MOTIVATION: Acute ischemic stroke is one of the leading causes of morbidity and disability worldwide, often followed by a long rehabilitation period. To improve and personalize stroke rehabilitation, it is essential to provide a reliable prognosis to caregivers and patients. Deep learning techniques might improve the predictions by incorporating different data modalities. We present a multimodal approach to predict the functional status of acute ischemic stroke patients after their discharge based on tabular data and CT perfusion imaging. METHODS: We conducted experiments on tabular, imaging, and multimodal deep learning architectures to predict dichotomized mRS scores 3 months after the event. The dataset was collected from a Dutch hospital and includes 98 CVA patients with a visible occlusion on their CT perfusion scan. Tabular data is based on the Dutch Acute Stroke Audit data, and imaging data consists of summed-up CT perfusion maps. RESULTS: On the tabular data, TabNet outperformed our baselines with an AUC of 0.71, while ResNet-10 on the imaging data performed comparably with an AUC of 0.70. Our implementation of the multimodal DAFT architecture outperforms baselines as well as comparable studies by achieving an 0.75 AUC, and 0.80 F1 score. This was achieved with a final model of less than a hundred thousand optimizable parameters, and a dataset less than half the size of reference papers. CONCLUSION: Overall, we demonstrate the feasibility of predicting the functional outcome for ischemic stroke patients and the usability of multimodal deep learning architectures for this purpose.

In healthcare and other fields, data of an individual is often vertically partitioned across multiple organizations. Creating a centralized data store for AI algorithm development is cumbersome in such cases because of concerns like privacy and data ownership. Methods of distributed learning over vertically partitioned data could offer a solution here. While several studies have evaluated the feasibility, privacy and efficiency of such methods, an extensive evaluation of their impact on predictive performance compared to a centralized approach is missing. Vertical Split Learning (VSL) aims to provide vertical distributed learning through distributed neural network architectures. Our study adapts and applies VSL to 8 datasets, both in medicine and beyond, evaluating the impact of different network and (vertical) feature distributions on predictive performance. In most configurations VSL yields comparable predictive performance to its centralized counterparts. However, certain data and network distributions give an unexpected and severe loss of performance. Based on our findings we give some initial recommendations under which conditions VSL can be applied as a suitable alternative for data centralization.

Agent-oriented solutions form a useful paradigm to design intelligent systems. For health-related behaviour change, this is also a promising approach. Designing an agent for lifestyle change interventions is a difficult task because socio-ecological models are involved that represent many conflicting desires and goals. Different types of cognitive architectures are available to design this type of health behavior agents but they are rarely used. In this paper, we used the BDI model to design a health behavior agent that will execute behavior change intervention for a better healthy lifestyle. We explain the working of the architecture by the example of an agent which uses adaptive goals-setting and a percentile scheduling technique for increasing physical activity.

The daily home-office commute of millions of people in crowded cities puts a strain on air quality, traveling time and noise pollution. This is especially problematic in western cities, where cars and taxis have low occupancy with daily commuters. To reduce these issues, authorities often encourage commuters to share their rides, also known as carpooling or ridesharing. To increase the ridesharing usage it is essential that commuters are efficiently matched. In this paper we present RideMatcher, a novel peer-to-peer system for matching car rides based on their routes and travel times. Unlike other ridesharing systems, RideMatcher is completely decentralized, which makes it possible to deploy it on distributed infrastructures, using fog and edge computing. Despite being decentralized, our system is able to efficiently match ridesharing users in near real-time. Our evaluations performed on a dataset with 34,837 real taxi trips from New York show that RideMatcher is able to reduce the number of taxi trips by up to 65%, the distance traveled by taxi cabs by up to 64%, and the cost of the trips by up to 66%.

Understanding the social dynamics of a group of people can give new insights into social behavior. Physical proximity between individuals results from the interactions between them. Hence, measuring physical proximity is an important step towards a better understanding of social behavior. We discuss a novel approach to sense proximity from within the social dynamics. Our primary objective is to construct a spatio-temporal social graph from noisy proximity data. We address the technical and algorithmic challenges of measuring proximity reliably and accurately. Simulations and real world experiments demonstrate the feasibility and scalability of our approach. Our algorithms doubles the sensitivity of proximity detections at the cost of a slight reduction in specificity.

Household food practices are complex. Many people are unable to effectively respond to challenges in their food environment to maintain diets considered to be in line with national and international standards for healthy eating. We argue that recognizing food practices as situated action affords opportunities to identify and design for practiced, local and achievable solutions to such food problems. Interviews and shop-a-longs were carried as part of a contextual inquiry with ten households. From this, we identify food practices, such as fitting food, stocking up, food value transitions, and having fun with others and how these practices are enacted in different ways with varied outcomes. We explore how HCI might respond to these practices through issues of social fooding, the presence of others, conceptions about food practices and food routines.

In this article, we describe the usage of persuasion profiles in a large scale, N = 1,129, field trial. Persuasive technologies—technologies intentionally designed to influence user behavior—are emergent and becoming more and more individualized and ubiquitous. Individual differences in people’s responses to often used persuasion principles —different psychological means by which to influence users—motivate personalization. We describe how, through identification, representation, and measurement, persuasive technologies can personalize their persuasive attempts. Next, we show that dynamically adapting a persuasive technology to the responses of its users increases the effectiveness of the system. Ubiquitous computing systems are, because of their ability to unobtrusively measure user behavior, very well suited for these types of applications.

