Search

A recent report from researchers at Birmingham Women's and Children's NHS Foundation Trust discusses the diagnosis and treatment of progressive familial intrahepatic cholestasis (PFIC), a rare liver disorder that primarily affects children. The researchers aimed to provide recommendations for the management of PFIC in clinical practice. They developed an algorithm for the diagnosis and treatment of children with suspected PFIC, which includes the use of licensed inhibitors of ileal bile acid transporters as the first-line treatment. The authors hope that these recommendations will help standardize the management of PFIC and raise awareness of current developments in the field. [Extracted from the article]

A recent study conducted by researchers at Hamad Bin Khalifa University in Doha, Qatar, focuses on the use of machine learning and digital signal processing to improve the analysis of electroencephalogram (EEG) signals for medical diagnosis. EEG is a safe and common test that records electrical signals from the brain and has various applications in diagnosing conditions such as epilepsy, Alzheimer's, brain tumors, sleep disorders, and stroke. The study provides a comprehensive review of recent approaches for EEG preprocessing, feature extraction, and diagnosis of brain disorders, aiming to identify reliable algorithms for improving the accuracy of automatic diagnostic systems. The research concludes that the paper can be a valuable resource for researchers and practitioners in the field, offering insights into recent developments and future research directions. [Extracted from the article]

Researchers at NYU Tandon School of Engineering have developed a new technology that could allow stroke patients to undergo rehabilitation exercises at home. The technology involves using a smartphone strapped to the forearm and a low-cost gaming controller called the Novint Falcon to track wrist movements. By combining data from the smartphone's motion sensors with the controller's measurements through machine learning, the researchers were able to predict wrist angles with over 90% accuracy. This technology has the potential to increase access to post-stroke care and improve the quality of life for stroke patients. [Extracted from the article]

Researchers from National Chung Hsing University in Taiwan have published a study on the detection of white matter hyperintensities (WMHs) in brain magnetic resonance images. WMHs are lesions associated with Alzheimer's disease and cognitive decline. The researchers developed a new algorithm called iterative KCEM (IKCEM) to improve the detection of WMHs. The algorithm was evaluated using two commonly used software packages and showed superiority over existing methods. The study provides valuable insights into the diagnosis of Alzheimer's disease and could contribute to advancements in medical imaging technology. [Extracted from the article]

A recent study conducted by researchers from the University of Minnesota explores the clustering of clinical response in patients with Major Depressive Disorder (MDD) after receiving transcranial magnetic stimulation (TMS) treatment. The researchers propose an algorithm based on hyperdimensional computing (HDC) to determine the number of clusters in a group of clinical responses. The study found that the HDC algorithm revealed different clusters compared to traditional latent class mixed modeling (LCMM), but both algorithms showed a similar clustering pattern. This research provides valuable insights into the variability of TMS treatment and its effect on cognitive control in MDD patients. [Extracted from the article]

A study conducted by researchers at Guangxi University in Nanning, China, explores the challenges of chemotherapy optimization due to toxic side effects, drug resistance, and individual variations in cancers. The researchers propose a new algorithm called TA-MOSSA, which takes into account these factors to provide effective chemotherapy regimens for patients with different treatment approaches. The study emphasizes the importance of precise dosing to improve efficacy and reduce toxicity. The findings of this research can be found in the journal Applied Sciences. [Extracted from the article]

A recent study conducted by researchers from the Icahn School of Medicine at Mount Sinai in New York City explores the use of advanced deep learning algorithms for magnetic resonance imaging (MRI) segmentation. Accurate MRI segmentation is crucial for diagnosing abnormalities and planning treatments, but traditional supervised learning techniques require a large amount of annotated data. The study reviews state-of-the-art algorithms that utilize a limited number of annotated samples, including self-supervised learning, generative models, few-shot learning, and semi-supervised learning. The researchers also discuss future directions in the field, such as contrastive language-image pretraining, to further improve image segmentation with limited labels. [Extracted from the article]

