Analysing disease trajectories in a cohort of 71,849 Patients: A visual analytics and statistical approach.

BACKGROUND: Disease trajectories have become increasingly relevant within the context of an aging population and the rising prevalence of chronic illnesses. Understanding the temporal progression of diseases is crucial for enhancing patient care, preventive measures, and effective management. OBJECTIVE: The objective of this study is to propose and validate a novel methodology for trajectory impact analysis and interactive visualization of disease trajectories over a cohort of 71,849 patients. METHODS: This article introduces an innovative comprehensive approach for analysis and interactive visualization of disease trajectories. First, Risk Increase (RI) index is defined that assesses the impact of the initial disease diagnosis on the development of subsequent illnesses. Secondly, visual graphics methods are used to represent cohort trajectories, ensuring a clear and semantically rich presentation that facilitates easy data interpretation. RESULTS: The proposed approach is demonstrated over the disease trajectories of a cohort comprising 71,849 patients from Tolosaldea, Spain. The study finds several clinically relevant trajectories in this cohort, such as that after suffering a cerebral ischemic stroke, the probability of suffering dementia increases 10.77 times. The clinical relevance of the study outcomes have been assessed by an in-depth analysis conducted by expert clinicians. The identified disease trajectories are in agreement with the latest advancements in the field. CONCLUSION: The proposed approach for trajectory impact analysis and interactive visualization offers valuable graphs for the comprehensive study of disease trajectories for improved clinical decision-making. The simplicity and interpretability of our methods make them valuable approach for healthcare professionals.

Clustering Cities over Features Extracted from Multiple Virtual Sensors Measuring Micro-Level Activity Patterns Allows One to Discriminate Large-Scale City Characteristics.

The impact of micro-level people's activities on urban macro-level indicators is a complex question that has been the subject of much interest among researchers and policymakers. Transportation preferences, consumption habits, communication patterns and other individual-level activities can significantly impact large-scale urban characteristics, such as the potential for innovation generation of the city. Conversely, large-scale urban characteristics can also constrain and determine the activities of their inhabitants. Therefore, understanding the interdependence and mutual reinforcement between micro- and macro-level factors is critical to defining effective public policies. The increasing availability of digital data sources, such as social media and mobile phones, has opened up new opportunities for the quantitative study of this interdependency. This paper aims to detect meaningful city clusters on the basis of a detailed analysis of the spatiotemporal activity patterns for each city. The study is carried out on a worldwide city dataset of spatiotemporal activity patterns obtained from geotagged social media data. Clustering features are obtained from unsupervised topic analyses of activity patterns. Our study compares state-of-the-art clustering models, selecting the model achieving a 2.7% greater Silhouette Score than the next-best model. Three well-separated city clusters are identified. Additionally, the study of the distribution of the City Innovation Index over these three city clusters shows discrimination of low performing from high performing cities relative to innovation. Low performing cities are identified in one well-separated cluster. Therefore, it is possible to correlate micro-scale individual-level activities to large-scale urban characteristics.

Mortality Risks after Two Years in Frail and Pre-Frail Older Adults Admitted to Hospital.

