippocampus segmentation is considered the gold standard but is extensive, time-consuming, and prone to rater bias. Therefore, it is often replaced by automated programs like FreeSurfer, one of the most commonly used tools in clinical research. Recently, deep learning-based methods have also been successfully applied to hippocampus segmentation. The basis of all approaches are clinically used T1-weighted whole-brain MR images with approximately 1 mm isotropic resolution. However, such T1 images show low contrast-to-noise ratios (CNRs), particularly for many hippocampal substructures, limiting delineation reliability. To overcome these limitations, high-resolution T2-weighted scans are suggested for better visualization and delineation, as they show higher CNRs and usually allow for higher resolutions. Unfortunately, such time-consuming T2-weighted sequences are not feasible in a clinical routine. We propose an automated hippocampus segmentation pipeline leveraging deep learning with T2-weighted MR images for enhanced hippocampus segmentation of clinical T1-weighted images based on a series of 3D convolutional neural networks and a specifically acquired multi-contrast dataset. This dataset consists of corresponding pairs of T1- and high-resolution T2-weighted images, with the T2 images only used to create more accurate manual ground truth annotations and to train the segmentation network. The T2-based ground truth labels were also used to evaluate all experiments by comparing the masks visually and by various quantitative measures. We compared our approach with four established state-of-the-art hippocampus segmentation algorithms (FreeSurfer, ASHS, HippoDeep, HippMapp3r) and demonstrated a superior segmentation performance. Moreover, we found that the automated segmentation of T1-weighted images benefits from the T2-based ground truth data. In conclusion, this work showed the beneficial use of high-resolution, T2-based ground truth data for training an automated, deep learning-based hippocampus segmentation and provides the basis for a reliable estimation of hippocampal atrophy in clinical studies.
RESUMEN
Background: Electronic health records (EHRs) are live, digital patient records that provide a thorough overview of a person's complete health data. Electronic health records (EHRs) provide better healthcare decisions and evidence-based patient treatment and track patients' clinical development. The EHR offers a new range of opportunities for analyzing and contrasting exam findings and other data, creating a proper information management mechanism to boost effectiveness, quick resolutions, and identifications. Aim: The aim of this studywas to implement an interoperable EHR system to improve the quality of care through the decision support system for the identification of lung cancer in its early stages. Objective: The main objective of the proposed system was to develop an Android application for maintaining an EHR system and decision support system using deep learning for the early detection of diseases. The second objective was to study the early stages of lung disease to predict/detect it using a decision support system. Methods: To extract the EHR data of patients, an android application was developed. The android application helped in accumulating the data of each patient. The accumulated data were used to create a decision support system for the early prediction of lung cancer. To train, test, and validate the prediction of lung cancer, a few samples from the ready dataset and a few data from patients were collected. The valid data collection from patients included an age range of 40 to 70, and both male and female patients. In the process of experimentation, a total of 316 images were considered. The testing was done by considering the data set into 80:20 partitions. For the evaluation purpose, a manual classification was done for 3 different diseases, such as large cell carcinoma, adenocarcinoma, and squamous cell carcinoma diseases in lung cancer detection. Results: The first model was tested for interoperability constraints of EHR with data collection and updations. When it comes to the disease detection system, lung cancer was predicted for large cell carcinoma, adenocarcinoma, and squamous cell carcinoma type by considering 80:20 training and testing ratios. Among the considered 336 images, the prediction of large cell carcinoma was less compared to adenocarcinoma and squamous cell carcinoma. The analysis also showed that large cell carcinoma occurred majorly in males due to smoking and was found as breast cancer in females. Conclusion: As the challenges are increasing daily in healthcare industries, a secure, interoperable EHR could help patients and doctors access patient data efficiently and effectively using an Android application. Therefore, a decision support system using a deep learning model was attempted and successfully used for disease detection. Early disease detection for lung cancer was evaluated, and the model achieved an accuracy of 93%. In future work, the integration of EHR data can be performed to detect various diseases early.
