Classification Capability Research Articles

Optimizing Kernel Transformations to Handle Binary Class Imbalanced Dataset Classification

ABSTRACT Imbalanced class distributions pose a prevalent challenge in numerous classification problems, requiring effective strategies for learning from such skewed data. Traditional machine learning algorithms often struggle with imbalanced datasets, as they tend to bias their classification functions toward the majority class, resulting in suboptimal performance for minority classes. In our research, we propose a novel approach to address this challenge specifically tailored for Support Vector Machines (SVM), a well-established family of learning algorithms. Our method leverages a kernel trick to enhance the SVM’s classification capabilities on imbalanced datasets named KTI. It aims to streamline the classification process by incorporating adaptive data transformations within the algorithm itself, offering a more efficient and integrated solution for handling imbalanced data. Experimental evaluations conducted on diverse real-world datasets demonstrate the superior performance of our proposed strategy compared to existing methods, showcasing its potential for practical applications in classification tasks with skewed class distributions.

A multi-class fundus disease classification system based on an adaptive scale discriminator and hybrid loss

Fundus images are crucial in the observation and detection of ophthalmic diseases. However, detecting multiple ophthalmic diseases from fundus images using deep learning techniques is a complex and challenging task One challenge is the complexity of fundus disease structures, which leads to low detection accuracy. Another challenge is the class imbalance problem common in multi-label image classification, which increases the difficulty of algorithm training and evaluation. To address these issues, this study leverages deep learning to propose an ophthalmic disease classification system. We first employ ResNet50 as the backbone network to extract image features, and then use our designed multi-dimensional attention module and adaptive scale discriminator to enhance the network's ability to detect disease features. During training, we innovatively propose a hybrid loss function method to improve the detection capability on imbalanced data. Finally, we conducted experiments on the ODRI-5K dataset with the proposed classification system. In the test set, our method achieved an AUC of 98.53 and an F1-score of 89.73. This result fully demonstrates the excellent disease classification capability of our method. In summary, the multi-label fundus image disease classification system we proposed exhibits outstanding recognition capability, providing an effective solution for the diagnosis of multi-label fundus image diseases.

An innovative attention infused- BiRecurrenTwin network assisted hybrid segmentation technique for accurate heart disease prediction

Heart disease is a noteworthy global health concern, profoundly affecting public health and individual well-being. Early prediction and diagnosis are crucial for reducing its negative impact. This work focuses on developing an effectual heart disease detection employing advanced optimization approaches and data processing. The proposed system incorporates a hybrid segmentation algorithm, Hybrid K-means-Fuzzy-C-means (HKMFCM) clustering, which assigns membership degrees to data points, enabling soft clustering and accommodating data uncertainty. Additionally, the Artificial Bee Colony (ABC) Optimization method is applied for feature selection, mimicking the foraging behavior of honeybees to identify the most discriminative attributes from the dataset. This optimization algorithm iteratively explores the feature space to select features that enhance the model's predictive accuracy. Furthermore, a novel classification architecture, termed the Attention-infused BiRecurrenTwin Network, is introduced to accurately predict heart disease based on segmented and extracted patient data profiles. This classifier leverages both Bidirectional Gated Recurrent Unit (BiGRU) and Bidirectional Long Short-Term Memory (BiLSTM) networks, with their bidirectional nature capturing both past and future contexts, thus increasing the classifier's capability to detect subtle temporal patterns in patient data. In addition, the proposed system addresses traditional approaches limitations through its advanced components like HKMFCM (effectively manages data uncertainty by enabling soft clustering), ABC Optimization technique (enabling an iterative, global search for the most discriminative attributes) Attention-infused BiRecurrenTwin Network (surpasses conventional classifiers by capturing both past and future temporal patterns in patient data). The simulation outcomes demonstrate that the developed system attains improved performance, with accuracy, ROC, F-measure, precision, recall, sensitivity, and specificity values of 96.09%, 97%, 95%, 96.6%, 94.3%, 95.14%, and 97.05%, respectively which is higher than the baseline models with an average of 7.75 % increase in accuracy. These results indicate that the developed predictive models show promise in accurately classifying individuals based on their extracted features, thereby facilitating the early recognition of heart disease.