BACKGROUND: Frailty is characterized by a progressive decline in the physiological functions of multiple body systems that lead to a more vulnerable condition, which is prone to the development of various adverse events, such as falls, hospitalization, and mortality. This study aims to determine whether frailty increases mortality compared to pre-frailty and to identify variables associated with a higher risk of mortality. MATERIALS: Two cohorts, frail and pre-frail subjects, are evaluated according to the Fried phenotype. A complete examination of frailty, cognitive status, comorbidities and pharmacology was carried out at hospital admission and was extracted through electronic health record (EHR). Mortality was evaluated from the EHR. METHODS: Kaplan-Meier estimates of survival probability functions were calculated at two years censoring time for frail and pre-frail cohorts. The log-rank test assessed significant differences between survival probability functions. Significant variables for frailty (p < 0-05) were extracted by independent sample t-test. Further selection was based on variable significance found in multivariate logistic regression discrimination between frail and pre-frail subjects. Cox regression over univariate t-test-selected variables was calculated to identify variables associated with higher proportional hazard risks (HR) at two years. RESULTS: Frailty is associated with greater mortality at two years censoring time than pre-frailty (log-rank test, p < 0.0001). Variables with significant (p < 0.05) association with mortality identified in both cohorts (HR 95% (CI in the frail cohort) are male sex (0.44 (0.29-0.66)), age (1.05 (1.01-1.09)), weight (0.98 (0.96-1.00)), and use of proton-pump inhibitors (PPIs) (0.60 (0.41-0.87)). Specific high-risk factors in the frail cohort are readmission at 30 days (0.50 (0.33-0.74)), SPPB sit and stand (0.62 (0.45-0.85)), heart failure (0.67 (0.46-0.98)), use of antiplatelets (1.80 (1.19-2.71)), and quetiapine (0.31 (0.12-0.81)). Specific high-risk factors in the pre-frail cohort are Barthel's score (120 (7.7-1700)), Pfeiffer test (8.4; (2.3-31)), Mini Nutritional Assessment (MNA) (1200 (18-88,000)), constipation (0.025 (0.0027-0.24)), falls (18,000 (150-2,200,000)), deep venous thrombosis (8400 (19-3,700,000)), cerebrovascular disease (0.01 (0.00064-0.16)), diabetes (360 (3.4-39,000)), thyroid disease (0.00099 (0.000012-0.085)), and the use of PPIs (0.062 (0.0072-0.54)), Zolpidem (0.000014 (0.0000000021-0.092)), antidiabetics (0.00015 (0.00000042-0.051)), diuretics (0.0003 (0.000004-0.022)), and opiates (0.000069 (0.00000035-0.013)). CONCLUSIONS: Frailty is associated with higher mortality at two years than pre-frailty. Frailty is recognized as a systemic syndrome with many links to older-age comorbidities, which are also found in our study. Polypharmacy is strongly associated with frailty, and several commonly prescribed drugs are strongly associated with increased mortality. It must be considered that frail patients need coordinated attention where the diverse specialist taking care of them jointly examines the interactions between the diversity of treatments prescribed.

Prediction and Analysis of Heart Failure Decompensation Events Based on Telemonitored Data and Artificial Intelligence Methods.

Cardiovascular diseases are the leading cause of death globally, taking an estimated 17.9 million lives each year. Heart failure (HF) occurs when the heart is not able to pump enough blood to satisfy metabolic needs. People diagnosed with chronic HF may suffer from cardiac decompensation events (CDEs), which cause patients' worsening. Being able to intervene before decompensation occurs is the major challenge addressed in this study. The aim of this study is to exploit available patient data to develop an artificial intelligence (AI) model capable of predicting the risk of CDEs timely and accurately. Materials and Methods: The vital variables of patients (n = 488) diagnosed with chronic heart failure were monitored between 2014 and 2022. Several supervised classification models were trained with these monitoring data to predict CDEs, using clinicians' annotations as the gold standard. Feature extraction methods were applied to identify significant variables. Results: The XGBoost classifier achieved an AUC of 0.72 in the cross-validation process and 0.69 in the testing set. The most predictive physiological variables for CAE decompensations are weight gain, oxygen saturation in the final days, and heart rate. Additionally, the answers to questionnaires on wellbeing, orthopnoea, and ankles are strongly significant predictors.

Neuro-semantic prediction of user decisions to contribute content to online social networks.

Understanding at microscopic level the generation of contents in an online social network (OSN) is highly desirable for an improved management of the OSN and the prevention of undesirable phenomena, such as online harassment. Content generation, i.e., the decision to post a contributed content in the OSN, can be modeled by neurophysiological approaches on the basis of unbiased semantic analysis of the contents already published in the OSN. This paper proposes a neuro-semantic model composed of (1) an extended leaky competing accumulator (ELCA) as the neural architecture implementing the user concurrent decision process to generate content in a conversation thread of a virtual community of practice, and (2) a semantic modeling based on the topic analysis carried out by a latent Dirichlet allocation (LDA) of both users and conversation threads. We use the similarity between the user and thread semantic representations to built up the model of the interest of the user in the thread contents as the stimulus to contribute content in the thread. The semantic interest of users in discussion threads are the external inputs for the ELCA, i.e., the external value assigned to each choice.. We demonstrate the approach on a dataset extracted from a real life web forum devoted to fans of tinkering with musical instruments and related devices. The neuro-semantic model achieves high performance predicting the content posting decisions (average F score 0.61) improving greatly over well known machine learning approaches, namely random forest and support vector machines (average F scores 0.19 and 0.21).

A Hybrid Time-Distributed Deep Neural Architecture for Speech Emotion Recognition.

