Many existing Graph Neural Networks (GNN) methods assume that labels are reliable and sufficient, which may not be the case in real-world scenarios. This paper addresses one such problem of Partial Label Learning (PLL) on graph-structured data. In the PLL for graphs, each node is represented by a candidate set of labels, where only one is true while the others are inaccurate. Despite advancements with PLL in tabular and vision domains, the graph-structured data still needs to be explored. In this work, we first define PLL for graphs. Subsequently, we propose a new PLD-Graph algorithm for PLL in homogeneous graphs with scarce labels. We utilize graph augmentation to reduce the effects of inexact labels and provide additional supervision from unlabeled nodes. Progressive label disambiguation is performed based on the model's ability to predict correct classes. Furthermore, an additional loss estimates the label corruption matrix to capture associations between correct and incorrect labels. We show the effectiveness of the proposed algorithm on multiple graph datasets, with two types of noise and varying levels of ambiguous labels. Overall, the proposed PLD-Graph algorithm outperforms state-of-the-art PLL methods.

Small object detection in radiological images has been a key challenge in the field of medical diagnosis for the last decade. Radiological modalities such as computed tomography scan imaging are often used to evaluate the condition of a patient. These modalities can capture anomalies from various parts of the body which are then analyzed by radiologists to identify tumors, stones, nodules, etc. However, it becomes a strenuous task for any radiologist to accurately identify tiny abnormalities contained within the radiological images. In this paper, a two-step lightweight computer-aided anomaly detection method is proposed that is suitable for small object detection in computed tomography scan images. The proposed method leverages the power of uniform superpixel generation, graph propagation, and positional convolutions to detect anomalies in computed tomography scan images with great accuracy. Furthermore, experiments conducted on imaging datasets from two distinct organs of the body that is the lungs and the kidneys, show the effectiveness of the proposed method in terms of small object detection in medical imaging.

Diabetic Retinopathy is one of the leading global causes of vision impairment and blindness in humans. It has seen a rise in prevalence, necessitating the development of advanced automatic detection methods. This paper presents a survey of the evolution in deep learning techniques for diabetic retinopathy detection, emphasizing the transition from traditional machine learning to sophisticated deep learning architectures such as convolutional neural networks. It discusses the role of transfer learning, end-to-end learning, and hybrid models in overcoming medical detection challenges while highlighting the need for artificial intelligence interpretability and real-time screening integration in clinical workflows. Building on this survey, the paper introduces a focused study on cross-dataset deployment of transfer learning for diabetic retinopathy detection and grading. Consequently, this paper evaluates 26 pre-trained models from various convolutional neural network families to provide a comprehensive comparison between the state-of-the-art CNN architectures in the field. Additionally, this study also employs Grad-CAM visualization to interpret the model’s decision-making, bridging advanced artificial intelligence techniques with practical healthcare applications for diabetic retinopathy.

Automatic detection of lung nodules has been a key element in modern medical research in the past decade. By utilizing accurate nodule detection approaches, lung cancer can be treated in its early stages, reducing its mortality rate. Manual detection of nodules is hindered by structures contained in the computed tomography scan images such as veins, bronchioles and lymphatics. Furthermore, nodules are tiny and sometimes have haphazard boundaries making them easy to miss. In this work, we propose a two-step lung nodule detection method that uses a novel hierarchical superpixel merging (HSPM) module to reduce the number of proposed regions of interest and couples it with a new graph neural network namely, the regional downsampled residual superpixel aggregation network (dRes-SPAN) to accurately detect nodules in the lung parenchyma. Experiments on the Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI) dataset show a high improvement in performance in terms of nodule classification and detection while achieving an overall low false positive rate. A significant reduction in training time and trainable parameters has also been observed along with the relaxation in the requirement for heavy annotation compared to the state-of-the-art object detection architectures.

Scalability and training time are crucial for any graph neural network model processing a knowledge graph (KG). While partitioning knowledge graphs helps reduce the training time, the prediction accuracy reduces significantly compared to training the model on the whole graph. In this paper, we propose CPa-WAC: a lightweight architecture that incorporates graph convolutional networks and modularity maximization-based constellation partitioning to harness the power of local graph topology. The proposed CPa-WAC method reduces the training time and memory cost of knowledge graph embedding, making the learning model scalable. The results from our experiments on standard databases, such as Wordnet and Freebase, show that by achieving meaningful partitioning, any knowledge graph can be broken down into subgraphs and processed separately to learn embeddings. Furthermore, these learned embeddings can be used for knowledge graph completion, retaining similar performance compared to training a GCN on the whole KG, while speeding up the training process by upto five times. Additionally, the proposed CPa-WAC method outperforms several other state-of-the-art KG in terms of prediction accuracy.