RESUMEN
Fine particulate matter (PM2.5) is a major air pollutant affecting human survival, development and health. By predicting the spatial distribution concentration of PM2.5, pollutant sources can be better traced, allowing measures to protect human health to be implemented. Thus, the purpose of this study is to predict and analyze the PM2.5 concentration of stations based on the integrated deep learning of a convolutional neural network long short-term memory (CNN-LSTM) model. To solve the complexity and nonlinear characteristics of PM2.5 time series data problems, we adopted the CNN-LSTM deep learning model. We collected the PM2.5data of Qingdao in 2020 as well as meteorological factors such as temperature, wind speed and air pressure for pre-processing and characteristic analysis. Then, the CNN-LSTM deep learning model was integrated to capture the temporal and spatial features and trends in the data. The CNN layer was used to extract spatial features, while the LSTM layer was used to learn time dependencies. Through comparative experiments and model evaluation, we found that the CNN-LSTM model can achieve excellent PM2.5 prediction performance. The results show that the coefficient of determination (R2) is 0.91, and the root mean square error (RMSE) is 8.216 µg/m3. The CNN-LSTM model achieves better prediction accuracy and generalizability compared with those of the CNN and LSTM models (R2 values of 0.85 and 0.83, respectively, and RMSE values of 11.356 and 14.367, respectively). Finally, we analyzed and explained the predicted results. We also found that some meteorological factors (such as air temperature, pressure, and wind speed) have significant effects on the PM2.5 concentration at ground stations in Qingdao. In summary, by using deep learning methods, we obtained better prediction performance and revealed the association between PM2.5 concentration and meteorological factors. These findings are of great significance for improving the quality of the atmospheric environment and protecting public health.
Asunto(s)
Contaminantes Atmosféricos , Redes Neurales de la Computación , Material Particulado , Material Particulado/análisis , Material Particulado/efectos adversos , Contaminantes Atmosféricos/análisis , Contaminantes Atmosféricos/efectos adversos , Humanos , Monitoreo del Ambiente/métodos , Aprendizaje Profundo , China , Algoritmos , Contaminación del Aire/análisisRESUMEN
Beef is an important food product in human nutrition. The evaluation of the quality and safety of this food product is a matter that needs attention. Non-destructive determination of beef quality by image processing methods shows great potential for food safety, as it helps prevent wastage. Traditionally, beef quality determination by image processing methods has been based on handcrafted color features. It is, however, difficult to determine meat quality based on the color space model alone. This study introduces an effective beef quality classification approach by concatenating learning-based global and handcrafted color features. According to experimental results, the convVGG16 + HLS + HSV + RGB + Bi-LSTM model achieved high performance values. This model's accuracy, precision, recall, F1-score, AUC, Jaccard index, and MCC values were 0.989, 0.990, 0.989, 0.990, 0.992, 0.979, and 0.983, respectively.
RESUMEN
With the establishment of the "double carbon" goal, various industries are actively exploring ways to reduce carbon emissions. Cloud data centers, represented by cloud computing, often have the problem of mismatch between load requests and resource supply, resulting in excessive carbon emissions. Based on this, this paper proposes a complete method for cloud computing carbon emission prediction. Firstly, the convolutional neural network and bidirectional long-term and short-term memory neural network (CNN-BiLSTM) combined model are used to predict the cloud computing load. The real-time prediction power is obtained by real-time prediction load of cloud computing, and then the carbon emission prediction is obtained by power calculation. Develop a dynamic server carbon emission prediction model, so that the server carbon emission can change with the change of CPU utilization, so as to achieve the purpose of low carbon emission reduction. In this paper, Google cluster data is used to predict the load. The experimental results show that the CNN-BiLSTM combined model has good prediction effect. Compared with the multi-layer feed forward neural network model (BP), long short-term memory network model (LSTM ), bidirectional long short-term memory network model (BiLSTM), modal decomposition and convolution long time series neural network model (CEEMDAN-ConvLSTM), the MSE index decreased by 52 % , 50 % , 34 % and 45 % respectively.