In recent years, speech emotion recognition (SER) has emerged as one of the most active human-machine interaction research areas. Innovative electronic devices, services and applications are increasingly aiming to check the user emotional state either to issue alerts under some predefined conditions or to adapt the system responses to the user emotions. Voice expression is a very rich and noninvasive source of information for emotion assessment. This paper presents a novel SER approach based on that is a hybrid of a time-distributed convolutional neural network (TD-CNN) and a long short-term memory (LSTM) network. Mel-frequency log-power spectrograms (MFLPSs) extracted from audio recordings are parsed by a sliding window that selects the input for the TD-CNN. The TD-CNN transforms the input image data into a sequence of high-level features that are feed to the LSTM, which carries out the overall signal interpretation. In order to reduce overfitting, the MFLPS representation allows innovative image data augmentation techniques that have no immediate equivalent on the original audio signal. Validation of the proposed hybrid architecture achieves an average recognition accuracy of 73.98% on the most widely and hardest publicly distributed database for SER benchmarking. A permutation test confirms that this result is significantly different from random classification ([Formula: see text]). The proposed architecture outperforms state-of-the-art deep learning models as well as conventional machine learning techniques evaluated on the same database trying to identify the same number of emotions.

Survival of Frail Elderly with Delirium.

This study aims to determine when frailty increases the risks of delirium mortality. Hospital patients falling into the elderly frail or pre-frail category were recruited, some without delirium, some with delirium at admission, and some who developed delirium during admission. We screened for frailty, cognitive status, and co-morbidities whenever possible and extracted drug information and mortality data from electronic health records. Kaplan-Meier estimates of survival probability functions were computed at four times, comparing delirium versus non delirium patients. Differences in survival were assessed by a log-rank test. Independent Cox's regression was carried out to identify significant hazard risks (HR) at 1 month, 6 months, 1 year, and 2 years. Delirium predicted mortality (log-rank test, p < 0.0001) at all four censoring points. Variables with significant HRs were frailty indicators, comorbidities, polypharmacy, and the use of specific drugs. For the delirium cohort, variables with the most significant 2-year hazard risks (HR(95%CI)) were: male gender (0.43 20 (0.26,0.69)), weight loss (0.45 (0.26,0.74)), sit and stand up test (0.67 (0.49,0.92)), readmission within 30 days of discharge (0.50 (0.30,0.80)), cerebrovascular disease (0.45 (0.27,0.76)), head trauma (0.54 22 (0.29,0.98)), number of prescribed drugs (1.10 (1.03,1.18)), and the use of diuretics (0.57 (0.34,0.96)). These results suggest that polypharmacy and the use of diuretics increase mortality in frail elderly patients with delirium.

Photometric Stereo-Based Defect Detection System for Steel Components Manufacturing Using a Deep Segmentation Network.

This paper presents an automatic system for the quality control of metallic components using a photometric stereo-based sensor and a customized semantic segmentation network. This system is designed based on interoperable modules, and allows capturing the knowledge of the operators to apply it later in automatic defect detection. A salient contribution is the compact representation of the surface information achieved by combining photometric stereo images into a RGB image that is fed to a convolutional segmentation network trained for surface defect detection. We demonstrate the advantage of this compact surface imaging representation over the use of each photometric imaging source of information in isolation. An empirical analysis of the performance of the segmentation network on imaging samples of materials with diverse surface reflectance properties is carried out, achieving Dice performance index values above 0.83 in all cases. The results support the potential of photometric stereo in conjunction with our semantic segmentation network.

Machine Learning Algorithms as a Computer-Assisted Decision Tool for Oral Cancer Prognosis and Management Decisions: A Systematic Review.

INTRODUCTION: Despite multiple prognostic indicators described for oral cavity squamous cell carcinoma (OCSCC), its management still continues to be a matter of debate. Machine learning is a subset of artificial intelligence that enables computers to learn from historical data, gather insights, and make predictions about new data using the model learned. Therefore, it can be a potential tool in the field of head and neck cancer. METHODS: We conducted a systematic review. RESULTS: A total of 81 manuscripts were revised, and 46 studies met the inclusion criteria. Of these, 38 were excluded for the following reasons: use of a classical statistical method (N = 16), nonspecific for OCSCC (N = 15), and not being related to OCSCC survival (N = 7). In total, 8 studies were included in the final analysis. CONCLUSIONS: ML has the potential to significantly advance research in the field of OCSCC. Advantages are related to the use and training of ML models because of their capability to continue training continuously when more data become available. Future ML research will allow us to improve and democratize the application of algorithms to improve the prediction of cancer prognosis and its management worldwide.