Lung nodule detection is key in early diagnosis of lung cancer. Expert radiologists dedicate a significant amount of time and effort to detecting such nodules manually by going through computed tomography scan images slice by slice. This endeavor results in the slow processing of radiological images and possible misdiagnosis due to nodules being tiny by nature. In this paper, we introduce a two-step automatic computer-aided nodule detection method that encompasses a novel uniform superpixel generation algorithm, namely, equivalent patchwise iterative agglomerative clustering. This superpixel generation algorithm can generate the same number of superpixels for every image making it suitable for training neural networks. This method is then coupled with a novel variant of graph neural networks, namely, the curtailed residual nested superpixel propagation network, and an unsupervised region proposal method, namely, pixel nesting region proposal mechanism to detect nodules with high accuracy. The results show an accelerated training process compared to state-of-the-art convolutional neural networks and good generalization capability. Furthermore, the proposed method displays a significant reduction in trainable parameters while achieving high performance in the detection and segmentation of nodules on the Lung Image Database Consortium and Image Database Resource Initiative dataset.

Recent advances in deep learning have given rise to high performance in image analysis operations in healthcare. Lung diseases are of particular interest, as most can be identified using non-invasive image modalities. Deep learning techniques such as convolutional neural networks, convolution autoencoders, and graph convolutional networks have been implemented in several pulmonary disease identification applications, e.g., lung nodule classification, Covid-19, and pneumonia detection. Various sources of medical images such as X-rays, computed tomography scans, magnetic resonance imaging, and positron emission tomography scans make deep learning techniques favorable to identify lung diseases with great accuracy. This paper discusses state-of-the-art methods that use deep learning on various medical imaging modalities to detect and classify diseases in the lungs. A description of a few publicly available databases is included in this study, along with some distinct deep learning techniques developed in recent times. Furthermore, several challenges and open research areas for pulmonary disease diagnosis using deep learning are discussed. The objective of this work is to direct researchers in the field of diagnosis of lung diseases.

Graph neural networks (GNN) have recently been integrated into knowledge graph representation learning. The efficient message-passing functions in GNNs capture latent relationships between entities within these semantic networks, which aids in various downstream tasks such as link prediction, node classification, and entity alignment. However, there is a general deficiency in representation learning on graphs with loops (cycles) and self-loops 1 . Traditional message-passing functions induce biased learning on knowledge graphs, leading to skewed predictions. This work presents a detailed analysis of representation bias generated by these functions on knowledge graphs containing short and self-loops. We demonstrate the variance in performance on knowledge graphs with varying topology over two downstream: link prediction and entity alignment. The experiments show that the representations from popular learning algorithms are prone to capturing biases in the graphs’ structures. These biases, however, have different effects on the formulated downstream tasks, motivating research in the domain of topology-invariant representation algorithms for knowledge graphs.

One of the most important steps in lung nodule diagnosis is the automatic segmentation of nodules irrespective of their position and size in the lung parenchyma. In this paper, we propose a new way of applying graph clustering to nodule segmentation. Firstly, the image is preprocessed to extract the lung parenchyma from the CT scan image and identify the region of interest. This is followed by the application of Patchwise Iterative Graph Clustering to spilt the patches and generate superpixels. Next, a region adjacency graph is generated, and agglomerative hierarchical clustering is used to merge the superpixels into different structures such as nodules, and blood vessels. A thresholding algorithm is then used to extract the nodules from the clusters. The proposed method has shown good performance in segmentation with an average dice score of 0.88, an intersection over union score of 0.81, and a high average sensitivity of 89.32 %. Furthermore, the proposed method has been compared to several state-of-the-art methods in the field and has shown an increase in performance in terms of the evaluation metrics.

Skin cancer, characterized by the abnormal growth of skin cells, is a severe and prevalent condition. Despite the advancement in digital diagnosis techniques, existing skin cancer detection methods often fail to achieve satisfactory performance in melanoma identification in dermoscopy imaging. This study presents a new melanoma classification algorithm that utilizes a superpixel-based segmentation technique, simple linear iterative clustering, and transfer learning-based convolutional neural networks to achieve the objective. A key innovation of our methodology is the introduction of a novel approach for efficiently merging superpixels, that enhances the quality of segmentation resulting in better classification performance. Following the segmentation phase, several convolutional neural networks have been utilized for feature extraction from segmented images and classification. The proposed method shows an enhanced performance in melanoma classification which is 91.67%, 95.33%, 91.23%, and 90.48% in terms of accuracy, precision, recall and F1-score, respectively. Our results indicate that the superpixel segmentation technique considerably enhances the classification models’ accuracy compared to k-means segmentation methods. A comparative analysis between the proposed method and several state-of-the-art methods in the field has also been presented in the context.

Prediction of heart disease is one of the most important fields of study in modern science. By studying data such as cholesterol levels, blood sugar, and blood pressure, heart disease can be predicted. In recent years, several machine learning techniques have been used to aid in fast prediction by learning from the data. However, the prediction accuracy still remains low. This is due to lower number of records contained in the databases available. In this paper, we propose a new method of heart disease prediction using a modified variation of infinite feature selection and multilayer perceptron. The method shows a high accuracy of 87.70%, a high F1-score of 87.21%, a high sensitivity of 88.50%, a high specificity of 87.02%, and a high precision in prediction of 86.05%. on the Cleveland, Hungarian, Switzerland, Long Beach, and Statlog datasets. For evaluation purposes, we have combined all the datasets together and then divided the combined dataset into training and test samples with a 20 % percent of the samples allocated for testing.