Researchers from Brigham and Women's Hospital in Boston, Massachusetts have published a report on algorithms and software for real-time electric field modeling techniques for transcranial magnetic stimulation (TMS). TMS is a device-based neuromodulation technique used to treat brain diseases. The report discusses modern algorithms for real-time TMS electric field modeling, including basis representation techniques and deep neural-network-based methods. It also reviews software tools that integrate electric field modeling with navigated TMS. The research was supported by the National Institutes of Health (NIH) in the United States. [Extracted from the article]

A researcher from the University of Science and Technology Beijing has conducted a study on the predictive analysis of brain imaging data using deep learning algorithms. The study aimed to improve the accuracy and efficiency of brain imaging data prediction. The researcher developed a batch normalized convolutional neural network (BN-CNN) based prediction method, which showed promising results in predicting missing structural data in brain imaging. The algorithm was applied to epilepsy and Parkinson's cases, achieving an average prediction accuracy of 93.37%. The study suggests that deep learning algorithms have the potential to greatly benefit clinical diagnosis and treatment in the future. [Extracted from the article]

A new study conducted by researchers at Quantum University focuses on hydrocephalus, a central nervous system disorder that primarily affects infants and toddlers. The study aims to improve the accuracy of hydrocephalus diagnosis by developing an algorithm called H-detect. This algorithm utilizes image processing techniques and the CatBoost algorithm to predict hydrocephalus more effectively than traditional methods. The research concludes that the H-detect strategy outperforms conventional approaches. For more information, readers can refer to the journal article "H-Detect: an Algorithm for Early Detection of Hydrocephalus" published in Informatika i avtomatizacia. [Extracted from the article]

Researchers from Charles Sturt University (CSU) in Bathurst, Australia have developed an algorithm called the Enhanced Real-time Video (ERV) algorithm to improve the transmission of real-time video in surgical tele-education. The algorithm aims to reduce delays and improve video quality by optimizing bandwidth usage and choosing the most reliable communication route. Comparative analyses show that the ERV algorithm significantly improves the Peak Signal-to-Noise Ratio (PSNR) of videos and reduces processing time. The research concludes that the ERV method enhances transmission efficiency and improves the quality of real-time video in surgical tele-education. [Extracted from the article]

A recent report from the School of Electronics Engineering discusses research on brain cancer and the challenges of accurately separating tumor regions from healthy tissue in medical imaging. The researchers propose a novel semantic segmentation architecture called IC-Net, which combines elements from various models to improve brain tumor segmentation in MRI images. The IC-Net design outperforms other segmentation techniques, achieving notable outcomes in accuracy, loss, specificity, sensitivity, and Dice similarity coefficient. This research has the potential to enhance the accuracy of brain tumor segmentation and improve medical imaging analysis in the field of oncology. [Extracted from the article]

A recent study conducted by researchers at the University of Maryland explores the development and analysis of a context-aware decision-support algorithm for burn resuscitation. The algorithm, known as the Cardiac Output-Enabled Burn Resuscitation Decision-Support (CaRD) algorithm, combines a surrogate of cardiac output with urinary output to potentially improve the efficacy of burn resuscitation. The study found that while the CaRD algorithm showed promise, further development and tuning are necessary to optimize its performance. This research provides valuable insights into the field of burn resuscitation and highlights the challenges and potential for future algorithm development. [Extracted from the article]

A recent report from Universitas Multimedia Nusantara highlights the global concern of breast cancer, which is the leading cancer type in terms of new cases and among the top five cancer types with the most deaths. The research aims to improve the identification of breast cancer comorbidities and ultimately reduce the number of deaths from this condition. The study utilizes network analysis and community detection algorithms to identify comorbidities associated with breast cancer. The findings reveal several diseases that are commonly associated with breast cancer, providing valuable insights for medical professionals. [Extracted from the article]