A case study on the use of an existing photo sharing application in a spinal cord lesion rehabilitation centre is presented. The study focuses on enhancing social connectedness through sharing photos between rehabilitants and their families and friends. Four rehabilitants participated in this study for 6-7 weeks. Most photos sent related to sharing things in everyday life and keeping the rehabilitant informed about regular events. The combination of interviews and content analysis reveals that only a minority of the photos lead to follow-up communication about the contents of the photos. Rehabilitants were positively surprised how spontaneous photo sharing simplified the way they could reconnect to their friends and family, without the immediate need or obligation to engage in a (phone) conversation.

With the advent of ubiquitous computing, a user is surrounded by a variety of devices including tiny sensor nodes, handheld mobile devices and powerful computers as well as diverse communication networks. In this networked society, the role of a human being is evolving from the data consumer to the data producer. In these changing circumstances, pipelined processing finds applications where the data obtained from the human producer needs to be processed and interpreted in real-time. For example, in an M- Health system, the vital signs acquired from the patient are processed in the pipelined fashion. This paper proposes a genetic algorithm (GA) based approach for the optimal assignment of pipelined processing tasks onto a chain of networked devices that minimizes total end-to-end processing delay considering knowledge about the communication and computation resources as the context information. Although some existing graph-based algorithms can solve this problem in polynomial time, we expect that GA can take less computational time and requires less memory while providing a reasonably good assignment. We compare the performance of GA approach with the graph-based approaches. It is observed that when the number of devices and the number of processing tasks are large, the GA approach performs better in terms of the satisfactory quality of the obtained sub-optimal solution considering the advantage of reduced computational time.

Nowadays, a variety of handheld mobile devices are capable of connecting to the Internet using multiple network interfaces. This is referred to as multi-homing. In addition to this, enriched computation resources allow them to host nomadic mobile services and provide these services to the clients located anywhere in the Internet. Potential advantages of multi-homing for nomadic mobile services typically includes: an increased service availability and improved service performance. However, applications running on the handheld mobile devices either do not, or cannot, exploit multi-homing. In this paper we address the problem of providing infrastructural support to the nomadic mobile services that can fully exploit multi-homing. To this end we propose to incorporate multi-homing functionality and support in a middleware layer to reduce the complexity of the design and maintenance of these services. The proposed solution uses a context-aware computing approach to realize this functionality. We report the initial experimental results in the remote tele-monitoring of a patient equipped with a Body Area Network.

Context-awareness (CA) enables the development of personalized applications for highly mobile and demanding users in the pervasive environments. These applications rely on the components responsible for context sensing, processing, storage and inference. Various context-aware computing infrastructures exist to shield the application developers from the complexity of developing these components. However, some additional concerns including the distribution of context sources in the pervasive environments, quality of context (QoC) and user privacy demand that a context-aware computing infrastructure should also handle these aspects. This paper introduces our work in progress on the context distribution framework (CDF) aimed at providing a service oriented infrastructure for the context-aware applications hosted on a mobile device. In the current version of CDF, we focus on three aspects: a) transparent off-loading of resource intensive context manipulation from the mobile device; b) selection of the suitable context sources based on the quality of context (QoC); and c) modeling mobile context sources as services.

Abstract—Emerging wireless communication standards and more capable sensors and actuators have pushed further development of wireless sensor networks. Deploying a large number of sensor nodes requires a high-level framework enabling the devices to present themselves and the resources they hold. The device and the resources can be described as services, and in this paper, we review a number of well-known service discovery protocols. Bonjour stands out with its auto-configuration, distributed architecture, and sharing of resources. We also present a lightweight implementation in order to demonstrate that an emerging standards-based device and service discovery protocol can actually be deployed on small wireless sensor nodes. Index Terms: Service discovery, wireless sensor, sensor networks.

Ongoing miniaturization and power efficiency of mobile devices have led to widespread availability of devices that have an increasing amount of processing power and storage, and that support multiple wireless network interfaces connecting to various auxiliary devices and to the Internet. It is now feasible for a mobile device to host services and participate in a service discovery network. Roaming of a mobile device from one wireless network to another entails nomadic characteristics to the hosted services. We denote this class of services as nomadic mobile services. This paper discusses the requirements for nomadic mobile service provisioning and proposes the mobile service platform (MSP) as a supporting infrastructure and middleware which extends the service oriented architecture paradigm to the mobile device. The MSP design is based on the Jini surrogate architecture specification which enables devices that can not directly participate in a Jini network to join a Jini network with the aid of a third party. MSP consists of an HTTPInterconnect protocol to meet the specifications of Jini surrogate architecture and provides a custom set of APIs to develop and deploy a nomadic mobile service. This paper also presents case studies of MSP enabled services in diverse domains such as healthcare, robotics and positioning services. To conclude, we outline the need for a context-aware MSP

Developers of distributed multimedia applications face a diversity of multimedia formats, streaming platforms and streaming protocols. Furthermore, support for end-to-end quality-of-service (QoS) is a crucial factor for the development of future distributed multimedia systems. This paper discusses the architecture, design and implementation of a QoS-aware middleware platform for content delivery. The platform supports the development of distributed multimedia applications and can deliver content with QoS guarantees. QoS support is offered by means of an agent infrastructure for QoS negotiation and enforcement. Properties of content are represented using a generic content representation model described using the OMG Meta Object Facility (MOF) model. A content delivery framework manages stream paths for content delivery despite differences in streaming protocols and content encoding. The integration of the QoS support, content representation and content delivery framework results in a QoS-aware middleware that enables representation transparent and location transparent delivery of content.

Program visualisation is an attractive way for understanding collaboration structures of complex distributed systems. By using the concepts of the open distributed processing-reference model (ODP-RM) as entities for visualisation, a multiview visualisation architecture is presented, which provides a large degree of flexibility in visualising the actions of an ODP system. The architecture has been implemented for visualising the CORBA system resulting in a visualisation tool called OBVlouS.