RESUMEN
Cervical cancer is a prevalent and concerning disease affecting women, with increasing incidence and mortality rates. Early detection plays a crucial role in improving outcomes. Recent advancements in computer vision, particularly the Swin transformer, have shown promising performance in image classification tasks, rivaling or surpassing traditional convolutional neural networks (CNNs). The Swin transformer adopts a hierarchical and efficient approach using shifted windows, enabling the capture of both local and global contextual information in images. In this paper, we propose a novel approach called Swin-GA-RF to enhance the classification performance of cervical cells in Pap smear images. Swin-GA-RF combines the strengths of the Swin transformer, genetic algorithm (GA) feature selection, and the replacement of the softmax layer with a random forest classifier. Our methodology involves extracting feature representations from the Swin transformer, utilizing GA to identify the optimal feature set, and employing random forest as the classification model. Additionally, data augmentation techniques are applied to augment the diversity and quantity of the SIPaKMeD1 cervical cancer image dataset. We compare the performance of the Swin-GA-RF Transformer with pre-trained CNN models using two classes and five classes of cervical cancer classification, employing both Adam and SGD optimizers. The experimental results demonstrate that Swin-GA-RF outperforms other Swin transformers and pre-trained CNN models. When utilizing the Adam optimizer, Swin-GA-RF achieves the highest performance in both binary and five-class classification tasks. Specifically, for binary classification, it achieves an accuracy, precision, recall, and F1-score of 99.012, 99.015, 99.012, and 99.011, respectively. In the five-class classification, it achieves an accuracy, precision, recall, and F1-score of 98.808, 98.812, 98.808, and 98.808, respectively. These results underscore the effectiveness of the Swin-GA-RF approach in cervical cancer classification, demonstrating its potential as a valuable tool for early diagnosis and screening programs.
RESUMEN
Convolutional Neural Networks (CNNs) have demonstrated outstanding performance in various domains, such as face recognition, object detection, and image segmentation. However, the lack of transparency and limited interpretability inherent in CNNs pose challenges in fields such as medical diagnosis, autonomous driving, finance, and military applications. Several studies have explored the interpretability of CNNs and proposed various post-hoc interpretable methods. The majority of these methods are feature-based, focusing on the influence of input variables on outputs. Few methods undertake the analysis of parameters in CNNs and their overall structure. To explore the structure of CNNs and intuitively comprehend the role of their internal parameters, we propose an Attribution Graph-based Interpretable method for CNNs (AGIC) which models the overall structure of CNNs as graphs and provides interpretability from global and local perspectives. The runtime parameters of CNNs and feature maps of each image sample are applied to construct attribution graphs (At-GCs), where the convolutional kernels are represented as nodes and the SHAP values between kernel outputs are assigned as edges. These At-GCs are then employed to pretrain a newly designed heterogeneous graph encoder based on Deep Graph Infomax (DGI). To comprehensively delve into the overall structure of CNNs, the pretrained encoder is used for two types of interpretable tasks: (1) a classifier is attached to the pretrained encoder for the classification of At-GCs, revealing the dependency of At-GC's topological characteristics on the image sample categories, and (2) a scoring aggregation (SA) network is constructed to assess the importance of each node in At-GCs, thus reflecting the relative importance of kernels in CNNs. The experimental results indicate that the topological characteristics of At-GC exhibit a dependency on the sample category used in its construction, which reveals that kernels in CNNs show distinct combined activation patterns for processing different image categories, meanwhile, the kernels that receive high scores from SA network are crucial for feature extraction, whereas low-scoring kernels can be pruned without affecting model performance, thereby enhancing the interpretability of CNNs.
RESUMEN
Sandification can degrade the strength and quality of dolomite, and to a certain extent, compromise the stability of a tunnel's surrounding rock as an unfavorable geological boundary. Sandification degree classification of sandy dolomite is one of the non-trivial challenges faced by geotechnical engineering projects such as tunneling in complex geographical environments. The traditional methods quantitatively measuring the physical parameters or analyzing some visual features are either time-consuming or inaccurate in practical use. To address these issues, we, for the first time, introduce the convolutional neural network (CNN)-based image classification methods into dolomite sandification degree classification task. In this study, we have made a significant contribution by establishing a large-scale dataset comprising 5729 images, classified into four distinct sandification degrees of sandy dolomite. These images were collected from the vicinity of a tunnel located in the Yuxi section of the CYWD Project in China. We conducted comprehensive classification experiments using this dataset. The results of these experiments demonstrate the groundbreaking achievement of CNN-based models, which achieved an impressive accuracy rate of up to 91.4%. This accomplishment underscores the pioneering role of our work in creating this dataset and its potential for applications in complex geographical analyses.