A new study conducted by researchers at R.M.K. Engineering College in Tamil Nadu, India, explores the use of deep learning classification to assess biomedical data for the detection and classification of fetal arrhythmia from ECG signals. The study proposes a Deep Siamese domain adaptation convolutional Neural Network optimized with Honey Badger Algorithm (DSDACNN-HBA-FAC) to overcome the problem of data imbalance. The researchers utilized coherence shock filtering and the improved non-subsampled Shearlet transform method to preprocess the ECG data and extract features for classification. The proposed method achieved higher accuracy, lower computation time, and greater AUC compared to existing techniques. [Extracted from the article]

A new study conducted at the University of Kastamonu in Turkey has developed an innovative method for the automated detection of colon cancer using artificial intelligence. The researchers utilized a convolutional neural network (CNN) and the Fishier Mantis Optimizer algorithm to improve the accuracy and efficiency of colon cancer diagnosis. The proposed method was evaluated using a comprehensive dataset of colon cancer images and demonstrated superior performance compared to traditional diagnostic approaches. The integration of deep learning techniques and bio-inspired optimization algorithms holds promise for enhancing early detection and diagnosis of colon cancer. [Extracted from the article]

A recent study conducted at Kanazawa University in Japan explores the use of machine learning algorithms for the diagnosis of cervical cancer. The researchers propose a novel feature selection strategy called the Binary Harris Hawks Optimization (BHHO) algorithm, which aims to reduce the number of features used in diagnosis while improving accuracy. The BHHO algorithm demonstrates better stability and accuracy compared to other common wrapper-based feature selection methods, not only on cervical cancer datasets but also on other disease datasets. This research contributes to the ongoing efforts to improve early diagnosis and treatment of cervical cancer, which is one of the leading causes of cancer-related deaths among females worldwide. [Extracted from the article]

Researchers at the Beijing Institute of Technology have developed a multi-LiDAR human joint recognition algorithm to monitor the physical condition of patients in hospital wards and prevent convulsions and falls in the absence of medical staff. The algorithm addresses challenges such as object occlusion, human self-occlusion, and complex backgrounds. The improved model achieved a mean average precision of 91.6%, outperforming the original model and a comparison algorithm. The speed of the improved model also meets the requirements for real-time target detection. [Extracted from the article]

A recent study conducted by researchers at the University of Granada in Spain has developed a new app called PaiNEd that aims to assess pain experience in survivors of breast cancer. The app includes an algorithm that determines whether patients are experiencing pain predominantly related to central sensitization (CS) symptoms. The study found that the app demonstrated reliability in evaluating pain experience and could be a valuable tool for healthcare professionals involved in pain education programs. This research provides a potential solution for assessing and managing pain in breast cancer survivors. [Extracted from the article]

Researchers at Northern Border University have conducted a study comparing machine learning algorithms for investigating complications related to myocardial infarction (heart attack). The study aimed to classify these complications and predict fatal outcomes. The algorithms tested were Multilayer Perceptron (MLP), Naive Bayes (NB), and Decision Tree (DT). The results showed that the MLP classifier performed best with the full dataset, while the DT classifier was more favorable when the dataset size was reduced using an attribute selection algorithm. This research provides valuable insights into the use of machine learning in understanding and predicting myocardial infarction complications. [Extracted from the article]

A recent report from Universiti Malaya discusses research on cancer treatment using optimal control supported by swarm intelligence. The study aims to develop hybrid techniques that combine optimal control theory with evolutionary algorithms and multi-objective swarm algorithms to reduce cancerous cells while minimizing the use of chemotherapy medications and toxicity. Two hybrid techniques were proposed and evaluated, with MOEA/D emerging as the most effective algorithm in reducing tumor size and cancer drug concentration. The study highlights the importance of integrating optimal control theory and evolutionary algorithms for optimizing cancer chemotherapy treatment. [Extracted from the article]

A patent application has been submitted for a device and method to assist with the detection and classification of bone fractures based on medical images. The device uses algorithms and neural networks to identify and localize fractures within the bone structure. It can also suggest possible treatment options based on the classification of the fracture. The device can be used manually with user input or automatically with algorithm support. The invention aims to provide assistance to surgeons in detecting and classifying bone fractures, ultimately improving patient care. [Extracted from the article]