Detecting fine and elaborate movements with piezo sensors provides non-invasive access to overlooked behavioral components.

Behavioral phenotyping devices have been successfully used to build ethograms, but many aspects of behavior remain out of reach of available phenotyping systems. We now report on a novel device, which consists in an open-field platform resting on highly sensitive piezoelectric (electromechanical) pressure-sensors, with which we could detect the slightest movements (up to individual heart beats during rest) from freely moving rats and mice. The combination with video recordings and signal analysis based on time-frequency decomposition, clustering, and machine learning algorithms provided non-invasive access to previously overlooked behavioral components. The detection of shaking/shivering provided an original readout of fear, distinct from but complementary to behavioral freezing. Analyzing the dynamics of momentum in locomotion and grooming allowed to identify the signature of gait and neurodevelopmental pathological phenotypes. We believe that this device represents a significant progress and offers new opportunities for the awaited advance of behavioral phenotyping.

COLAEVA: Visual Analytics and Data Mining Web-Based Tool for Virtual Coaching of Older Adult Populations.

The global population is aging in an unprecedented manner and the challenges for improving the lives of older adults are currently both a strong priority in the political and healthcare arena. In this sense, preventive measures and telemedicine have the potential to play an important role in improving the number of healthy years older adults may experience and virtual coaching is a promising research area to support this process. This paper presents COLAEVA, an interactive web application for older adult population clustering and evolution analysis. Its objective is to support caregivers in the design, validation and refinement of coaching plans adapted to specific population groups. COLAEVA enables coaching caregivers to interactively group similar older adults based on preliminary assessment data, using AI features, and to evaluate the influence of coaching plans once the final assessment is carried out for a baseline comparison. To evaluate COLAEVA, a usability test was carried out with 9 test participants obtaining an average SUS score of 71.1. Moreover, COLAEVA is available online to use and explore.

Depth Data Denoising in Optical Laser Based Sensors for Metal Sheet Flatness Measurement: A Deep Learning Approach.

Surface flatness assessment is necessary for quality control of metal sheets manufactured from steel coils by roll leveling and cutting. Mechanical-contact-based flatness sensors are being replaced by modern laser-based optical sensors that deliver accurate and dense reconstruction of metal sheet surfaces for flatness index computation. However, the surface range images captured by these optical sensors are corrupted by very specific kinds of noise due to vibrations caused by mechanical processes like degreasing, cleaning, polishing, shearing, and transporting roll systems. Therefore, high-quality flatness optical measurement systems strongly depend on the quality of image denoising methods applied to extract the true surface height image. This paper presents a deep learning architecture for removing these specific kinds of noise from the range images obtained by a laser based range sensor installed in a rolling and shearing line, in order to allow accurate flatness measurements from the clean range images. The proposed convolutional blind residual denoising network (CBRDNet) is composed of a noise estimation module and a noise removal module implemented by specific adaptation of semantic convolutional neural networks. The CBRDNet is validated on both synthetic and real noisy range image data that exhibit the most critical kinds of noise that arise throughout the metal sheet production process. Real data were obtained from a single laser line triangulation flatness sensor installed in a roll leveling and cut to length line. Computational experiments over both synthetic and real datasets clearly demonstrate that CBRDNet achieves superior performance in comparison to traditional 1D and 2D filtering methods, and state-of-the-art CNN-based denoising techniques. The experimental validation results show a reduction in error than can be up to 15% relative to solutions based on traditional 1D and 2D filtering methods and between 10% and 3% relative to the other deep learning denoising architectures recently reported in the literature.

Non-Intrusive Assessment of COVID-19 Lockdown Follow-Up and Impact Using Credit Card Information: Case Study in Chile.