RESUMEN
When a person listens to natural speech, the relation between features of the speech signal and the corresponding evoked electroencephalogram (EEG) is indicative of neural processing of the speech signal. Using linguistic representations of speech, we investigate the differences in neural processing between speech in a native and foreign language that is not understood. We conducted experiments using three stimuli: a comprehensible language, an incomprehensible language, and randomly shuffled words from a comprehensible language, while recording the EEG signal of native Dutch-speaking participants. We modeled the neural tracking of linguistic features of the speech signals using a deep-learning model in a match-mismatch task that relates EEG signals to speech, while accounting for lexical segmentation features reflecting acoustic processing. The deep learning model effectively classifies coherent versus nonsense languages. We also observed significant differences in tracking patterns between comprehensible and incomprehensible speech stimuli within the same language. It demonstrates the potential of deep learning frameworks in measuring speech understanding objectively.
Asunto(s)
Electroencefalografía , Lenguaje , Percepción del Habla , Humanos , Percepción del Habla/fisiología , Electroencefalografía/métodos , Femenino , Masculino , Adulto , Adulto Joven , Aprendizaje Profundo , Habla/fisiología , LingüísticaRESUMEN
Craft beer brewers need to learn process control strategies from traditional industrial production to ensure the consistent quality of the finished product. In this study, FT-IR combined with deep learning was used for the first time to model and analyze the Plato degree and total flavonoid content of Qingke beer during the mashing and boiling stages and to compare the effectiveness with traditional chemometrics methods. Two deep learning neural networks were designed, the effect of variable input methods on the effectiveness of the models was discussed. The experimental results showed that the CARS-LSTM model had the best predictive performance, not only as the best quantitative model for Plato in the mashing (R2p = 0.9368) and boiling (R2p = 0.9398) phases but also as the best model for TFC in the boiling phase (R2p = 0.9154). This study demonstrates the great potential of deep learning and provides a new approach to quality control analysis in beer brewing.
RESUMEN
The exponential progress of image editing software has contributed to a rapid rise in the production of fake images. Consequently, various techniques and approaches have been developed to detect manipulated images. These methods aim to discern between genuine and altered images, effectively combating the proliferation of deceptive visual content. However, additional advancements are necessary to enhance their accuracy and precision. Therefore, this research proposes an image forgery algorithm that integrates error level analysis (ELA) and a convolutional neural network (CNN) to detect the manipulation. The system primarily focuses on detecting copy-move and splicing forgeries in images. The input image is fed to the ELA algorithm to identify regions within the image that have different compression levels. Afterward, the created ELA images are used as input to train the proposed CNN model. The CNN model is constructed from two consecutive convolution layers, followed by one max pooling layer and two dense layers. Two dropout layers are inserted between the layers to improve model generalization. The experiments are applied to the CASIA 2 dataset, and the simulation results show that the proposed algorithm demonstrates remarkable performance metrics, including a training accuracy of 99.05%, testing accuracy of 94.14%, precision of 94.1%, and recall of 94.07%. Notably, it outperforms state-of-the-art techniques in both accuracy and precision.