Researchers at Yildiz Technical University have conducted a comprehensive review of cerebral emboli detection algorithms. Stroke is a leading cause of death worldwide, and cerebral emboli is considered one of the main sources of stroke. The researchers discuss the physics of Doppler ultrasound and review various signal processing-based algorithms used for emboli detection. They also highlight the use of animal models to understand the nature of emboli development. The research provides valuable insights into the detection and understanding of cerebral emboli. [Extracted from the article]

A recent report from the University of Nebraska Medical Center discusses research on epilepsy and the localization of eloquent cortices, which is important for neurosurgical applications such as epilepsy and tumor resections. The researchers propose a method to post-process distributed source localization results to identify the area and boundary of the cortical tissue that primarily activates in response to somatosensory stimulation. The algorithm successfully localized somatosensory responses on the cortical surface in anatomically appropriate regions, reducing subjectivity associated with dipole localization. The research concludes that the algorithm is not inferior to dipole localization and provides additional pre-surgical information on the location and spatial distribution of eloquent cortical areas. [Extracted from the article]

Researchers from Quzhou University have developed a new algorithm called the quasi-oppositional learning-based double Gaussian sampling bare-bone imperialist competitive algorithm (QOLBDGSBB-ICA) for global optimization. This algorithm improves upon the Gaussian bare-bone imperialist competitive algorithm (GBB-ICA) by introducing a double Gaussian sampling strategy and incorporating the quasi-oppositional learning technique. The QOLBDGSBB-ICA algorithm was tested on benchmark functions and engineering design problems, and it outperformed other referenced ICA variants. This research contributes to the field of computational intelligence and offers a more effective algorithm for global optimization. [Extracted from the article]

A recent report from Sun Yat-sen University in Guangzhou, China discusses the research on natural killer cells (NKCs) and their potential in cancer immunotherapy. The researchers utilized deep learning techniques to analyze the morphology and motion characteristics of NKCs in tumor contexts. They proposed an improved two-shot tracking algorithm, Distance Cascade Matching and the Re-Search method, to overcome the challenges of tracking small NKCs in complex backgrounds. The study found that polarized NKCs are more likely to move along their polarization direction and approach cancer cells, and under equilibrium, NKCs display a Boltzmann distribution on the cancer cell surface. This research provides valuable insights into the dynamics of NKCs and their interactions with cancer cells. [Extracted from the article]

Researchers from the University of Malaysia Perlis have developed a new method for detecting and segmenting the nucleus cell region in Pap smear images for cervical cancer screening. The proposed method, which includes color adjustment, k-means, and a modified Bradley algorithm, has been found to significantly and efficiently segment the nucleus cell region compared to the original Bradley algorithm. The researchers achieved high accuracy, sensitivity, and F-measure results, indicating the effectiveness of their approach. This research provides a new perspective on improving the convenience and accuracy of cervical cancer screening. [Extracted from the article]

Researchers from Regeneron Pharmaceuticals Inc. conducted a study on the application of multiple validated algorithms for identifying incident breast cancer among individuals with atopic dermatitis (AD). They tested two algorithms in multiple insurance claims databases and found that Algorithm 2 provided similar estimates to those of the Surveillance, Epidemiology, and End Results (SEER) program. However, they noted that the replicability of these algorithms in different data sources or subpopulations is rarely tested, and caution should be exercised when using them to calculate incidence or prevalence estimates. The study was supported by Sanofi and published in Pharmacoepidemiology and Drug Safety. [Extracted from the article]

A recent study conducted by researchers at Shanghai University of Electric Power explores the use of privacy-preserving algorithms in breast cancer prediction. The researchers highlight the need for pre-diagnosis testing to alleviate the strain on healthcare resources. They propose a highly optimized two-party logistic regression algorithm based on the CKKS scheme, which aims to enhance the accuracy and efficiency of pre-diagnosis testing while upholding data privacy. Simulation experiments demonstrate that the model's prediction accuracy exceeds 96% for two-sided encrypted data. This research contributes to the field of machine learning in healthcare and offers potential solutions for collaborative training of models while protecting private data. [Extracted from the article]

