Dementia (DEM) is a growing global health challenge, underscoring the need for early and accurate diagnosis. Electroencephalography (EEG) provides a non-invasive window into brain activity, but conventional methods struggle to capture its transient complexity. We present the \textbf{EEG Microstate Analysis Framework (EEG-MSAF)}, an end-to-end pipeline that leverages EEG microstates discrete, quasi-stable topographies to identify DEM-related biomarkers and distinguish DEM, mild cognitive impairment (MCI), and normal cognition (NC). EEG-MSAF comprises three stages: (1) automated microstate feature extraction, (2) classification with machine learning (ML), and (3) feature ranking using Shapley Additive Explanations (SHAP) to highlight key biomarkers. We evaluate on two EEG datasets: the public Chung-Ang University EEG (CAUEEG) dataset and a clinical cohort from Thessaloniki Hospital. Our framework demonstrates strong performance and generalizability. On CAUEEG, EEG-MSAF-SVM achieves \textbf{89\% $\pm$ 0.01 accuracy}, surpassing the deep learning baseline CEEDNET by \textbf{19.3\%}. On the Thessaloniki dataset, it reaches \textbf{95\% $\pm$ 0.01 accuracy}, comparable to EEGConvNeXt. SHAP analysis identifies mean correlation and occurrence as the most informative metrics: disruption of microstate C (salience/attention network) dominates DEM prediction, while microstate F, a novel default-mode pattern, emerges as a key early biomarker for both MCI and DEM. By combining accuracy, generalizability, and interpretability, EEG-MSAF advances EEG-based dementia diagnosis and sheds light on brain dynamics across the cognitive spectrum.

EEG signals capture brain activity with high temporal but low spatial resolution, supporting applications such as neurological diagnosis, cognitive monitoring, and brain-computer interfaces. However, effective analysis remains challenging due to limited labeled data, high dimensionality, and the lack of scalable models that fully capture spatiotemporal dependencies. Existing self-supervised learning (SSL) methods often focus on either spatial or temporal features in isolation, leading to suboptimal representations. To this end, we propose EEG-VJEPA, a novel adaptation of the Video Joint Embedding Predictive Architecture (V-JEPA) for EEG classification. By treating EEG as video-like sequences, EEG-VJEPA learns semantically meaningful spatiotemporal representations using joint embeddings and adaptive masking. To our knowledge, this is the first work that exploits V-JEPA for EEG classification and explores the visual concepts learned by the model. EEG-VJEPA achieves state-of-the-art performance on the publicly available Temple University Hospital (TUH) Abnormal EEG dataset, outperforming both self-supervised and fully supervised baselines. Likewise, we demonstrate the model's good generalization ability on an independent, smaller clinical dataset from the General Hospital of Thessaloniki, involving dementia classification. Keywords: Electroencephalography (EEG), Joint Embedding Predictive Architecture (JEPA), Vision Transformer (ViT), Self-Supervised Learning, Foundation Model 1. Introduction Electroencephalography (EEG) is a non-invasive and cost-effective technique for capturing rhythmic brain activity, widely used in clinical neurology to monitor conditions such as epilepsy, encephalopathy, and cognitive disorders [1, 2].

Clinical guidelines underscore the importance of regularly monitoring and surveilling arteriovenous fistula (AVF) access in hemodialysis patients to promptly detect any dysfunction. Although phono-angiography/sound analysis overcomes the limitations of standardized AVF stenosis diagnosis tool, prior studies have depended on conventional feature extraction methods, restricting their applicability in diverse contexts. In contrast, representation learning captures fundamental underlying factors that can be readily transferred across different contexts. We propose an approach based on deep denoising autoencoders (DAEs) that perform dimensionality reduction and reconstruction tasks using the waveform obtained through one-level discrete wavelet transform, utilizing representation learning. Our results demonstrate that the latent representation generated by the DAE surpasses expectations with an accuracy of 0.93. The incorporation of noise-mixing and the utilization of a noise-to-clean scheme effectively enhance the discriminative capabilities of the latent representation. Moreover, when employed to identify patient-specific characteristics, the latent representation exhibited performance by surpassing an accuracy of 0.92. Appropriate light-weighted methods can restore the detection performance of the excessively reduced dimensionality version and enable operation on less computational devices. Our findings suggest that representation learning is a more feasible approach for extracting auscultation features in AVF, leading to improved generalization and applicability across multiple tasks. The manipulation of latent representations holds immense potential for future advancements. Further investigations in this area are promising and warrant continued exploration.