RESUMEN
Traditional methods may be inefficient when processing large-scale data in the field of text mining, often struggling to identify and cluster relevant information accurately and efficiently. Additionally, capturing nuanced sentiment and emotional context within news text is challenging with conventional techniques. To address these issues, this article introduces an improved bidirectional-Kmeans-long short-term memory network-convolutional neural network (BiK-LSTM-CNN) model that incorporates emotional semantic analysis for high-dimensional news text visual extraction and media hotspot mining. The BiK-LSTM-CNN model comprises four modules: news text preprocessing, news text clustering, sentiment semantic analysis, and the BiK-LSTM-CNN model itself. By combining these components, the model effectively identifies common features within the input data, clusters similar news articles, and accurately analyzes the emotional semantics of the text. This comprehensive approach enhances both the accuracy and efficiency of visual extraction and hotspot mining. Experimental results demonstrate that compared to models such as Transformer, AdvLSTM, and NewRNN, BiK-LSTM-CNN achieves improvements in macro accuracy by 0.50%, 0.91%, and 1.34%, respectively. Similarly, macro recall rates increase by 0.51%, 1.24%, and 1.26%, while macro F1 scores improve by 0.52%, 1.23%, and 1.92%. Additionally, the BiK-LSTM-CNN model shows significant improvements in time efficiency, further establishing its potential as a more effective approach for processing and analyzing large-scale text data.
RESUMEN
Agriculture is the main source of livelihood for most of the population across the globe. Plants are often considered life savers for humanity, having evolved complex adaptations to cope with adverse environmental conditions. Protecting agricultural produce from devastating conditions such as stress is essential for the sustainable development of the nation. Plants respond to various environmental stressors such as drought, salinity, heat, cold, etc. Abiotic stress can significantly impact crop yield and development posing a major threat to agriculture. SNARE proteins play a major role in pathological processes as they are vital proteins in the life sciences. These proteins act as key players in stress responses. Feature extraction is essential for visualizing the underlying structure of the SNARE proteins in analyzing the root cause of abiotic stress in plants. To address this issue, we developed a hybrid model to capture the hidden structures of the SNAREs. A feature fusion technique has been devised by combining the potential strengths of convolutional neural networks (CNN) with a high dimensional radial basis function (RBF) network. Additionally, we employ a bi-directional long short-term memory (Bi-LSTM) network to classify the presence of SNARE proteins. Our feature fusion model successfully identified abiotic stress in plants with an accuracy of 74.6%. When compared with various existing frameworks, our model demonstrates superior classification results.
RESUMEN
This work presents the application of an Encoder-Decoder convolutional neural network (ED-CNN) model to automatically segment COVID-19 computerised tomography (CT) data. By doing so we are producing an alternative model to current literature, which is easy to follow and reproduce, making it more accessible for real-world applications as little training would be required to use this. Our simple approach achieves results comparable to those of previously published studies, which use more complex deep-learning networks. We demonstrate a high-quality automated segmentation prediction of thoracic CT scans that correctly delineates the infected regions of the lungs. This segmentation automation can be used as a tool to speed up the contouring process, either to check manual contouring in place of a peer checking, when not possible or to give a rapid indication of infection to be referred for further treatment, thus saving time and resources. In contrast, manual contouring is a time-consuming process in which a professional would contour each patient one by one to be later checked by another professional. The proposed model uses approximately 49 k parameters while others average over 1,000 times more parameters. As our approach relies on a very compact model, shorter training times are observed, which make it possible to easily retrain the model using other data and potentially afford "personalised medicine" workflows. The model achieves similarity scores of Specificity (Sp) = 0.996 ± 0.001, Accuracy (Acc) = 0.994 ± 0.002 and Mean absolute error (MAE) = 0.0075 ± 0.0005.
RESUMEN
Background: For space object detection tasks, conventional optical cameras face various application challenges, including backlight issues and dim light conditions. As a novel optical camera, the event camera has the advantages of high temporal resolution and high dynamic range due to asynchronous output characteristics, which provides a new solution to the above challenges. However, the asynchronous output characteristic of event cameras makes them incompatible with conventional object detection methods designed for frame images. Methods: Asynchronous convolutional memory network (ACMNet) for processing event camera data is proposed to solve the problem of backlight and dim space object detection. The key idea of ACMNet is to first characterize the asynchronous event streams with the Event Spike Tensor (EST) voxel grid through the exponential kernel function, then extract spatial features using a feed-forward feature extraction network, and aggregate temporal features using a proposed convolutional spatiotemporal memory module ConvLSTM, and finally, the end-to-end object detection using continuous event streams is realized. Results: Comparison experiments among ACMNet and classical object detection methods are carried out on Event_DVS_space7, which is a large-scale space synthetic event dataset based on event cameras. The results show that the performance of ACMNet is superior to the others, and the mAP is improved by 12.7% while maintaining the processing speed. Moreover, event cameras still have a good performance in backlight and dim light conditions where conventional optical cameras fail. This research offers a novel possibility for detection under intricate lighting and motion conditions, emphasizing the superior benefits of event cameras in the realm of space object detection.
