Search

Recent success in Artificial Intelligence (AI) and Machine Learning (ML) allow problem solving automatically without any human intervention. Autonomous approaches can be very convenient. However, in certain domains, e.g., in the medical domain, it is necessary to enable a domain expert to understand, why an algorithm came up with a certain result. Consequently, the field of Explainable AI (xAI) rapidly gained interest worldwide in various domains, particularly in medicine. Explainable AI studies transparency and traceability of opaque AI/ML and there are already a huge variety of methods. For example with layer-wise relevance propagation relevant parts of inputs to, and representations in, a neural network which caused a result, can be highlighted. This is a first important step to ensure that end users, e.g., medical professionals, assume responsibility for decision making with AI/ML and of interest to professionals and regulators. Interactive ML adds the component of human expertise to AI/ML processes by enabling them to re-enact and retrace AI/ML results, e.g. let them check it for plausibility. This requires new human-AI interfaces for explainable AI. In order to build effective and efficient interactive human-AI interfaces we have to deal with the question of how to evaluate the quality of explanations given by an explainable AI system. In this paper we introduce our System Causability Scale (SCS) to measure the quality of explanations. It is based on our notion of Causability (Holzinger et al., 2019) combined with concepts adapted from a widely accepted usability scale., Comment: 6 pages, 1 figure, 1 table, will appear in Springer/Nature KI - K\"unstliche Intelligenz (2020), Volume 34, Issue 2

Advancements in Artificial Intelligence (AI) and Machine Learning (ML) are enabling new diagnostic capabilities. In this paper we argue that the very first step before introducing AI/ML into diagnostic workflows is a deep understanding of how pathologists work. To contribute to a deeper understanding of the diagnostic process in histopathology, we developed a visualization concept, including: (a) the sequence of the views observed by the pathologist (Observation Path), (b) the sequence of the spoken comments and statements of the pathologist (Dictation Path), (c) the underlying knowledge and experience of the pathologist (Knowledge Path), (d) information about the current phase of the diagnostic process and (e) the current magnification factor of the microscope chosen by the pathologist. We implemented the proofof-concept prototype as HTML5/CSS3/JavaScript application.

AI follows the notion of human intelligence which is unfortunately not a clearly defined term. The most common definition given by cognitive science as mental capability, includes, among others, the ability to think abstract, to reason, and to solve problems from the real world. A hot topic in current AI/machine learning research is to find out whether and to what extent algorithms are able to learn abstract thinking and reasoning similarly as humans can do – or whether the learning outcome remains on purely statistical correlation. In this paper we provide some background on testing intelligence, report some preliminary results from 271 participants of our online study on explainability, and propose to use our Kandinsky Patterns as an IQ-Test for machines. Kandinsky Patterns are mathematically describable, simple, self-contained hence controllable test data sets for the development, validation and training of explainability in AI. Kandinsky Patterns are at the same time easily distinguishable from human observers. Consequently, controlled patterns can be described by both humans and computers. The results of our study show that the majority of human explanations was made based on the properties of individual elements in an image (i.e., shape, color, size) and the appearance of individual objects (number). Comparisons of elements (e.g., more, less, bigger, smaller, etc.) were significantly less likely and the location of objects, interestingly, played almost no role in the explanation of the images. The next step is to compare these explanations with machine explanations.

In this paper an automatic classification system for pathological findings is presented. The starting point in our undertaking was a pathologic tissue collection with about 1.4 million tissue samples described by free text records over 23 years. Exploring knowledge out of this “big data” pool is a challenging task, especially when dealing with unstructured data spanning over many years. The classification is based on an ontology-based term extraction and decision tree build with a manually curated classification system. The information extracting system is based on regular expressions and a text substitution system. We describe the generation of the decision trees by medical experts using a visual editor. Also the evaluation of the classification process with a reference data set is described. We achieved an F-Score of 89,7% for ICD-10 and an F-Score of 94,7% for ICD-O classification. For the information extraction of the tumor staging and receptors we achieved am F-Score ranging from 81,8 to 96,8%.