Road target recognition based on radar range-Doppler spectrum with GS – ResNet

ABSTRACT Road target recognition ability of automotive millimetre-wave (mmWave) radar is crucial in areas such as autonomous driving, advanced driver assistance systems (ADAS), and automated emergency braking systems (AEBS). Current mmWave radar systems are primarily utilized for speed, distance, and angle measurements, with limited capability or unsatisfactory performance in target classification. In response to this limitation, this paper proposes a novel method for road target recognition, using radar range-Doppler spectrum as input data and introducing a global attention mechanism and SCConv module. First, the range-Doppler spectrum is processed by the clutter removal algorithm block to remove background clutter and vibration interference. Second, considering the small features and susceptibility to noise interference in the range-Doppler spectrum, we have devised a deep residual network based on a global attention mechanism, significantly enhancing the model’s accuracy in range-Doppler spectrum classification. Finally, we introduce the SCConv module and improve the downsampling module in ResNet to enhance the model’s classification accuracy. Experimental results demonstrate that the model achieves an average accuracy of 99.79% in classifying six types of road targets, significantly outperforming other methods. This research is of significant importance in advancing the understanding of road traffic conditions by autonomous driving systems and enhancing system safety.

2D-convolutional neural network based fault detection and classification of transmission lines using scalogram images

The reliable operation of power transmission systems is essential for maintaining the stability and efficiency of the electrical grid. Rapid and accurate detection of faults in transmission lines is crucial for minimizing downtime and preventing cascading failures. This research presents a novel approach to fault detection and classification in transmission lines employing 2D Convolutional Neural Networks (2D-CNN).The proposed methodology leverages the inherent spatial characteristics of fault signals, converting them as 2D scalogram images for input to the CNN model. By converting fault signals into scalogram representations, the network can capture both temporal and frequency domain features, enabling a more comprehensive analysis of fault patterns. The 2D-CNN architecture is designed to automatically learn hierarchical features, allowing for effective discrimination between different fault types. To evaluate the performance of the proposed approach, extensive simulations and experiments were conducted using MATLAB/SIMULINK modeled transmission line data. The results demonstrate the superior fault detection accuracy and classification capabilities of the 2D-CNN model. The performance of the proposed model is evaluated using 10-fold cross-validation, and its effectiveness is assessed by comparing it with current state-of-the-art techniques. Proposed 2D-CNN model has evidenced an accuracy of 99.9074 with ideal dataset for 12- class fault classification and performing consistently in presence of noise, having an accuracy of 99.629 %,99.72 % and 99.814 % in 20.30 and 40 dB noises respectively. The proposed model also verified in high resistance fault condition. The model exhibits robustness to noise and is capable of generalizing well to various fault scenarios. The proposed methodology offers a scalable and efficient solution for transmission line fault analysis, paving the way for the integration of advanced machine learning techniques into the operation and maintenance of power transmission infrastructure.

A Novel Strategy Coupling Optimised Sampling with Heterogeneous Ensemble Machine-Learning to Predict Landslide Susceptibility

The accuracy of data-driven landslide susceptibility prediction depends heavily on the quality of non-landslide samples and the selection of machine-learning algorithms. Current methods rely on artificial prior knowledge to obtain negative samples from landslide-free regions or outside the landslide buffer zones randomly and quickly but often ignore the reliability of non-landslide samples, which will pose a serious risk of including potential landslides and lead to erroneous outcomes in training data. Furthermore, diverse machine-learning models exhibit distinct classification capabilities, and applying a single model can readily result in over-fitting of the dataset and introduce potential uncertainties in predictions. To address these problems, taking Chenxi County, a hilly and mountainous area in southern China, as an example, this research proposes a strategy-coupling optimised sampling with heterogeneous ensemble machine learning to enhance the accuracy of landslide susceptibility prediction. Initially, 21 landslide impact factors were derived from five aspects: geology, hydrology, topography, meteorology, human activities, and geographical environment. Then, these factors were screened through a correlation analysis and collinearity diagnosis. Afterwards, an optimised sampling (OS) method was utilised to select negative samples by fusing the reliability of non-landslide samples and certainty factor values on the basis of the environmental similarity and statistical model. Subsequently, the adopted non-landslide samples and historical landslides were combined to create machine-learning datasets. Finally, baseline models (support vector machine, random forest, and back propagation neural network) and the stacking ensemble model were employed to predict susceptibility. The findings indicated that the OS method, considering the reliability of non-landslide samples, achieved higher-quality negative samples than currently widely used sampling methods. The stacking ensemble machine-learning model outperformed those three baseline models. Notably, the accuracy of the hybrid OS–Stacking model is most promising, up to 97.1%. The integrated strategy significantly improves the prediction of landslide susceptibility and makes it reliable and effective for assessing regional geohazard risk.

