Novel Meta Attention Guided Framework for Breast Abnormality Classification With Combination of FSL and DA

Detecting lung abnormalities via chest X-rays is challenging due to understated tissue variations often ignored by traditional methods. Augmentation techniques like rotation or flipping risk distorting critical anatomical features, actually leading to misdiagnosis. This paper proposes a novel two-stage ASCE (Anatomical Segmentation and Color-Based Enhancement) framework for precise and efficient classification of lung abnormalities while preserving anatomical integrity.Stage 1 classifies Normal vs. Pneumonia with 95 % accuracy, an AUC of 0.98, and an F1-score of 0.92. Stage 2 distinguishes Pneumonia into Viral and Bacterial subtypes with 100 % accuracy and F1-score. This approach integrates segmentation and tissue-specific color enhancements with Kullback-Leibler (KL) divergence, quantifying deviations from healthy lung regions for improved classification. The lightweight pipeline ensures computational efficiency (∼0.06s/image) and clinical interpretability by preserving diagnostic features, enhancing visibility, and enabling quantitative analysis.1.Preserving Anatomical Structures: The methodology ensures that diagnostic features are preserved and highlighted with Anatomy-Preserved Segmentation2.Enhancing Diagnostic Visibility: The system employs targeted colour-based enhancement that improves the visibility of potential abnormalities3.Quantitative Analysis with Kullback-Leibler (KL) divergence: The model enhances precise identification of abnormal tissue by comparing the probability distributions of healthy lungs and abnormal areas

Breast cancer is the most common form of cancer among women globally. Detecting a tumor at its early stages is very crucial for a higher possibility of successful treatment. Cancerous cells have high metabolic rate which generate more heat compared to healthy tissue and will be transferred to the skin surface. Thermography technique has distinguished itself as an adjunctive imaging modality to the current gold standard mammography approach due to its capability in measuring the heat radiated from the skin surface for early detection of breast cancer. It provides an additional set of functional information, describing the physiological changes of the underlying thermal and vascular properties of the tissues. However, the thermography technique is shown to be highly dependent on the trained analyst for image interpretation and most of the analyses were conducted qualitatively. Therefore, the current ability of this technique is still limited especially for massive screening activity. This chapter presented a proposed technical framework for automatic segmentation and classification of abnormality on multiple in vivo thermography-based images. A new two-tier automatic segmentation algorithm was developed using a series of thermography screening conducted on both pathological and healthy Sprague-Dawley rats. Features extracted show that the mean values for temperature standard deviation and pixel intensity of the abnormal thermal images are distinctively higher when compared to normal thermal images. For classification, Artificial Neural Network system was developed and produced a validation accuracy performance of 92.5% for thermal image abnormality detection. In conclusion, this study has successfully demonstrated that for massive or routine screening activities, the proposed technical framework could provide a highly reliable clinical decision support to the clinicians in making a diagnosis based on the medical thermal images.

Wirekess body area network(WBAN) has become a promising type of networks to efficiently collecting as well as analyzing human physiological information. These collected data are then transformed to outside database support real-time monitoring or other health applications. However, due to the resource constrained of the networks, the monitored data needs to be analyzed for automatic detection of physiological changes indeed trigger an alarm of patient health degradation. Moreover, false alarm result by abnormal changes or faulty measurements should be classified to reduce unnecessary medical intervention.. In this paper, we propose a novel framework for data abnormal detection and classification on real-time ECG signals to distinguish faulty measurements form clinical emergency. Our work is based on adaptive 1-D convolutional neural networks(CNNs). We use small common ECG dataset to build a dedicated CNN models offline, which can be used to classify possibly long ECG data stream in a fast and accurate manner, such a solution can conveniently be used for real-time monitoring and early alert system in WBANs.The state-of-the-art performance on efficiency and accuracy for ECG classification over real dataset is achieved by the proposed method.

The visual similarity between normal breast tissues and abnormal lesions in digital mammogram images makes computer-aided diagnosis of breast cancer using automatically detected features a highly error-prone task. Our contribution in this paper is a novel feature reduction framework for selecting the most discriminative features that achieves both efficiency and classification accuracy. Our approach applies five individual feature-ranking methods including Fisher score, minimum redundancy-maximum relevance, relief-f, sequential forward feature selection, and genetic algorithm for sorting the extracted features and selecting the features with highest ranking to setup a classifier. Our method achieves an accuracy of 94.27% and a sensitivity of 98.36% with a specificity of 99.27% on a set of 1,100 mammogram patches taken from image retrieval in medical applications database using a neural network classifier, which competes with state-of-the-art classification accuracy 93.11%. Furthermore, we demonstrate that only 49 out of the 119 extracted features are sufficient to achieve the reported accuracy of normal vs. abnormal classification.