Rare diseases (RD) patient registries are powerful instruments that help develop clinical research, facilitate the planning of appropriate clinical trials, improve patient care, and support healthcare management. They constitute a key information system that supports the activities of European Reference Networks (ERNs) on rare diseases. A rapid proliferation of RD registries has occurred during the last years and there is a need to develop guidance for the minimum requirements, recommendations and standards necessary to maintain a high-quality registry. In response to these heterogeneities, in the framework of RD-Connect, a European platform connecting databases, registries, biobanks and clinical bioinformatics for rare disease research, we report on a list of recommendations, developed by a group of experts, including members of patient organizations, to be used as a framework for improving the quality of RD registries. This list includes aspects of governance, Findable, Accessible, Interoperable and Reusable (FAIR) data and information, infrastructure, documentation, training, and quality audit. The list is intended to be used by established as well as new RD registries. Further work includes the development of a toolkit to enable continuous assessment and improvement of their organizational and data quality. Supported by the RD-CONNECT: an integrated platform connecting registries, biobanks and clinical bioinformatics for rare disease research, which received funding from the European Union within the framework of FP7 Collaborative projectHEALTH.2012.2.1.1-1-C [Grant agreement number: 305444]. Supported partly also by EuRRECa: European Registries for Rare Endocrine Conditions, which received funding from the European Union within the framework of CHAFEA Health Programme (2014–2020) [Grant agreement number: 777215] and the COST Action CA16210 “Maximising Impact of research in NeuroDevelopmental Disorders”. Sí

Sample collections and data are hosted within different biobanks at diverse institutions across Europe. Our data integration framework aims at incorporating data about sample collections from different biobanks into a common research infrastructure, facilitating researchers' abilities to obtain high-quality samples to conduct their research. The resulting information must be locally gathered and distributed to searchable higher level information biobank directories to maximize the visibility on the national and European levels. Therefore, biobanks and sample collections must be clearly described and unambiguously identified. We describe how to tackle the challenges of integrating biobank-related data between biobank directories using heterogeneous data schemas and different technical environments.To establish a data exchange infrastructure between all biobank directories involved, we propose the following steps: (A) identification of core entities, terminology, and semantic relationships, (B) harmonization of heterogeneous data schemas of different Biobanking and Biomolecular Resources Research Infrastructure (BBMRI) directories, and (C) formulation of technical core principles for biobank data exchange between directories.(A) We identified the major core elements to describe biobanks in biobank directories. Since all directory data models were partially based on Minimum Information About BIobank Data Sharing (MIABIS) 2.0, the MIABIS 2.0 core model was used for compatibility. (B) Different projection scenarios were elaborated in collaboration with all BBMRI.at partners. A minimum set of mandatory and optional core entities and data items was defined for mapping across all directory levels. (C) Major core data exchange principles were formulated and data interfaces implemented by all biobank directories involved.We agreed on a MIABIS 2.0-based core set of harmonized biobank attributes and established a list of data exchange core principles for integrating biobank directories on different levels. This generic approach and the data exchange core principles proposed herein can also be applied in related tasks like integration and harmonization of biobank data on the individual sample and patient levels.

During the last decade pathology has benefited from the rapid progress of image digitizing technologies, which led to the development of scanners, capable to produce so-called Whole Slide images (WSI) which can be explored by a pathologist on a computer screen comparable to the conventional microscope and can be used for diagnostics, research, archiving and also education and training. Digital pathology is not just the transformation of the classical microscopic analysis of histological slides by pathologists to just a digital visualization. It is a disruptive innovation that will dramatically change medical work-flows in the coming years and help to foster personalized medicine. Really powerful gets a pathologist if she/he is augmented by machine learning, e.g. by support vector machines, random forests and deep learning. The ultimate benefit of digital pathology is to enable to learn, to extract knowledge and to make predictions from a combination of heterogenous data, i.e. the histological image, the patient history and the *omics data. These challenges call for integrated/integrative machine learning approach fostering transparency, trust, acceptance and the ability to explain step-by-step why a decision has been made.

Next-generation sequencing (NGS) has transformed genomic research by decreasing the cost of sequencing and increasing the throughput. Today the translation of NGS technology to (cancer) diagnostics and therapeutics is widely anticipated. In this paper we give an overview of design, user experience and usability (DUXU) requirements for NGS applications with a special focus on clinical and diagnostic settings and describe a data driven GUI and visualization framework for a NGS based clinical decision support system (DSS). In such a DSS it is hard to oversee the huge amount of information, especially if given only in textual form. To overcome this we developed a visual language for electronic health records (EHR) for navigation within a case and comparison of medical cases.

