Whole slide imaging (WSI) scanners and automatic image analysis algorithms, in order to be used for clinical applications, including primary diagnosis in pathology, are subject to specific regulatory frameworks in each country. Until May 25, 2018, in the European Union (EU), in vitro diagnostic (IVD) medical devices were regulated by directive 98/79/EC (in vitro diagnostic medical device directive [IVDD]). Main scanner vendors have obtained a Conformité Européenne mark of their products that in Europe were classified as General Class IVDD, so that conformity is only based on a self-declaration of the manufacturer. This contrasts with the initial classification of the US Food and Drug Administration (FDA) of WSI system as Class III medical devices, although the first digital pathology WSI system to be cleared by FDA was classified as Class II, with special controls. Other digital pathology solutions (automated cervical cytology slide reader) are considered of higher risk by US and European regulations. There is also some disparity in the classification of image analysis solutions between Europe and the United States. All IVD-MDs must be approved under the new European regulation (in vitro diagnostic medical device regulation) 2017/746 after May 26, 2024. This means the need of a performance evaluation, including a scientific validity report, an analytical performance report, and a clinical performance report. According to its clinical use (e.g., screening, diagnosis, or staging of cancer), a WSI slide scanner can be now classified as Class C device. A special regulation is applied to companion diagnostics. The new EU regulation 2017/746 contemplates the use of standard unique identifiers for medical devices and the creation of a European database on medical devices (Eudamed). Existing validation studies and clinical guidelines already available in the literature are a sound basis to avoid that this new regulation becomes a barrier for digital pathology development in Europe.

A probabilistic deformable model for the representation of multiple brain structures is described. The statistically learned deformable model represents the relative location of different anatomical surfaces in brain magnetic resonance images (MRIs) and accommodates their significant variability across different individuals. The surfaces of each anatomical structure are parameterized by the amplitudes of the vibration modes of a deformable spherical mesh. For a given MRI in the training set, a vector containing the largest vibration modes describing the different deformable surfaces is created. This random vector is statistically constrained by retaining the most significant variation modes of its Karhunen-Loève expansion on the training population. By these means, the conjunction of surfaces are deformed according to the anatomical variability observed in the training set. Two applications of the joint probabilistic deformable model are presented: isolation of the brain from MRI using the probabilistic constraints embedded in the model and deformable model-based registration of three-dimensional multimodal (magnetic resonance/single photon emission computed tomography) brain images without removing nonbrain structures. The multi-object deformable model may be considered as a first step toward the development of a general purpose probabilistic anatomical atlas of the brain.

Secure electronic identification (eID) is one of the key enablers of data protection, privacy, and the prevention of online fraud. However, until now, the lack of common legal basis prevented European Member States from recognizing and accepting eIDs issued in the other Member States. The electronic identification and trust services (eIDAS) regulation provides a solution to these issues by ensuring the cross-border mutual recognition of eIDs. FIWARE is a European initiative that provides a rather simple yet powerful set of application programming interfaces (APIs) that ease the development of smart applications in multiple vertical sectors and oriented to the future internet. In this paper, we propose a model that enables the connection of FIWARE OAuth 2.0-based services with the eID authentication provided by eIDAS reference. Thanks to this model, services already connected with an OAuth 2.0 identity provider can be automatically connected with eIDAS nodes for providing eID authentication to European citizens. For validating the proposed model, we have deployed an instance of the FIWARE identity manager connected to the Spanish eIDAS node. Then, we have registered two services, a private videoconferencing system, and a public smart city deployment, and extended their functionalities for enriching the user experience leveraging the eID authentication. We have evaluated the integration of both services in the eIDAS network with real users from seven different countries. We conclude that the proposed model facilitates the integration of generic and FIWARE-based OAuth 2.0 services to the eIDAS infrastructure, making the connection transparent for developers. INDEX TERMS Access Control, eIDAS, electronic identification, identity, FIWARE.

Abstract. This paper presents a generic strategy to facilitate the segmentation of anatomical structures in medical images. The segmentation is performed using an adapted PDM by fuzzy c-means classification, which also uses the fuzzy decision to evolve PDM into the final contour. Furthermore, the fuzzy reasoning exploits a priori statistical information from several knowledge sources based on histogram analysis and the intensity values of the structures under consideration. The fuzzy reasoning is also applied and compared to a geometrical active contour model (or level set). The method has been developed to assist clinicians and radiologists in conformal RTP. Experimental results and their quantitative validation to assess the accuracy and efficiency are given segmenting the bladder on CT images. To assess precision, results are also presented in CT images with added Gaussian noise. The fuzzy-snake is free of parameter and it is able to properly segment the structures by using the same initial spline curve for a whole study image-patient set.

This paper reviews existing techniques to deliver radiotherapy treatment and describes the requirements from a control engineering view point to adapt radiation delivery to organs motion. The first part describes recent evolution of radiotherapy research; the second part focuses on imaging modalities used in radiotherapy; the third part reviews image processing techniques to detect organ movement; the final section compares traditional PID with predictive control strategies for use in patient positioning devices.

Deep learning (henceforth DL) has become most powerful machine learning methodology. Under specific circumstances recognition rates even surpass those obtained by humans. Despite this, several works have shown that deep learning produces outputs that are very far from human responses when confronted with the same task. This the case of the so-called "adversarial examples" (henceforth AE). The fact that such implausible misclassifications exist points to a fundamental difference between machine and human learning. This paper focuses on the possible causes of this intriguing phenomenon. We first argue that the error in adversarial examples is caused by high bias, i.e. by regularization that has local negative effects. This idea is supported by our experiments in which the robustness to adversarial examples is measured with respect to the level of fitting to training samples. Higher fitting was associated to higher robustness to adversarial examples. This ties the phenomenon to the trade-off that exists in machine learning between fitting and generalization.

The Eyes of Things (EoT) EU H2020 project envisages a computer vision platform that can be used both standalone and embedded into more complex artifacts, particularly for wearable applications, robotics, home products, surveillance etc. The core hardware will be based on a number of technologies and components that have been designed for maximum performance of the always-demanding vision applications while keeping the lowest energy consumption. An important functionality is to be able to communicate with other devices that we use everyday (say, configuring and controlling the EoT device from a tablet). Apart from low-power hardware components, an efficient protocol is necessary. Text-oriented protocols like HTTP are not appropriate in this context. Instead, the lightweight publish/subscribe MQTT protocol was selected. Still, the typical scenario is that of a device that sends/receives messages, the messages being forwarded by a cloud-based message broker. In this paper we propose a novel approach in which each EoT device acts as an MQTT broker instead of the typical cloud-based architecture. This eliminates the need for an external Internet server, which not only makes the whole deployment more affordable and simpler but also more secure by default.

No abstract