RESUMEN
Objective.Cardiovascular diseases are a major cause of mortality globally, and electrocardiograms (ECGs) are crucial for diagnosing them. Traditionally, ECGs are stored in printed formats. However, these printouts, even when scanned, are incompatible with advanced ECG diagnosis software that require time-series data. Digitizing ECG images is vital for training machine learning models in ECG diagnosis, leveraging the extensive global archives collected over decades. Deep learning models for image processing are promising in this regard, although the lack of clinical ECG archives with reference time-series data is challenging. Data augmentation techniques using realistic generative data models provide a solution.Approach.We introduceECG-Image-Kit, an open-source toolbox for generating synthetic multi-lead ECG images with realistic artifacts from time-series data, aimed at automating the conversion of scanned ECG images to ECG data points. The tool synthesizes ECG images from real time-series data, applying distortions like text artifacts, wrinkles, and creases on a standard ECG paper background.Main results.As a case study, we used ECG-Image-Kit to create a dataset of 21 801 ECG images from the PhysioNet QT database. We developed and trained a combination of a traditional computer vision and deep neural network model on this dataset to convert synthetic images into time-series data for evaluation. We assessed digitization quality by calculating the signal-to-noise ratio and compared clinical parameters like QRS width, RR, and QT intervals recovered from this pipeline, with the ground truth extracted from ECG time-series. The results show that this deep learning pipeline accurately digitizes paper ECGs, maintaining clinical parameters, and highlights a generative approach to digitization.Significance.The toolbox has broad applications, including model development for ECG image digitization and classification. The toolbox currently supports data augmentation for the 2024 PhysioNet Challenge, focusing on digitizing and classifying paper ECG images.
Asunto(s)
Aprendizaje Profundo , Electrocardiografía , Procesamiento de Imagen Asistido por Computador , Procesamiento de Imagen Asistido por Computador/métodos , Humanos , Procesamiento de Señales Asistido por Computador , Artefactos , Programas InformáticosRESUMEN
BACKGROUND: Wide-field calcium imaging (WFCI) with genetically encoded calcium indicators allows for spatiotemporal recordings of neuronal activity in mice. When applied to the study of sleep, WFCI data are manually scored into the sleep states of wakefulness, non-REM (NREM) and REM by use of adjunct EEG and EMG recordings. However, this process is time-consuming, invasive and often suffers from low inter- and intra-rater reliability. Therefore, an automated sleep state classification method that operates on spatiotemporal WFCI data is desired. NEW METHOD: A hybrid network architecture consisting of a convolutional neural network (CNN) to extract spatial features of image frames and a bidirectional long short-term memory network (BiLSTM) with attention mechanism to identify temporal dependencies among different time points was proposed to classify WFCI data into states of wakefulness, NREM and REM sleep. RESULTS: Sleep states were classified with an accuracy of 84â¯% and Cohen's κ of 0.64. Gradient-weighted class activation maps revealed that the frontal region of the cortex carries more importance when classifying WFCI data into NREM sleep while posterior area contributes most to the identification of wakefulness. The attention scores indicated that the proposed network focuses on short- and long-range temporal dependency in a state-specific manner. COMPARISON WITH EXISTING METHOD: On a held out, repeated 3-hour WFCI recording, the CNN-BiLSTM achieved a κ of 0.67, comparable to a κ of 0.65 corresponding to the human EEG/EMG-based scoring. CONCLUSIONS: The CNN-BiLSTM effectively classifies sleep states from spatiotemporal WFCI data and will enable broader application of WFCI in sleep research.