The paper presents a method of predicting classification accuracy of remote sensing data by means of training set analysis. Various sampling plans were applied to satellite image and its complete ground truth to derive different training sets. The quality of these training sets was determined by quantifying the similarity of the training set distributions to the ones of the entire satellite image. Each training set was then used to learn a classifier.The paper shows how the accuracy of classifications that were carried out using these classifiers depends upon the quality of the corresponding training sets.

Next-generation sequencing (NGS) has transformed genomic research by decreasing the cost of sequencing and increasing the throughput. Today, translation of NGS technology to (cancer) diagnostics and therapeutics is widely anticipated. In this paper we give an overview of design, user experience and usability (DUXU) requirements for NGS applications with a special focus on clinical and diagnostic settings.

Keratin 8 (K8) and keratin 18 (K18) form the major hepatocyte cytoskeleton. We investigated the impact of genetic loss of either K8 or K18 on liver homeostasis under toxic stress with the hypothesis that K8 and K18 exert different functions. krt8⁻/⁻ and krt18⁻/⁻ mice crossed into the same 129-ola genetic background were treated by acute and chronic administration of 3,5-diethoxy-carbonyl-1,4-dihydrocollidine (DDC). In acutely DDC-intoxicated mice, macrovesicular steatosis was more pronounced in krt8⁻/⁻ and krt18⁻/⁻ compared with wild-type (wt) animals. Mallory-Denk bodies (MDBs) appeared in krt18⁻/⁻ mice already at an early stage of intoxication in contrast to krt8⁻/⁻ mice that did not display MDB formation when fed with DDC. Keratin-deficient mice displayed significantly lower numbers of apoptotic hepatocytes than wt animals. krt8⁻/⁻, krt18⁻/⁻ and control mice displayed comparable cell proliferation rates. Chronically DDC-intoxicated krt18⁻/⁻ and wt mice showed a similarly increased degree of steatohepatitis with hepatocyte ballooning and MDB formation. In krt8⁻/⁻ mice, steatosis was less, ballooning, and MDBs were absent. krt18⁻/⁻ mice developed MDBs whereas krt8⁻/⁻ mice on the same genetic background did not, highlighting the significance of different structural properties of keratins. They are independent of the genetic background as an intrinsic factor. By contrast, toxicity effects may depend on the genetic background. krt8⁻/⁻ and krt18⁻/⁻ mice on the same genetic background show similar sensitivity to DDC intoxication and almost resemble wt animals regarding survival, degree of porphyria, liver-to-body weight ratio, serum bilirubin and liver enzyme levels. This stands in contrast to previous work where krt8⁻/⁻ and krt18⁻/⁻ mice on different genetic backgrounds were investigated.

Biobanks are essential for the realization of P4-medicine, hence indispensable for smart health. One of the grand challenges in biobank research is to close the research cycle in such a way that all the data generated by one research study can be consistently associated to the original samples, therefore data and knowledge can be reused in other studies. A catalogue must provide the information hub connecting all relevant information sources. The key knowledge embedded in a biobank catalogue is the availability and quality of proper samples to perform a research project. Depending on the study type, the samples can reflect a healthy reference population, a cross sectional representation of a certain group of people (healthy or with various diseases) or a certain disease type or stage. To overview and compare collections from different catalogues, we introduce visual analytics techniques, especially glyph based visualization techniques, which were successfully applied for knowledge discovery of single biobank catalogues. In this paper, we describe the state-of-the art in the integration of biobank catalogues addressing the challenge of combining heterogeneous data sources in a unified and meaningful way, consequently enabling the discovery and visualization of data from different sources. Finally we present open questions both in data integration and visualization of unified catalogues and propose future research in data integration with a linked data approach and the fusion of multi level glyph and network visualization.