Due to rapid development in Artificial Intelligence (AI) and Deep Learning (DL), it is difficult to maintain the security and robustness of these techniques and algorithms due to emergence of novel term adversary sampling. Such technique is sensitive to these models. Thus, fake samples cause AI and DL model to produce diverse results. Adversarial attacks that successfully implemented in real world scenarios highlight their applicability even further. In this regard, minor modifications of input images cause “Adversarial Attacks” that altered the performance of competing attacks dramatically. Recently, such attacks and defensive strategies are gaining lot of attention by the machine learning and security researchers. Doctors use different kinds of technologies to examine the patient abnormalities including Wireless Capsule Endoscopy (WCE). However, using WCE it is very difficult for doctors to detect an abnormality within images since it takes enough time while inspection and deciding abnormality. As a result, it took weeks to generate patients test report, which is tiring and strenuous for them. Therefore, researchers come out with the solution to adopt computerized technologies, which are more suitable for the classification and detection of such abnormalities. As far as the classification is concern, the adversarial attacks generate problems in classified images. Now days, to handle this issue machine learning is mainstream defensive approach against adversarial attacks. Hence, this research exposes the attacks by altering the datasets with noise including salt and pepper and Fast Gradient Sign Method (FGSM) and then reflects that how machine learning algorithms work fine to handle these noises in order to avoid attacks. Results obtained on the WCE images which are vulnerable to adversarial attack are 96.30% accurate and prove that the proposed defensive model is robust when compared to competitive existing methods.

Abstract: The detection of abnormalities in medical images plays a pivotal role in early diagnosis and treatment. This paper presents a hybrid approach that combines K-Nearest Neighbors (KNN) and deep learning techniques to improve medical image classification and anomaly detection. The method applies KNN for classifying images such as MRI, CT, and X-ray scans, focusing on abnormality detection in brain images. By integrating KNN classifiers with feature extraction methods, the approach addresses challenges such as class imbalance and small datasets, resulting in improved detection accuracy. The effectiveness of the proposed method is demonstrated on a medical image dataset, showing significant improvements in both classification and anomaly detection tasks.

Breast cancer, impacting both men and women, though it is notably more prevalent in women, ranks among the most widespread types of cancer worldwide. Early detection of breast cancer significantly improves treatment outcomes and increases the likelihood of successful recovery, underscoring the importance of regular screenings and prompt medical attention upon noticing any changes or abnormalities in breast health. Utilizing a Computer-Aided Diagnosis (CAD) system further enhances early detection capabilities, assisting clinicians in accurate diagnosis and treatment planning. Unlike existing binary methods, the proposed method employs a three-class classification system distinguishing between mass, micro calcification, and normal tissues. Initial preprocessing is followed by patch extraction and augmentation, playing a pivotal role in mitigating overfitting and enabling the model to learn robust features that generalize effectively to new medical images. A deep-learning architecture, with various pre-trained models includes R<tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$e$</tex>sNet50, ResNetl0l, ResNetl52, Inception ResNet V2, Inception V3, VGG-16, and VGG-19, is trained on the augmented dataset to classify patches into normal, micro calcification, and mass categories. The proposed model is validated on the combined dataset which demonstrates the model's accuracy in diagnosing breast abnormalities, with Inception ResN et V2 achieving an accuracy of 97.93 %, outperforming other pre-trained models. The system achieved an overall sensitivity, specificity, precision, and Fl-score approximate 0.9411, 0.9885, 0.9740 and 0.9557, respectively, illustrating the model's robustness and its ability to reliably detect and classify breast abnormalities.

Chapter 7 - Big data classification with IoT-based application for e-health care

Radiographs are used as the most important imaging tool for identifying spine anomalies in clinical practice. The evaluation of spinal bone lesions, however, is a challenging task for radiologists. This work aims at developing and evaluating a deep learning-based framework, named VinDr-SpineXR, for the classification and localization of abnormalities from spine X-rays. First, we build a large dataset, comprising 10,468 spine X-ray images from 5,000 studies, each of which is manually annotated by an experienced radiologist with bounding boxes around abnormal findings in 13 categories. Using this dataset, we then train a deep learning classifier to determine whether a spine scan is abnormal and a detector to localize 7 crucial findings amongst the total 13. The VinDr-SpineXR is evaluated on a test set of 2,078 images from 1,000 studies, which is kept separate from the training set. It demonstrates an area under the receiver operating characteristic curve (AUROC) of 88.61% (95% CI 87.19%, 90.02%) for the image-level classification task and a mean average precision (mAP@0.5) of 33.56% for the lesion-level localization task. These results serve as a proof of concept and set a baseline for future research in this direction. To encourage advances, the dataset, codes, and trained deep learning models are made publicly available.

