Corticosteroid sensitivity detection in sepsis patients using a personalized data mining approach: A clinical investigation.

BACKGROUND AND OBJECTIVE: Sepsis is a life-threatening disease with high mortality, incidence, and morbidity. Corticosteroids (CS) are a recommended treatment for sepsis, but some patients respond negatively to CS therapy. Early prediction of corticosteroid responsiveness can help intervene and reduce mortality. In this study, we aim to develop a data mining methodology for predicting CS responsiveness of septic patients. METHODS: We used data from a randomized controlled trial called APROCCHSS, which recruited 1241 septis patients to study the effectiveness of corticotherapy. We conducted a thorough study of multiple machine learning models to select the most efficient prediction model, called "signature". We evaluated the performance of the signature using precision, sensitivity, and specificity values. RESULTS: We found that Logistic Regression was the best model with an AUC of 72%. We conducted further experiments to examine the impact of additional features and the model's generalizability to different groups of patients. We also performed a statistical analysis to analyze the effect of the treatment at the individual level and on the population as a whole. CONCLUSIONS: Our data mining methodology can accurately predict cortico-sensitivity or resistance in septis patients. The signature has been deployed into the Assistance Publique - Hôpitaux de Paris (APHP) information system as a web service, taking patient information as input and providing a prediction of cortico-sensitivity or resistance. Early prediction of corticosteroid responsiveness can help clinicians intervene promptly and improve patient outcomes.

NO2, BC and PM Exposure of Participants in the Polluscope Autumn 2019 Campaign in the Paris Region.

The Polluscope project aims to better understand the personal exposure to air pollutants in the Paris region. This article is based on one campaign from the project, which was conducted in the autumn of 2019 and involved 63 participants equipped with portable sensors (i.e., NO2, BC and PM) for one week. After a phase of data curation, analyses were performed on the results from all participants, as well as on individual participants' data for case studies. A machine learning algorithm was used to allocate the data to different environments (e.g., transportation, indoor, home, office, and outdoor). The results of the campaign showed that the participants' exposure to air pollutants depended very much on their lifestyle and the sources of pollution that may be present in the vicinity. Individuals' use of transportation was found to be associated with higher levels of pollutants, even when the time spent on transport was relatively short. In contrast, homes and offices were environments with the lowest concentrations of pollutants. However, some activities performed in indoor air (e.g., cooking) also showed a high levels of pollution over a relatively short period.

Modeling Trajectories Obtained from External Sensors for Location Prediction via NLP Approaches.

Representation learning seeks to extract useful and low-dimensional attributes from complex and high-dimensional data. Natural language processing (NLP) was used to investigate the representation learning models to extract words' feature vectors using their sequential order in the text via word embeddings and language models that maintain their semantic meaning. Inspired by NLP, in this paper, we tackle the representation learning problem for trajectories, using NLP methods to encode external sensors positioned in the road network and generate the features' space to predict the next vehicle movement. We evaluate the vector representations of on-road sensors and trajectories using extrinsic and intrinsic strategies. Our results have shown the potential of natural language models to describe the space of features on trajectory applications as the next location prediction.

Personal Exposure to Black Carbon, Particulate Matter and Nitrogen Dioxide in the Paris Region Measured by Portable Sensors Worn by Volunteers.

Portable sensors have emerged as a promising solution for personal exposure (PE) measurement. For the first time in Île-de-France, PE to black carbon (BC), particulate matter (PM), and nitrogen dioxide (NO2) was quantified based on three field campaigns involving 37 volunteers from the general public wearing the sensors all day long for a week. This successful deployment demonstrated its ability to quantify PE on a large scale, in various environments (from dense urban to suburban, indoor and outdoor) and in all seasons. The impact of the visited environments was investigated. The proximity to road traffic (for BC and NO2), as well as cooking activities and tobacco smoke (for PM), made significant contributions to total exposure (up to 34%, 26%, and 44%, respectively), even though the time spent in these environments was short. Finally, even if ambient outdoor levels played a role in PE, the prominent impact of the different environments suggests that traditional ambient monitoring stations is not a proper surrogate for PE quantification.

Towards a semantic indoor trajectory model: application to museum visits.

In this paper we present a new conceptual model of trajectories, which accounts for semantic and indoor space information and supports the design and implementation of context-aware mobility data mining and statistical analytics methods. Motivated by a compelling museum case study, and by what we perceive as a lack in indoor trajectory research, we combine aspects of state-of-the-art semantic outdoor trajectory models, with a semantically-enabled hierarchical symbolic representation of the indoor space, which abides by OGC's IndoorGML standard. We drive the discussion on modeling issues that have been overlooked so far and illustrate them with a real-world case study concerning the Louvre Museum, in an effort to provide a pragmatic view of what the proposed model represents and how. We also present experimental results based on Louvre's visiting data showcasing how state-of-the-art mining algorithms can be applied on trajectory data represented according to the proposed model, and outline their advantages and limitations. Finally, we provide a formal outline of a new sequential pattern mining algorithm and how it can be used for extracting interesting trajectory patterns.

A methodology for the characterization of portable sensors for air quality measure with the goal of deployment in citizen science.

The field of small air quality sensors is of growing interest within the scientific community, especially because this new technology is liable to improve air pollutant monitoring as well as be used for personal exposure quantification. Amongst the myriad existing devices, the performances are highly variable; this is why the sensors must be rigorously assessed before deployment, according to the intended use. This study is included in the Polluscope project; its purpose is to quantify personal exposure to air pollutants by using portable sensors. This paper designs and applies a methodology for the evaluation of portable air quality sensors to eight devices measuring PM, BC, NO2 and O3. The dedicated testing protocol includes static ambient air measurements compared with reference instruments, controlled chamber and mobility tests, as well as reproducibility evaluation. Three sensors (AE51, Cairclip and Canarin) were retained to be used for the field campaigns. The reliability of their performances were robustly quantified by using several metrics. These three devices (for a total of 36 units) were deployed to be worn by volunteers for a week. The results show the ability of sensors to discriminate between different environments (i.e., cooking, commuting or in an office). This work demonstrates, first, the ability of the three selected sensors to deliver data reliable enough to enable personal exposure estimations, and second, the robustness of this testing methodology.

A tamper-resistant and portable healthcare folder.

Electronic health record (EHR) projects have been launched in most developed countries to increase the quality of healthcare while decreasing its cost. The benefits provided by centralizing the healthcare information in database systems are unquestionable in terms of information quality, availability, and protection against failure. Yet, patients are reluctant to give to a distant server the control over highly sensitive data (e.g., data revealing a severe or shameful disease). This paper capitalizes on a new hardware portable device, associating the security of a smart card to the storage capacity of a USB key, to give back to the patient the control over his medical data. This paper shows how this device can complement a traditional EHR server to (1) protect and share highly sensitive data among trusted parties and (2) provide a seamless access to the data even in disconnected mode. The proposed architecture is experimented in the context of a medicosocial network providing medical care and social services at home for elderly people.