In this paper, we propose and validate with data extracted from the city of Santiago, capital of Chile, a methodology to assess the actual impact of lockdown measures based on the anonymized and geolocated data from credit card transactions. Using unsupervised Latent Dirichlet Allocation (LDA) semantic topic discovery, we identify temporal patterns in the use of credit cards that allow us to quantitatively assess the changes in the behavior of the people under the lockdown measures because of the COVID-19 pandemic. An unsupervised latent topic analysis uncovers the main patterns of credit card transaction activity that explain the behavior of the inhabitants of Santiago City. The approach is non-intrusive because it does not require the collaboration of people for providing the anonymous data. It does not interfere with the actual behavior of the people in the city; hence, it does not introduce any bias. We identify a strong downturn of the economic activity as measured by credit card transactions (down to 70%), and thus of the economic activity, in city sections (communes) that were subjected to lockdown versus communes without lockdown. This change in behavior is confirmed by independent data from mobile phone connectivity. The reduction of activity emerges before the actual lockdowns were enforced, suggesting that the population was spontaneously implementing the required measures for slowing virus propagation.

Combining Geostatistics and Remote Sensing Data to Improve Spatiotemporal Analysis of Precipitation.

The wide availability of satellite data from many distributors in different domains of science has provided the opportunity for the development of new and improved methodologies to aid the analysis of environmental problems and to support more reliable estimations and forecasts. Moreover, the rapid development of specialized technologies in satellite instruments provides the opportunity to obtain a wide spectrum of various measurements. The purpose of this research is to use publicly available remote sensing product data computed from geostationary, polar and near-polar satellites and radar to improve space-time modeling and prediction of precipitation on Crete island in Greece. The proposed space-time kriging method carries out the fusion of remote sensing data with data from ground stations that monitor precipitation during the hydrological period 2009/10-2017/18. Precipitation observations are useful for water resources, flood and drought management studies. However, monitoring stations are usually sparse in regions with complex terrain, are clustered in valleys, and often have missing data. Satellite precipitation data are an attractive alternative to observations. The fusion of the datasets in terms of the space-time residual kriging method exploits the auxiliary satellite information and aids in the accurate and reliable estimation of precipitation rates at ungauged locations. In addition, it represents an alternative option for the improved modeling of precipitation variations in space and time. The obtained results were compared with the outcomes of similar works in the study area.

Machine Learning Models for Predicting Facial Nerve Palsy in Parotid Gland Surgery for Benign Tumors.

BACKGROUND: Despite the increasing use of intraoperative facial nerve monitoring during parotid gland surgery (PGS) and the improvement in the preoperative radiological assessment, facial nerve injury (FNI) remains the most severe complication after PGS. Until now, no studies have been published regarding the application of machine learning (ML) for predicting FNI after PGS. We hypothesize that ML would improve the prediction of patients at risk. METHODS: Patients who underwent PGS for benign tumors between June 2010 and June 2019 were included. RESULTS: Regarding prediction accuracy and performance of each ML algorithm, the K-nearest neighbor and the random forest achieved the highest sensitivity, specificity, positive predictive value, negative predictive value F-score, receiver operating characteristic (ROC)-area under the ROC curve, and accuracy globally. The K-nearest neighbor algorithm achieved performance values above 0.9 for specificity, negative predictive value, F-score and ROC-area under the ROC curve, and the highest sensitivity and positive predictive value. CONCLUSIONS: This study demonstrates that ML prediction models can provide evidence-based predictions about the risk of FNI to otolaryngologists and patients. It is hoped that such algorithms, which use clinical, radiological, histological, and cytological information, can improve the information given to patients before surgery so that they can be better informed of any potential complications.

Impact of Machine Learning Pipeline Choices in Autism Prediction From Functional Connectivity Data.

Autism Spectrum Disorder (ASD) is a largely prevalent neurodevelopmental condition with a big social and economical impact affecting the entire life of families. There is an intense search for biomarkers that can be assessed as early as possible in order to initiate treatment and preparation of the family to deal with the challenges imposed by the condition. Brain imaging biomarkers have special interest. Specifically, functional connectivity data extracted from resting state functional magnetic resonance imaging (rs-fMRI) should allow to detect brain connectivity alterations. Machine learning pipelines encompass the estimation of the functional connectivity matrix from brain parcellations, feature extraction, and building classification models for ASD prediction. The works reported in the literature are very heterogeneous from the computational and methodological point of view. In this paper, we carry out a comprehensive computational exploration of the impact of the choices involved while building these machine learning pipelines. Specifically, we consider six brain parcellation definitions, five methods for functional connectivity matrix construction, six feature extraction/selection approaches, and nine classifier building algorithms. We report the prediction performance sensitivity to each of these choices, as well as the best results that are comparable with the state of the art.