A research study conducted in Bucharest, Romania, analyzed the management of ectopic mediastinal thyroid (EMT) in relation to cancer and non-malignant findings. The study reviewed 89 original articles and classified them into four groups: studies/case series, malignant EMTs, benign EMTs without thyroid anomalies, and benign EMTs with thyroid anomalies. The researchers proposed a 10-item algorithm to navigate the EMT domain. The study found that the clinical index of suspicion for EMT remains low, but the rate of cancer is higher than previously reported. Surgery was found to have a good overall outcome. [Extracted from the article]

Researchers at Fuzhou University in Fujian, China have developed a deep learning model fusion algorithm for the diagnosis of gastric mucosa-associated lymphoid tissue (MALT) lymphoma. The algorithm uses hematoxylin-eosin (H&E) whole slide images (WSIs) to accurately diagnose MALT lymphoma, which is a type of malignant tumor originating from the lymphohematopoietic system. The proposed framework achieved an accuracy of 98.53% using image patches and 94.96% on 258 WSIs, demonstrating its potential use in clinical practice. The research was funded by the National Natural Science Foundation of China and the Fujian Health Commission. [Extracted from the article]

A researcher from the Pan African University Institute for Basic Sciences has discussed their research on improving magnetic resonance imaging (MRI) using compressed sensing and adaptive multi-extreme particle swarm optimization. MRI is a powerful technique for examining the body's internal structure, but its lengthy acquisition times can limit its clinical usefulness. The researcher proposes a novel method that combines adaptive multi-extreme particle swarm optimization and dual tree complex wavelet transform to speed up image acquisition in MRI. The results show improvements in image quality and suggest that this method could potentially enable high-quality, real-time MRI imaging, leading to better detection and treatment of medical conditions. [Extracted from the article]

A study conducted by researchers from the Chinese Academy of Sciences discusses a new automatic detection method for bladder tumor cells based on color and shape features. Bladder urothelial carcinoma is a common malignant tumor disease in the urinary system, and this study aims to improve the efficiency of its pathological diagnosis. The researchers propose an algorithm that combines color and shape features extracted from microscopic hyperspectral data, achieving a sensitivity of 0.952, specificity of 0.897, and accuracy of 0.920. This method could potentially assist doctors in choosing treatment plans and predicting the prognosis of bladder urothelial cancer. [Extracted from the article]

Researchers from the School of Information and Control Engineering have developed a bone stick image matching algorithm based on an improved ConvNeXt and Siamese network. The algorithm addresses the challenges posed by the close resemblance in shape and complex texture features of bone sticks, which make traditional image matching methods inefficient and less accurate. The proposed method achieves a higher matching accuracy compared to other algorithms, providing an effective solution for automated bone stick matching. This research has implications for the field of engineering and bone research. [Extracted from the article]

A recent study conducted by researchers at George Emil Palade University of Medicine in Targu Mures, Romania, has developed a second trimester prediction algorithm for early-onset hypertensive disorders of pregnancy (HDP). The algorithm combines biochemical markers, such as the sFlt-1/PlGF ratio, with uterine artery Doppler ultrasound parameters to accurately predict the occurrence and severity of HDP in high-risk pregnant women. The study found that the combination of these markers improved the prediction of severe preeclampsia and its complications compared to using the markers or ultrasound parameters alone. Implementing these algorithms could enhance the accuracy of HDP screening and help manage maternal and fetal complications. [Extracted from the article]

Researchers at Fuzhou University in the People's Republic of China have developed an improved optimal foraging algorithm (OFA) for global optimization. The OFA is a swarm-based algorithm inspired by animal behavioral ecology theory. The researchers introduced a new algorithm called QOS-OFA, which incorporates quasi-opposition-based learning, an efficient cosine-based scale factor, and a search strategy with social behavior. The QOS-OFA was compared to seven other meta-heuristic algorithms and performed better on most test problems. This research has been peer-reviewed and published in the journal Computing. [Extracted from the article]