Enhancing ROP plus form diagnosis: An automatic blood vessel segmentation approach for newborn fundus images

BackgroundROP Plus Form is an eye disease that can lead to blindness, and diagnosing it requires medical experts to manually examine the retinal condition. This task is challenging due to its subjective nature and poor image quality. Therefore, developing automatic tools for Retinal Blood Vessel Segmentation in fundus images could assist healthcare experts in diagnosing, monitoring, and prognosing the disease. ObjectiveThis study focuses on developing a novel pipeline for automatically segmenting retinal blood vessels. The main requirements are that it can correctly identify the blood vessels in fundus images and perform well on different systems used for newborn evaluation. MethodsThe pipeline uses different methods, including CIELAB Enhancement, Background Normalization, Bell-Shaped Gaussian Matched Filtering, Modified Top-Hat operation, and a combination of vesselness filtering composed of Frangi and Jerman Filters. The segmentation is done by determining a threshold using the Triangle Threshold algorithm. A novel filter is also proposed to remove the Optical Disc artifacts from the primary segmentation based on the Circular Hough Transform. The segmentation pipeline is combined with different pre-trained Convolution Neural Network architectures to evaluate its automatic classification capabilities. ResultsThe pipeline was tested with newborn fundus images acquired with Clarity RetCam3 and Phoenix ICON systems. The results were compared against annotations from three ophthalmologic experts. Clarity RetCam3 images achieved an accuracy of 0.94, specificity of 0.95, and sensitivity of 0.81, while Phoenix ICON images achieved an accuracy of 0.94, specificity of 0.97, and sensitivity of 0.83. The pipeline was also tested for the DRIVE Database, achieving an accuracy of 0.95, specificity of 0.97, and sensitivity of 0.82. For the classification task, the best results were achieved with the DenseNet121 architecture with an accuracy of 0.946. ConclusionThe segmentation scores were auspicious and confirmed the clinical relevance of the proposed pipeline. It has also proven to have a good generalization performance, essential for easier clinic integration. Finally, preliminary results on using CNNs showed how our work can be used to develop fully automatic tools for diagnosing ROP Plus form disease.

Application of logistic regression to detect an attack on financial monitoring

This paper discusses the application of logistic regression to detect attacks on financial monitoring. Financial organizations are facing increasing security threats, which requires effective methods to detect fraudulent activities. Logistic regression, with its binary classification capability and interpretability of results, is a powerful tool for transaction analysis. The paper also emphasizes the features of this method, including its applicability to large amounts of data and opportunities for model development. Thus, the use of logistic regression is a promising solution for securing financial transactions and identifying potential threats.

High-performance virtual sensors array based on a single-chip FBAR for volatile organic compounds (VOCs) detection and classification

The detection and classification of volatile organic compounds (VOCs) are of great significance in atmospheric environment monitoring and have garnered extensive attention in recent decades. This study proposes a virtual sensor array (VSA) for VOC detection and classification fabricating a single-chip high-Q-factor film bulk acoustic resonator (FBAR) functionalized with MIL-101(Cr) thin film, known for its porous crystal structure and high specific surface area. The sensing performance of five different VOCs was evaluated to characterize the VOCs, achieving high detection accuracy calibrated with the Langmuir adsorption model. The FBAR’s frequency response was utilized to determine the sensitivity and limit of detection (LOD), yielding sensitivity and LOD of 85.45 Hz/ppm and 51 ppm, respectively, for isopropanol. The identification and classification capabilities of the proposed VSA for VOCs and their mixtures were assessed using statistical methods and machine learning algorithms, realizing classification accuracies of 100 % for pure VOCs and 94.6 % for mixtures. The developed VSA demonstrates superior performances in VOC detection and classification, with potential applications across various domains.

