Search

Screening for ocular anomalies using fundus photography is key to prevent vision impairment and blindness. With the growing and aging population, automated algorithms that can triage fundus photographs and provide instant referral decisions are relevant to scale-up screening and face the shortage of ophthalmic expertise.This study aimed to develop a deep learning algorithm that detects any ocular anomaly in fundus photographs and to evaluate this algorithm for "normal versus anomalous" eye examination classification in the diabetic and general populations.The deep learning algorithm was developed and evaluated in two populations: the diabetic and general populations. Our patient cohorts consist of 37,129 diabetic patients from the OPHDIAT diabetic retinopathy screening network in Paris, France, and 7356 general patients from the OphtaMaine private screening network, in Le Mans, France. Each data set was divided into a development subset and a test subset of more than 4000 examinations each. For ophthalmologist/algorithm comparison, a subset of 2014 examinations from the OphtaMaine test subset was labeled by a second ophthalmologist. First, the algorithm was trained on the OPHDIAT development subset. Then, it was fine-tuned on the OphtaMaine development subset.On the OPHDIAT test subset, the area under the receiver operating characteristic curve for normal versus anomalous classification was 0.9592. On the OphtaMaine test subset, the area under the receiver operating characteristic curve was 0.8347 before fine-tuning and 0.9108 after fine-tuning. On the ophthalmologist/algorithm comparison subset, the second ophthalmologist achieved a specificity of 0.8648 and a sensitivity of 0.6682. For the same specificity, the fine-tuned algorithm achieved a sensitivity of 0.8248.The proposed algorithm compares favorably with human performance for normal versus anomalous eye examination classification using fundus photography. Artificial intelligence, which previously targeted a few retinal pathologies, can be used to screen for ocular anomalies comprehensively.

We address the problem of abnormal behaviour recognition of the inhabitant of a smart home in the presence of unreliable sensors. The corner stone of this work is a two-level architecture sensor fusion based on the Transferable Belief Model (TBM). The novelty of our work lies in the way we detect both unreliable sensors and abnormal behaviour within our architecture by using a temporal analysis of conflict resulting from the fusion of sensors. Detection of abnormal behaviour is based on a prediction/observation process and the influence of the faulty sources is discarded by discounting coefficients. Our architecture is tested in a real-life setting using three heterogeneous sensors enabling the detection of impossible transitions between three possible postures: Sitting, Standing and Lying. The impact of having a faulty sensor management is also tested in the real-life experiment for posture detection.

To follow up our works on a service-oriented architecture dedicated to context perception for the smart home [7], we present in this paper how to move complex reasoning on top of the ontology. We focus on a multi-sensors data fusion architecture using the Transferable Belief Model. High level symbolic data are deduced using fusion from redundant, complementary and heterogeneous sensors data. Symbolic data describe complex contexts and enable to provide more contextualised services. To show the contribution of using the Transferable Belief Model in context awareness systems we detail an experiment about human posture characterisation using heterogeneous and uncertain sensors data.

Firstly this article presents a multi-level architecture permitting the localization of a mobile platform and secondly an incremental construction of the environment's map. The environment will be modeled by an occupancy grid built with information provided by the stereovision system situated on the platform. The reliability of these data is introduced to the grid by the propagation of uncertainties managed thanks to the theory of the Transferable Belief Model.

This paper deals with a context awareness system based on a service-oriented architecture dedicated to the Smart Home. In our approach, a Smart Home must be able to provide services adapted to the needs of its inhabitants. To achieve this, we propose to build a Smart environment that takes into account the ambient context of its inhabitants, providing contextually-aware services. In previous work [18], we described a four-layer architecture to manage data acquisition and context information in distributed sensor network for smart home environments. In this paper, we focus on the context model based on an OWL ontology and then on the inference layer, based on SWRL rules language. Finally, we present our current implementation.

International audience

This general paper aims at presenting the Smart Home concept. In this paper, we will detail a) The Smart Home concept b) The various networks infrastructures specific to the habitat c) Our concepts to model the habitat and to provide the most adapted services to the inhabitants. Contrary to the other projects, we direct our work towards a sensors approach and an ontology modelling of the Smart Home. Our work has the originality to take into account the real heterogeneity of information present in an habitat and use a Service Oriented Approach (SOA). We can say that our paper is a good overview to present what is a Smart Home and which are the necessary hardware and software components to make a Smart Home.

In this paper we will talk about the smart home concept. Even if smart home domain is not like the robotics domain, we can notice resemblances. In V. Ricquebourg et al. (2005), we shown that a smart home can be seen as a robot that turn over itself. A robot needs sensors to discover its environment. In the same way, a smart home needs sensors to watch in itself. This paper deals with a service oriented architecture of context perception for the smart home. In our approach, the smart house must be able to provide services adapted to the needs for its inhabitants. To answer to the needs, the environment, which will be described as smart, must take into account the context around the inhabitant: it will make it possible to provide contextualised services. Here, we present our service oriented architecture which integrates the various layers necessary to the determination of this context. Four layers are considered. Perception layer to obtain necessary information on the context, the context layer which makes it possible to give a sense to the perceived data, the inference layer which makes it possible to elaborate an artificial intelligence layer and to reason on the contextual data and the action layer which makes it possible to act following the decisions induced by the inference layer