Researchers at Chalmers University of Technology in Gothenburg, Sweden have developed a framework for using reinforcement learning in de novo drug design. The researchers studied various reinforcement learning algorithms and replay buffers to generate novel molecules predicted to be active against the dopamine receptor DRD2. The findings suggest that using both top-scoring and low-scoring molecules improves performance stability, and replaying molecules can enhance structural diversity and the number of active molecules generated. The research provides an open-source framework for investigating reinforcement learning methods in drug design. [Extracted from the article]

A recent study conducted at the University of Tikrit explores the impact of poisoning attacks on deep learning-based Network Intrusion Detection Systems (NIDSs). The researchers found that these attacks, which aim to distort the functionality of ML systems, can compromise the integrity of smart technology-based systems and render them susceptible to unauthorized access. The study specifically investigates the resilience of deep learning models against untethered white-box attacks by injecting altered instances into training datasets. The results indicate that the poisoning attacks significantly impact the model's performance and are difficult to detect. The findings highlight the vulnerability of DL-based NIDSs and emphasize the need for defense techniques to secure these systems against sophisticated threats. [Extracted from the article]

Researchers from Sohag University have developed a new algorithm, called RBNRO-DE, for gene selection in cancer classification using high-dimensional data. The algorithm was tested on 22 cancer datasets and compared to other meta-heuristic algorithms. The results showed that the RBNRO-DE algorithm achieved optimal convergence and classification accuracy up to 100%, with a feature reduction size of 98%. This research highlights the potential of RNA sequencing as a technique for gene profiling and quantification in cancer research. [Extracted from the article]

A recent study conducted by researchers at Hebei University of Technology in Tianjin, China, addresses the problem of home health care routing and scheduling in a pandemic environment. The study focuses on assigning caregivers to patients and optimizing caregiver routes while minimizing travel costs, workload differences, and negative patient preferences. The researchers propose a customized algorithm called AMOALNS, which combines archived multi-objective simulated annealing with adaptive large neighborhood search. The algorithm generates new solutions iteratively and employs a multi-objective optimization framework. The study concludes that minimizing travel costs is the most challenging objective, while optimizing workload balancing is worthwhile. Sensitivity analyses show that time windows classifications and instance sizes significantly impact the compromising solutions. [Extracted from the article]

Researchers at Lovely Professional University in Phagwara, India have developed a hybrid feature selection technique to improve the speed and performance of machine learning models in diagnosing breast cancer. The technique combines Regularized Neighborhood Component Analysis (RNCA) and Binary Teaching Learning Based Optimization (BTLBO) algorithms to select the best features for classification. The results of the study showed that the hybrid model outperformed other feature selection methods in terms of classification accuracy, recall, and AUC measures. This research has been peer-reviewed and published in The International Arab Journal of Information Technology. [Extracted from the article]

Researchers at the University of Belgrade have developed a logic-based machine learning algorithm called IBA-VNS that aims to improve the interpretability and explainability of machine learning models. The algorithm incorporates interpolative Boolean algebra (IBA) and variable neighborhood search (VNS) to create logic aggregation functions that are easy to understand and provide additional insight to decision-makers. The researchers applied the IBA-VNS algorithm to predict hospital length of stay (LOS) for colorectal surgeries and achieved exceptional results in terms of interpretability and accuracy. The algorithm does not require domain knowledge and can be applied in various domains. [Extracted from the article]

Researchers at Zhejiang University of Technology in China have developed an intelligent electrocardiogram (ECG) recognition method for the rapid diagnosis and evaluation of cardiovascular diseases. The method utilizes a convolutional neural network (CNN) and support vector machines (SVM) with an improved antlion algorithm (ALO) to classify ECG signals in real time. The algorithm achieved an average accuracy, sensitivity, and specificity of 99.97%, 99.97%, and 99.99%, respectively. This approach shows promise for improving the recognition performance of cardiovascular diseases compared to existing methods. [Extracted from the article]