In Silico Modeling of Fabry Disease Pathophysiology for the Identification of Early Cellular Damage Biomarker Candidates.

Fabry disease (FD) is an X-linked lysosomal disease whose ultimate consequences are the accumulation of sphingolipids and subsequent inflammatory events, mainly at the endothelial level. The outcomes include different nervous system manifestations as well as multiple organ damage. Despite the availability of known biomarkers, early detection of FD remains a medical need. This study aimed to develop an in silico model based on machine learning to identify candidate vascular and nervous system proteins for early FD damage detection at the cellular level. A combined systems biology and machine learning approach was carried out considering molecular characteristics of FD to create a computational model of vascular and nervous system disease. A data science strategy was applied to identify risk classifiers by using 10 K-fold cross-validation. Further biological and clinical criteria were used to prioritize the most promising candidates, resulting in the identification of 36 biomarker candidates with classifier abilities, which are easily measurable in body fluids. Among them, we propose four candidates, CAMK2A, ILK, LMNA, and KHSRP, which have high classification capabilities according to our models (cross-validated accuracy ≥ 90%) and are related to the vascular and nervous systems. These biomarkers show promise as high-risk cellular and tissue damage indicators that are potentially applicable in clinical settings, although in vivo validation is still needed.

Electronic Nose Humidity Compensation System Based on Rapid Detection.

In this study, we present an electronic nose (e-nose) humidity compensation system based on rapid detection to solve the issue of humidity drift's potential negative impact on the performance of electronic noses. First, we chose the first ten seconds of non-steady state (rapid detection mode) sensor data as the dataset, rather than waiting for the electronic nose to stabilize during the detection process. This was carried out in the hope of improving the detection efficiency of the e-nose and to demonstrate that the e-nose can collect gasses efficiently in rapid detection mode. The random forest approach is then used to optimize and reduce the dataset's dimensionality, filtering critical features and improving the electronic nose's classification capacity. Finally, this study builds an electronic nose humidity compensation system to compensate for the datasets generated via rapid real-time detection, efficiently correcting the deviation of the sensor response caused by humidity variations. This method enhanced the average resolution of the electronic nose in this trial from 87.7% to 99.3%, a 12.4% improvement, demonstrating the efficacy of the humidity compensation system based on rapid detection for the electronic nose. This strategy not only improves the electronic nose's anti-drift and classification capabilities but also extends its service life, presenting a new solution for the electronic nose in practical detecting applications.

Lobish: Symbolic Language for Interpreting Electroencephalogram Signals in Language Detection Using Channel-Based Transformation and Pattern.

Electroencephalogram (EEG) signals contain information about the brain's state as they reflect the brain's functioning. However, the manual interpretation of EEG signals is tedious and time-consuming. Therefore, automatic EEG translation models need to be proposed using machine learning methods. In this study, we proposed an innovative method to achieve high classification performance with explainable results. We introduce channel-based transformation, a channel pattern (ChannelPat), the t algorithm, and Lobish (a symbolic language). By using channel-based transformation, EEG signals were encoded using the index of the channels. The proposed ChannelPat feature extractor encoded the transition between two channels and served as a histogram-based feature extractor. An iterative neighborhood component analysis (INCA) feature selector was employed to select the most informative features, and the selected features were fed into a new ensemble k-nearest neighbor (tkNN) classifier. To evaluate the classification capability of the proposed channel-based EEG language detection model, a new EEG language dataset comprising Arabic and Turkish was collected. Additionally, Lobish was introduced to obtain explainable outcomes from the proposed EEG language detection model. The proposed channel-based feature engineering model was applied to the collected EEG language dataset, achieving a classification accuracy of 98.59%. Lobish extracted meaningful information from the cortex of the brain for language detection.