General Hospital has saved $40 million in system-wide reduced inefficiencies since launching its CareComm artificial intelligence command center with GE Healthcare's software in August 2019. The program has helped TGH operate at maximum capacity, decrease average length of stay and reduced emergency room diversion by 25 percent for its Level 1 trauma center, according to an Oct. 29 news release. "CareComm is not only the center of gravity for our artificial intelligence platform, it's the center of gravity for the entire hospital system," said John Couris, CEO of TGH. "We feel sometimes that to fix a problem, we've got to build a building or build more capacity. We started to think a little differently saying, how do we drive value to the consumer by doing better with what we have and not just simply building more."

Changi General Hospital (CGH), a 1000-bed academic medical institution under SingHealth located in the eastern part of Singapore, together with the Integrated Health Information System (IHiS), Singapore's national HIT agency, have developed a Community Acquired Pneumonia and COVID-19 Artificial Intelligence (AI) Predictive Engine (CAPE) that can determine the likelihood of whether the patient has mild or severe pneumonia, based on the chest X-ray image. The ability to quickly predict the patient's expected severity of pneumonia would enable clinicians and administrators to efficiently allocate healthcare resources and treat patients, particularly in pandemic situations, where there may be an increased need for inpatient care and critical care support. As pneumonia severity correlates to the degree of Chest X-Ray (CXR) lung image abnormality, CGH's Respiratory and Critical Care Medicine and Radiology teams recognized the potential in leveraging AI to predict the severity of pneumonia from CXR images, and worked with the IHiS Health Insights team to develop CAPE. Using more than 3,000 CXR images and 200,000 data points including lab results and clinical history, CAPE was trained to generate a score for (a) low-risk pneumonia with anticipated short inpatient hospitalization; (b) the risk of mortality (death); and (c) the risk of requiring critical care support – indicators of pneumonia severity – from CXR images. Initial results have been promising – validation tests at CGH showed that CAPE has an approximate accuracy of 80% in predicting the future presence or absence of severe pneumonia.

How machine intelligence is helping healthcare providers fight the pandemic. In times of crisis, help can come from the most unexpected places. We're seeing that right now in the innovative ways AI is being used to protect healthcare workers and aid the effort to overcome COVID-19. And we're just scratching the surface on the potential for AI to make the entire healthcare journey safer and more humane for nurses, doctors, and patients. At Tampa General Hospital in Florida, an AI-driven technology screens individuals for COVID-19 symptoms before they interact with hospital staff and patients.

We've made our coronavirus coverage free for all readers. To get all of HBR's content delivered to your inbox, sign up for the Daily Alert newsletter. On Monday March 9, in an effort to address soaring patient demand in Boston, Partners HealthCare went live with a hotline for patients, clinicians, and anyone else with questions and concerns about Covid-19. The goals are to identify and reassure the people who do not need additional care (the vast majority of callers), to direct people with less serious symptoms to relevant information and virtual care options, and to direct the smaller number of high-risk and higher-acuity patients to the most appropriate resources, including testing sites, newly created respiratory illness clinics, or in certain cases, emergency departments. As the hotline became overwhelmed, the average wait time peaked at 30 minutes.

Despite significant advancements in genetics and modern imaging technology, for the vast majority of breast cancer patients, the diagnosis catches them by surprise. For some, it comes too late. Later diagnosis means aggressive treatments, anxiety and uncertain outcomes. Therefore, identifying patients at risk before the disease develops has been a central pillar to breast cancer research and effective early detection programs. A team from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL) and Massachusetts General Hospital (MGH) has created a new deep learning model that can predict from a mammogram if a patient is likely to develop breast cancer in the future.

For one 43-year-old Beijing patient, relief had seemed an impossible dream. His arm had been numb for 14 months and every hospital he went to gave him the same answer to his questions about a remedy. Surgeons told him that the risks of mass bleeding, stroke or even paralysis were too great with the delicate operation needed to fix the abnormalities in his spine and skull that were causing the condition. Then three years ago the patient met Tian Wei, a top spinal surgeon at Beijing's Jishuitan Hospital and an advocate of using robotics in medical operations. Tian and his team used a technology called the TiRobot system to create a 3D scan of the patient's torso and plot a surgical path to the affected area.

Approximately 330,000 cataract operations are performed in England alone and it all started with the introduction of the intraocular lens. Sir Harold Ridley was the first to successfully implant an intraocular lens on 29 November 1949, at St Thomas' Hospital at London, however, it wasn't until the 1970s, following further developments in lens design and surgical techniques, that the lens found acceptance in cataract surgery. Laser eye surgery then followed on from the intraocular lens in the 1990s.