Researchers at Yanshan University in Qinhuangdao, China have developed an improved fault diagnosis algorithm for extracting features and diagnosing periodic transient impact faults. The algorithm combines a modified fast kurtogram method with the lightweight convolutional neural network GhostNet. By converting vibration signals into two-dimensional kurtosis graphs, the algorithm achieves accurate fault diagnosis and classification with improved computational efficiency and resource utilization. The research was supported by the National Natural Science Foundation of China and the Province Natural Science Foundation of Hebei. [Extracted from the article]

A recent study conducted by researchers at the University of Zurich explores the concept of the "Right to Audit" in algorithm auditing. The researchers discuss power asymmetries in the field, including legal issues, disparities between early-career and senior researchers, and other important asymmetries related to access to data, targets of auditing, and user rights. They also highlight the implications of these asymmetries, such as Western-centrism and a lack of diversity in perspectives. The study suggests that these asymmetries not only affect algorithm auditing but also Computational Social Science as a whole and should be addressed. [Extracted from the article]

A researcher from Shanghai Dianji University has conducted a study on algorithms in the field of prosthetics. The study focuses on improving the interaction between a shoulder-disconnected upper limb prosthesis and the environment. The researcher proposes a reinforcement learning-based method for controlling the forearm interaction of the prosthesis, which reduces uncertainty and ensures stability during interaction. Simulation experiments showed positive results, indicating that the control method effectively improved the interaction between the robotic arm and the environment. The study provides valuable insights into the development of more advanced and user-friendly prosthetic devices. [Extracted from the article]

Researchers from the Universidad Autonoma de Manizales have conducted a comparative study on the use of artificial intelligence (AI) algorithms to classify epileptic seizures. The study aimed to optimize the diagnosis process, which currently relies on lengthy electroencephalogram (EEG) tests. By comparing traditional machine learning models with cutting-edge models like Capsule-Net architectures and Transformer Encoder, the researchers found that AI models achieved high accuracy in classifying EEG signals. The study concludes that AI is essential in diagnosing epilepsy and that state-of-the-art models outperform conventional models, although data processing also plays a crucial role in evaluating accuracy. [Extracted from the article]

Researchers from the University of Sydney have conducted a scoping review to assess the efficacy of machine learning (ML) algorithms in detecting voice disorders. The review aims to identify the critical issues that have hindered the use of ML algorithms in clinical settings by examining standard audio tasks, acoustic features, processing algorithms, and environmental factors that affect their efficacy. While ML algorithms have shown high accuracy in detecting voice disorders, none have been deemed reliable enough for clinical use. The findings of this review will be disseminated through peer-reviewed publications related to voice pathology. [Extracted from the article]

A recent study conducted at Henan University in China has proposed a deep learning model called SE-DenseNet for the early diagnosis and detection of Alzheimer's disease (AD) and mild cognitive impairment (MCI). The model combines channel attention and dense connectivity networks to extract high-quality features from magnetic resonance imaging (MRI) data. The model was validated on the Alzheimer's Disease Neuroimaging Initiative (ADNI) dataset and achieved high accuracy, sensitivity, and specificity in classifying AD-NC, AD-MCI, NC-MCI, and AD-NC-MCI cases. The study suggests that the proposed model could be a powerful tool for the early intervention and treatment of AD. [Extracted from the article]

Researchers at Shandong University in Jinan, China have developed an automatic segmentation algorithm for conglutinated bone fragments in 3D CT images of lower limb fractures. The algorithm aims to address the over-segmentation and error-segmentation problems commonly encountered with existing automatic segmentation algorithms. By using a pixel classification and clustering method based on morphology, as well as a concave points detection method, the algorithm achieves good segmentation accuracy for separated and conglutinated bone fragments. The research concludes that this algorithm significantly improves the segmentation accuracy and efficiency of bone fragments under complex fracture conditions. [Extracted from the article]