Intelligent ultrasonic aspirator: Advancing tissue differentiation through hierarchical classification during hand-held resection

Modern neurosurgery strives to maximize tumor removal while preserving healthy tissue integrity. Accurate intraoperative differentiation between tumor and healthy tissue is crucial yet challenging. Often neurosurgeons rely on their experience and haptic feedback during palpation to distinguish between tumor and healthy tissue. A commonly used hand-held tool for tissue removal during neurosurgery is the ultrasonic aspirator, which changes its electrical properties as it interacts with tissue. The goal is to equip the ultrasonic aspirator with the ability to differentiate between different types of tissue while at the same time not interfering with the surgical workflow and providing comprehensible outcomes. To this end, a hierarchical classification approach is employed as a proof of concept, enabling precise identification of tissue stiffness during resection.The hierarchical approach is compared with the standard flat classification, commonly used in machine learning. Within the hierarchical approach, two strategies are employed: mandatory leaf-node predictions (MLNP) and non-mandatory leaf-node predictions (NMLNP). The NMLNP allows prediction to revert to a parent node when certainty is low. Data are acquired on three artificial tissue models – differing in stiffness – with an ultrasonic aspirator in a hand-held manner. The dataset comprises 1,821 data points for training and 186 for testing after balancing.The results indicate a slight performance advantage for the hierarchical classification MLNP approach over the flat classification approach in the absence of confidence thresholds, with weighted F2-scores of 0.781 and 0.762, respectively. However, the application of confidence thresholds results in both approaches exhibiting comparable performance, with the hierarchical NMLNP approach achieving a weighted F1-score of 0.920, thereby demonstrating superior overall performance. The effects of enforcing these thresholds and excluding data with low certainty are thoroughly investigated. This work emphasizes the feasibility of tissue differentiation using a hand-held ultrasound aspirator while resecting tissue. Moreover, it highlights the capability of hierarchical classification in advancing tissue differentiation accuracy during neurosurgical procedures, which could ultimately aid surgeons and enhance the safety of intraoperative workflows.

Enhancing trustworthiness in ML-based network intrusion detection with uncertainty quantification

A crucial role in the security of modern networks is played by Intrusion Detection Systems (IDSs), security devices designed to identify and mitigate attacks to networks structure. Data-driven approaches based on Machine Learning (ML) have gained more and more popularity for executing the classification tasks required by signature-based IDSs. However, typical ML models adopted for this purpose do not properly take into account the uncertainty associated with their prediction. This poses significant challenges, as they tend to produce misleadingly high classification scores for both misclassified inputs and inputs belonging to unknown classes (e.g. novel attacks), limiting the trustworthiness of existing ML-based solutions. In this paper, we argue that ML-based IDSs should always provide accurate uncertainty quantification to avoid overconfident predictions. In fact, an uncertainty-aware classification would be beneficial to enhance closed-set classification performance, would make it possible to carry out Active Learning, and would help recognize inputs of unknown classes as truly unknowns, unlocking open-set classification capabilities and Out-of-Distribution (OoD) detection. To verify it, we compare various ML-based methods for uncertainty quantification and open-set classification, either specifically designed for or tailored to the domain of network intrusion detection. Moreover, we develop a custom model based on Bayesian Neural Networks that stands out for its OoD detection capabilities and robustness, with a lower variance in the results over different scenarios, compared to other baselines, thus showing how proper uncertainty quantification can be exploited to significantly enhance the trustworthiness of ML-based IDSs.

You only look once v8 for fish species identification

<p>This research aims to test the performance of you only look once (YOLOv8) in identifying fish species in Indonesian waters. Fish image data is obtained from various sources to conduct testing. The results show that YOLOv8 is able to identify fish species with a mAP accuracy rate of 97%. These results reveal the great potential of deep learning technology in supporting the preservation of marine biodiversity as well as the development of various applications, such as fisheries monitoring, conservation, and marine-based tourism development in Indonesia. With its efficient object detection and classification capabilities, YOLOv8 can simplify and accelerate the process of identifying fish species, even on a large scale. Thus, this technology is a highly effective solution to overcome the challenges of manual fish species identification, which requires a lot of time and effort. The results of this study provide valuable insights into the potential utilization of Indonesia's natural resources in the context of scientific development, the tourism industry, and the fisheries sector, which is vital to the country's economy.</p>

Telomere-related prognostic signature for survival assessments in lung adenocarcinoma.