The heart is the single most important organ of the human body. By studying the ECG signal any abnormalities in the heart’s tempo can be identified. The QRS complex is the most prominent feature of an ECG signal. However, the detection of this feature is hampered by the presence of high and low-frequency noises in the ECG signal and abrupt changes in the signal’s morphology. This paper presents a new method of QRS detection using advanced adaptive multilevel thresholding (AAMT) with selective statistical false peak elimination (SSFPE). Firstly, a band-pass filter is used to filter out most of the unwanted noise and interference to aid the detection operation. Then, AAMT is applied to the entire ECG record to find the location and amplitudes of the pseudo peaks. Next, SSFPE is employed to eliminate false peaks resulting from electromyogram (EMG) and any other high-frequency noise that has not been eliminated in the filtering stage. Finally, after most of the peaks are correctly identified a search back stage is included to find any low amplitude true peaks that might have been missed in the peak detection stage. The proposed method is tested on the MIT-BIH arrhythmia and Fantasia databases and shows high accuracy in detection compared to many state-of-the-art QRS detection methods. The method yields high sensitivity, positive predictivity, and a low detection error rate of 99.85%, 99.91%, and 0.25%, respectively, for the MIT-BIH arrhythmia and 99.98%, 99.90%, and 0.12%, respectively for the Fantasia database.

Heartbeats are important aspects for the study of heart diseases in medical sciences as they provide vital information on heart disorders and abnormalities in heart rhythms. Each heartbeat provides a QRS complex in the electrocardiogram (ECG) which is centered at the R-peak. The analysis of ECG is hindered by low-frequency noise, high-frequency noise, interference from P and T waves, and changes in QRS morphology. This paper presents a new peak detection algorithm that can suppress the noise and adapt to changes in ECG signal morphology for better detection performance. The proposed algorithm is based on median and moving average (MA) filtering, segmentation, time and amplitude thresholds, and statistical false peak elimination (SFPE). The filters are first used in preprocessing to reduce unwanted noise and interference. The data is then divided into smaller segments and each segment is then analyzed using two distinct thresholds, a time axis (x-axis) threshold and an amplitude (y-axis) threshold. Next, the false peaks are eliminated resulting from any residue of noise using an average value of peak-to-peak interval. A post-processing stage is added to eliminate any peak that is detected twice and to search for missed low-amplitude peaks. The proposed method is tested on MIT-BIH arrhythmia and Fantasia databases and provides better results in comparison to several state-of-the-art methods in the field. The mean sensitivity, positive predictivity, and detection error rates for the proposed method are 99.82%, 99.88%, and 0.31%, respectively, for the MIT-BIH arrhythmia database and 99.92%, 99.90%, and 0.18%, respectively, for the Fantasia database.

The study of heartbeats in electrocardiogram (ECG) signals is very important to sustain good health. Any anomalies in the heart rhythm can be detected by carefully studying the ECG signal. The detection of the QRS is obstructed by external and internal sources of noise. Automatic detection of the QRS is achieved by diminishing these noises to a minimum by different types of filtering such as band-pass filtering, wavelet transform, and applying thresholds. This paper presents a new method of QRS detection using discrete wavelet transform (DWT), median filtering, and adaptive multilevel thresholding (AMT). The proposed method is tested for the MIT-BIH Arrhythmia database and shows a high sensitivity of 99.74%, positive predictivity of 99.88%, and a detection error rate of 0.38%. In addition to this, the proposed technique is quite robust and can adapt to signals with a low signal-to-noise ratio.

Power electronics is a means of controlling/processing in an appropriate way that is suitable for the load. Power conditioning is done to ensure higher efficiency and power density. Improved efficiency can reduce emission of heat/wasted power so that it can reduces environmental pollution. Power electronic converters use semi-conductor switching devices that are operated in ON-OFF states. These converters have higher efficiency with the control of distorted input signals. The input signal distortion increases with the increased use of these converters. Various standards have been set to limit the level of distortion. One of the ways of measuring these distortions is THD. A signal having THD more than the specified range has some detrimental effect on the line to which it is connected. Lighting loads are one of the major powering applications. Approximately 19% to 20% of the power is used for this purpose. As a result high efficient lighting system is a major research area in power electronics. In this regard incandescent bulbs were replaced by discharge lights (those are operated in negative resistance region) long ago. A ballast circuit is included to limit the current flow in this circuit. Initially magnetic ballasts were used. However due to some disadvantages like flickering, size, high cost, low efficiency etc electronic ballast has taken place instead of magnetic ballast. The focus of this thesis is in the electronic ballasts in discharge lamps and the total harmonic distortion produced due to the non-linear circuit situated in the lamps. The main objective is to minimize this distortion as far as possible. Various simulations have been done in order to measure the total harmonic distortion of ballast circuit situated on the CFL bulb. Moreover a power factor rectification circuit has been provided to improve the power factor and reduce the THD value of the ballast circuit.