Big trajectory data feature analysis for mobile networks is a popular big data analysis task. Due to the large coverage and complexity of the mobile networks, it is difficult to define and detect anomalies in urban motion behavior. Some existing methods are not suitable for the detection of abnormal urban vehicle trajectories because they use the limited single detection techniques, such as determining the common patterns. In this study, we propose a framework for urban trajectory modeling and anomaly detection. Our framework takes into account the fact that anomalous behavior manifests the overall shape of unusual locations and trajectories in the spatial domain as well as the way these locations appear. Therefore, this study determines the peripheral features required for anomaly detection, including spatial location, sequence, and behavioral features. Then, we explore sports behaviors from the three types of features and build a taxi trajectory model for anomaly detection. Anomaly detection, including sports behaviors, are (i) detour behavior detection using an algorithm for global router anomaly detection of trajectories having a pair of same starting and ending points; this method is based on the isolation forest algorithm; (ii) local speed anomaly detection based on the DBSCAN algorithm; and (iii) local shape anomaly detection based on the local outlier factor algorithm. Using a real-life dataset, we demonstrate the effectiveness of our methods in detecting outliers. Furthermore, experiments show that the proposed algorithms perform better than the classical algorithm in terms of high accuracy and recall rate; thus, the proposed methods can accurately detect drivers’ abnormal behavior.

A framework for ship abnormal behaviour detection and classification using AIS data

Cervical cancer is one of the fastest growing global health problems and leading cause of mortality among women of developing countries. Automated Pap smear cell recognition and classification in early stage of cell development is crucial for effective disease diagnosis and immediate treatment. Thus, in this article, we proposed a novel internet of health things (IoHT)-driven deep learning framework for detection and classification of cervical cancer in Pap smear images using concept of transfer learning. Following transfer learning, convolutional neural network (CNN) was combined with different conventional machine learning techniques like K nearest neighbor, naive Bayes, logistic regression, random forest and support vector machines. In the proposed framework, feature extraction from cervical images is performed using pre-trained CNN models like InceptionV3, VGG19, SqueezeNet and ResNet50, which are fed into dense and flattened layer for normal and abnormal cervical cells classification. The performance of the proposed IoHT frameworks is evaluated using standard Pap smear Herlev dataset. The proposed approach was validated by analyzing precision, recall, F1-score, training–testing time and support parameters. The obtained results concluded that CNN pre-trained model ResNet50 achieved the higher classification rate of 97.89% with the involvement of random forest classifier for effective and reliable disease detection and classification. The minimum training time and testing time required to train model were 0.032 s and 0.006 s, respectively.

In most maternity hospitals, an ultrasound scan in the mid-trimester is now a standard element of antenatal care. More fetal abnormalities are being detected in scans as technology advances and ability improves. Fetal anomalies are developmental abnormalities in a fetus that arise during pregnancy, birth defects and congenital abnormalities are related terms. Fetal abnormalities have been commonly observed in industrialized countries over the previous few decades. Three out of every 1000 pregnant mothers suffer a fetal anomaly. This research work proposes an Adaptive Stochastic Gradient Descent Algorithm to evaluate the risk of fetal abnormality. Findings of this work suggest that proposed innovative method can successfully classify the anomalies linked with nuchal translucency thickening. Parameters such an accuracy, recall, precision, and F1-score are analyzed. The accuracy achieved through the suggested technique is 98.642.%.

Medical image based breast cancer diagnosis: State of the art and future directions

Coronary artery disease (CAD) is the most common cardiovascular disease. Risk factors impact CAD progression. Diagnostic and therapy methods for this illness include several costly side effects. Consequently, researchers are seeking economical and precise techniques for diagnosing this condition. Machine learning algorithms may assist doctors in the early diagnosis of the condition. Hence this work presents an efficient approach for feature selection and classification of abnormal heart rate patterns by combining Joint Mutual Information (JMI), Quantum Annealing, and a Bayesian ensemble model using CatBoost and XGBoost classifiers. The method starts with Joint Mutual Information to rank features based on their dependency with the target variable, identifying the most informative features for classification. Quantum Annealing, specifically simulated annealing in this case, is then used to optimize the subset of features by exploring the feature space and selecting the most relevant combinations, thus improving the model's performance by avoiding suboptimal solutions. The selected characteristics are then input into a Bayesian ensemble of CatBoost and XGBoost classifiers, which are trained to forecast heart rate irregularities. This ensemble method integrates the advantages of gradient boosting models to improve forecast accuracy and mitigate overfitting. The proposed technique is termed Probability-based Bayesian Statistics with Ensemble Boost (PBSEnsBoost), achieving an accuracy of 98.2%, specificity of 93.6%, sensitivity of 92.6%, and an F1-score of 95.7%.