Telomere-related genes (TRGs) are important in many different types of cancers. However, there is a lack of research on the relationship between their expression and prognosis in lung adenocarcinoma (LUAD) patients. This study is to investigate the prognostic value of TRGs in LUAD and to develop a TRG signature that can predict patient survival. A total of 2,086 TRGs were obtained from a database of genes involved in telomere maintenance (TelNet), while the clinical information and tumor RNA expression profiles of 513 LUAD patients were acquired from The Cancer Genome Atlas (TCGA) database. Statistical methodologies, such as least absolute shrinkage and selection operator (LASSO)-Cox, were employed to construct a prognostic model with predictive capabilities. We analyzed 1,339 telomere-associated differentially expressed genes and identified a ten-gene predictive signature for LUAD. This signature exhibited effective prognostic classification capabilities across multiple datasets, including GSE3141 (58 samples), GSE8894 (63 samples), GSE50081 (127 samples), and GSE72094 (398 samples). Furthermore, we screened tumor-sensitive drugs targeting this signature. High telomere levels were associated with reduced survival in lung cancer patients who underwent surgery. Compared to the traditional TNM (tumor node metastasis classification) grading method, our telomere-associated gene panel demonstrated superior prediction accuracy. Notably, patients in the high-risk group, defined by the telomere-associated signature, exhibited improved responses to immunotherapy, suggesting potential benefits for this subgroup of patients. This study presents a comprehensive molecular signature comprising TRGs, which holds potential for functional and therapeutic investigations. Additionally, it serves as an integrated tool to identify crucial molecules for immunotherapy in lung cancer.

Intelligent Gesture Recognition Based on Screen Reflectance Multi-Band Spectral Features.

Human-computer interaction (HCI) with screens through gestures is a pivotal method amidst the digitalization trend. In this work, a gesture recognition method is proposed that combines multi-band spectral features with spatial characteristics of screen-reflected light. Based on the method, a red-green-blue (RGB) three-channel spectral gesture recognition system has been developed, composed of a display screen integrated with narrowband spectral receivers as the hardware setup. During system operation, emitted light from the screen is reflected by gestures and received by the narrowband spectral receivers. These receivers at various locations are tasked with capturing multiple narrowband spectra and converting them into light-intensity series. The availability of multi-narrowband spectral data integrates multidimensional features from frequency and spatial domains, enhancing classification capabilities. Based on the RGB three-channel spectral features, this work formulates an RGB multi-channel convolutional neural network long short-term memory (CNN-LSTM) gesture recognition model. It achieves accuracies of 99.93% in darkness and 99.89% in illuminated conditions. This indicates the system's capability for stable operation across different lighting conditions and accurate interaction. The intelligent gesture recognition method can be widely applied for interactive purposes on various screens such as computers and mobile phones, facilitating more convenient and precise HCI.

Classification of Multi-Parametric Body MRI Series Using Deep Learning.

Multi-parametric magnetic resonance imaging (mpMRI) exams have various series types acquired with different imaging protocols. The DICOM headers of these series often have incorrect information due to the sheer diversity of protocols and occasional technologist errors. To address this, we present a deep learning-based classification model to classify 8 different body mpMRI series types so that radiologists read the exams efficiently. Using mpMRI data from various institutions, multiple deep learning-based classifiers of ResNet, EfficientNet, and DenseNet are trained to classify 8 different MRI series, and their performance is compared. Then, the best-performing classifier is identified, and its classification capability under the setting of different training data quantities is studied. Also, the model is evaluated on the out-of-training-distribution datasets. Moreover, the model is trained using mpMRI exams obtained from different scanners in two training strategies, and its performance is tested. Experimental results show that the DenseNet-121 model achieves the highest F1-score and accuracy of 0.966 and 0.972 over the other classification models with p-value 0.05. The model shows greater than 0.95 accuracy when trained with over 729 studies of the training data, whose performance improves as the training data quantities grew larger. On the external data with the DLDS and CPTAC-UCEC datasets, the model yields 0.872 and 0.810 accuracy for each. These results indicate that in both the internal and external datasets, the DenseNet-121 model attains high accuracy for the task of classifying 8 body MRI series types.