This paper describes how to make use of e-books that look like printed books in a knowledge network. After an overview of digitalization efforts and current digital library initiatives we introduce quality measures for the digitalization process. After digitalization an Interactive Internet Book (IIB) has to offer a kind of digital binding, annotation efforts and sophisticated ways for user interaction. We claim that the quality and the enhancements of an Interactive Internet Book go far beyond what is traditionally assumed: it is not enough to scan books. The scans have to be of high quality, allow good OCR to permit full text searches; books need not only be "packaged" but also need meta-data and functionalities that one can expect from a computer supported medium that go far beyond what is possible with traditional printed books. Those factors are critical for the use of e-books in social media environments, yet this is often still overlooked. Finally, we describe a working prototype and demonstrate the advantages obtained with a use case.

As a hub of information controlled by the patient, personal health records (PHR) collect information from the patient medical history including a wide variety of data sources as patient's observations, lab results, clinical findings and in the future maybe even personal genetic data and automatic recordings from monitoring devices. This development will on the one hand make health care more personalized and user controlled but on the other hand also overloads consumers with a huge amount of data. To address this issue we developed a framework for adaptive visual symbols (AVS). An AVS can adapt its appearance and level of detail during the communication process. Finally we demonstrate the AVS principle for the visualization of personal health records.

We present an architectural model for adaptive interfaces based on eye-gaze patterns and facial expression analysis. In our approach, each basic visual sign can adapt its appearance and level of detail during the communication process. Atomic Communication Units (ACUs) -- analogous to graphical output primitives - encapsulate the intended denotation, the encoding of the message and a method for the judgment of the communication goal. We have analyzed feedback cycles in human-human communication tasks, and propose applications scenarios for ACUs.

Head injury with or without polytrauma is the most important cause of death and severe morbidity in an age bracket of up to 45 years. Two major factors are determining its outcome, the extent and nature of the primary irreversible brain injury, and the subsequently developing manifestations of secondary brain damage, which in principle can be prevented by the management procedures and therapeutical interventions. Therefore, a better outcome from severe head injury depends exclusively on a higher efficiency of the management and treatment in order to inhibit secondary brain damage.

The change from analogue broadcasting to digital MPEG-2 channels among the satellite programs resulted in new demands on video databases and archives. Digital archives offer on the one hand a reduction of storage costs, and enable on the other hand easy reuse of already existing material. However, searching for appropriate film material in large archives is still a tedious problem. This paper describes a generic annotation model for MPEG movies which enables the user to structure the film in as many hierarchical levels as needed and to annotate any physical or logical part of the film with generic definable attributes. The model was implemented in prototype system which additionally offers a query and ordering facility per web browser and Internet.

The domain of biobanking has gone through many stages and as a result there are a wide range of commercial and open source software solutions available. The utilization of these software tools requires different levels of domain and technical skills for installation, configuration and ultimate us of these biobank software tools. To compound this complexity the biobanking community are required to work together in order to share knowledge and jointly build solutions to underpin the research infrastructure. We have evaluated the available tools, described them in a catalogue (BiobankApps) and made a selection of tools available to biobanks in a reference toolbox (BIBBOX) that are use-case driven. In the BiobankApps tool catalogue, both commercial and open source software solutions related to the biobanking domain are included, classified and evaluated. The evaluation covers: 1) "user review" by an authenticated user 2) domain expert: quick analysis by BBMRI members and 3) domain expert: detailed analysis and test installation with real world data. The evaluation is paired with a survey across the more "advanced" (from a technology perspective) biobanks to investigate what tools are currently used and summarises known benefits/drawbacks of the respective packages. In the second step we recommend tools for specific use cases, and install, configure and connect these in the BIBBOX framework. This service also builds on the existing work in the United Kingdom in seeking to establish the motivations for different stakeholders to become involved and therefore assisting in prioritising the use-cases based on the level of need and support within the research community. All tools associated to a use-case are available as BIBBOX applications (technically this is achieved by docker containers), which are integrated in the BIBBOX framework with central identification and user management. In future work we plan to share the acquired knowledge with other networks, develop an Application Programmable Interface (API) for the exchange of metadata with other tool catalogues and work on an ontology for the evaluation of biobank software.