Utilizing language models for advanced electrocardiogram analysis

Electrocardiography (ECG) signals are often referred to as the language of the heart and have been widely utilized for diagnosing various heart ailments, particularly arrhythmias. Consequently, numerous machine learning models have been employed to automatically detect heart disorders using ECG signals. In this research, the primary objective is to detect arrhythmias using a center-symmetric self-organized textual pattern. A novel feature engineering model has been introduced, which includes the following components: (i) multilevel feature extraction using the proposed center-symmetric self-organized textual pattern (CSSOTP), (ii) iterative neighborhood component analysis (INCA), and (iii) classification using k-nearest neighbors (kNN) with 10-fold cross-validation. During the feature extraction phase, a multilevel feature extraction model incorporating maximum absolute pooling (MAP) and the proposed CSSOTP was applied. The CSSOTP was designed to select the most appropriate pattern for the given data block. A large language model (LLM) was leveraged to generate text for creating patterns, and ChatGPT was used to assist in text generation. To identify the most informative features, the INCA feature selector was employed, and the selected features were subsequently classified using the kNN classifier. An impressive 96.20 % classification accuracy was achieved by the CSSOTP-based feature engineering model when tested on an ECG dataset containing 17 classes. Furthermore, comparisons with state-of-the-art feature engineering models were conducted, demonstrating that the proposed model has superior classification capabilities.

Artificial intelligence-based classification of cardiac autonomic neuropathy from retinal fundus images in patients with diabetes: The Silesia Diabetes Heart Study

BackgroundCardiac autonomic neuropathy (CAN) in diabetes mellitus (DM) is independently associated with cardiovascular (CV) events and CV death. Diagnosis of this complication of DM is time-consuming and not routinely performed in the clinical practice, in contrast to fundus retinal imaging which is accessible and routinely performed. Whether artificial intelligence (AI) utilizing retinal images collected through diabetic eye screening can provide an efficient diagnostic method for CAN is unknown.MethodsThis was a single center, observational study in a cohort of patients with DM as a part of the Cardiovascular Disease in Patients with Diabetes: The Silesia Diabetes-Heart Project (NCT05626413). To diagnose CAN, we used standard CV autonomic reflex tests. In this analysis we implemented AI-based deep learning techniques with non-mydriatic 5-field color fundus imaging to identify patients with CAN. Two experiments have been developed utilizing Multiple Instance Learning and primarily ResNet 18 as the backbone network. Models underwent training and validation prior to testing on an unseen image set.ResultsIn an analysis of 2275 retinal images from 229 patients, the ResNet 18 backbone model demonstrated robust diagnostic capabilities in the binary classification of CAN, correctly identifying 93% of CAN cases and 89% of non-CAN cases within the test set. The model achieved an area under the receiver operating characteristic curve (AUCROC) of 0.87 (95% CI 0.74–0.97). For distinguishing between definite or severe stages of CAN (dsCAN), the ResNet 18 model accurately classified 78% of dsCAN cases and 93% of cases without dsCAN, with an AUCROC of 0.94 (95% CI 0.86–1.00). An alternate backbone model, ResWide 50, showed enhanced sensitivity at 89% for dsCAN, but with a marginally lower AUCROC of 0.91 (95% CI 0.73–1.00).ConclusionsAI-based algorithms utilising retinal images can differentiate with high accuracy patients with CAN. AI analysis of fundus images to detect CAN may be implemented in routine clinical practice to identify patients at the highest CV risk.Trial registrationThis is a part of the Silesia Diabetes-Heart Project (Clinical-Trials.gov Identifier: NCT05626413).Graphical