Artificial Intelligence (AI) for the biomedical domain is gaining significant interest and holds considerable potential for the future of healthcare, particularly also in the context of in vitro diagnostics. The European In Vitro Diagnostic Medical Device Regulation (IVDR) explicitly includes software in its requirements. This poses major challenges for In Vitro Diagnostic devices (IVDs) that involve Machine Learning (ML) algorithms for data analysis and decision support. This can increase the difficulty of applying some of the most successful ML and Deep Learning (DL) methods to the biomedical domain, just by missing the required explanatory components from the manufacturers. In this context, trustworthy AI has to empower biomedical professionals to take responsibility for their decision-making, which clearly raises the need for explainable AI methods. Explainable AI, such as layer-wise relevance propagation, can help in highlighting the relevant parts of inputs to, and representations in, a neural network that caused a result and visualize these relevant parts. In the same way that usability encompasses measurements for the quality of use, the concept of causability encompasses measurements for the quality of explanations produced by explainable AI methods. This paper describes both concepts and gives examples of how explainability and causability are essential in order to demonstrate scientific validity as well as analytical and clinical performance for future AI-based IVDs.

Biological samples such as tissues, blood and other body fluids, plants or seeds, prokaryotic and eukaryotic cells or isolated biomolecules as well as associated data are the essential raw material for research and development in medicine, biotechnology and agriculture. The collection, processing, preservation, and storage of these resources, in addition to provision of access, are key activities of biobanks or biological resource centres. Biobanks have to ensure proper quality of samples and data, ethical and legal compliance as well as transparent and efficient access procedures. In this context the review places special emphasis on pre-analytical procedures and international standards, which are essential to improving analytical data reliability and reproducibility, as well as on the increasing importance of data management. These requirements of biobanks are demonstrated using the example of pathogen-containing and microbiome biobanks, and refer to needs in cancer research and development.

Provenance is information describing the lineage of an object, such as a dataset or biological material. Since these objects can be passed between organizations, each organization can document only parts of the objects life cycle. As a result, interconnection of distributed provenance parts forms distributed provenance chains. Dependant on the actual provenance content, complete provenance chains can provide traceability and contribute to reproducibility and FAIRness of research objects. In this paper, we define a lightweight provenance model based on W3C PROV that enables generation of distributed provenance chains in complex, multi-organizational environments. The application of the model is demonstrated with a use case spanning several steps of a real-world research pipeline — starting with the acquisition of a specimen, its processing and storage, histological examination, and the generation/collection of associated data (images, annotations, clinical data), ending with training an AI model for the detection of tumor in the images. The proposed model has become an open conceptual foundation of the currently developed ISO 23494 standard on provenance for biotechnology domain.

Background This paper presents multilevel data glyphs optimized for the interactive knowledge discovery and visualization of large biomedical data sets. Data glyphs are three- dimensional objects defined by multiple levels of geometric descriptions (levels of detail) combined with a mapping of data attributes to graphical elements and methods, which specify their spatial position. Methods In the data mapping phase, which is done by a biomedical expert, meta information about the data attributes (scale, number of distinct values) are compared with the visual capabilities of the graphical elements in order to give a feedback to the user about the correctness of the variable mapping. The spatial arrangement of glyphs is done in a dimetric view, which leads to high data density, a simplified 3D navigation and avoids perspective distortion. Results We show the usage of data glyphs in the disease analyser a visual analytics application for personalized medicine and provide an outlook to a biomedical web visualization scenario. Conclusions Data glyphs can be successfully applied in the disease analyser for the analysis of big medical data sets. Especially the automatic validation of the data mapping, selection of subgroups within histograms and the visual comparison of the value distributions were seen by experts as an important functionality.

Quantum Key Distribution typically delivers symmetric keys to protect data in transit. For data at rest the method of secret sharing describes the distribution of data among a given number of storages in such a way that for reconstruction only a certain smaller number of shares is strictly required. This is proven mathematically with the consequence to categorize secret sharing to be ITS matching well to QKD. The research of the FP7-funded projects Archistar and Prismacloud [1] have been commercialized by the spin-off fragmentiX and subsequently used for OpenQKD [2] . As a medical use case in OpenQKD Whole Slide Images from digital pathology together with genetic data have been split into three fragments with the requirement to combine two fragments to retrieve full information. These fragments have been stored at different locations and the distribution to the datacenters has been secured by QKD.