Data infrastructures for AI in medical imaging: a report on the experiences of five EU projects.

Artificial intelligence (AI) is transforming the field of medical imaging and has the potential to bring medicine from the era of 'sick-care' to the era of healthcare and prevention. The development of AI requires access to large, complete, and harmonized real-world datasets, representative of the population, and disease diversity. However, to date, efforts are fragmented, based on single-institution, size-limited, and annotation-limited datasets. Available public datasets (e.g., The Cancer Imaging Archive, TCIA, USA) are limited in scope, making model generalizability really difficult. In this direction, five European Union projects are currently working on the development of big data infrastructures that will enable European, ethically and General Data Protection Regulation-compliant, quality-controlled, cancer-related, medical imaging platforms, in which both large-scale data and AI algorithms will coexist. The vision is to create sustainable AI cloud-based platforms for the development, implementation, verification, and validation of trustable, usable, and reliable AI models for addressing specific unmet needs regarding cancer care provision. In this paper, we present an overview of the development efforts highlighting challenges and approaches selected providing valuable feedback to future attempts in the area.Key points• Artificial intelligence models for health imaging require access to large amounts of harmonized imaging data and metadata.• Main infrastructures adopted either collect centrally anonymized data or enable access to pseudonymized distributed data.• Developing a common data model for storing all relevant information is a challenge.• Trust of data providers in data sharing initiatives is essential.• An online European Union meta-tool-repository is a necessity minimizing effort duplication for the various projects in the area.

CHAIMELEON Project: Creation of a Pan-European Repository of Health Imaging Data for the Development of AI-Powered Cancer Management Tools.

The CHAIMELEON project aims to set up a pan-European repository of health imaging data, tools and methodologies, with the ambition to set a standard and provide resources for future AI experimentation for cancer management. The project is a 4 year long, EU-funded project tackling some of the most ambitious research in the fields of biomedical imaging, artificial intelligence and cancer treatment, addressing the four types of cancer that currently have the highest prevalence worldwide: lung, breast, prostate and colorectal. To allow this, clinical partners and external collaborators will populate the repository with multimodality (MR, CT, PET/CT) imaging and related clinical data. Subsequently, AI developers will enable a multimodal analytical data engine facilitating the interpretation, extraction and exploitation of the information stored at the repository. The development and implementation of AI-powered pipelines will enable advancement towards automating data deidentification, curation, annotation, integrity securing and image harmonization. By the end of the project, the usability and performance of the repository as a tool fostering AI experimentation will be technically validated, including a validation subphase by world-class European AI developers, participating in Open Challenges to the AI Community. Upon successful validation of the repository, a set of selected AI tools will undergo early in-silico validation in observational clinical studies coordinated by leading experts in the partner hospitals. Tool performance will be assessed, including external independent validation on hallmark clinical decisions in response to some of the currently most important clinical end points in cancer. The project brings together a consortium of 18 European partners including hospitals, universities, R&D centers and private research companies, constituting an ecosystem of infrastructures, biobanks, AI/in-silico experimentation and cloud computing technologies in oncology.

Serverless Workflows for Containerised Applications in the Cloud Continuum.

This paper introduces an open-source platform to support serverless computing for scientific data-processing workflow-based applications across the Cloud continuum (i.e. simultaneously involving both on-premises and public Cloud platforms to process data captured at the edge). This is achieved via dynamic resource provisioning for FaaS platforms compatible with scale-to-zero approaches that minimise resource usage and cost for dynamic workloads with different elasticity requirements. The platform combines the usage of dynamically deployed auto-scaled Kubernetes clusters on on-premises Clouds and automated Cloud bursting into AWS Lambda to achieve higher levels of elasticity. A use case in public health for smart cities is used to assess the platform, in charge of detecting people not wearing face masks from captured videos. Faces are blurred for enhanced anonymity in the on-premises Cloud and detection via Deep Learning models is performed in AWS Lambda for this data-driven containerised workflow. The results indicate that hybrid workflows across the Cloud continuum can efficiently perform local data processing for enhanced regulations compliance and perform Cloud bursting for increased levels of elasticity.

ReMindCare, an app for daily clinical practice in patients with first episode psychosis: A pragmatic real-world study protocol.

AIM: Despite the potential benefits of e-health interventions for patients with psychosis, the integration of these applications into the clinical workflow and analysis of their long-term effects still face significant challenges. To address these issues, we developed the ReMindCare app. This app aims to improve the treatment quality for patients with psychosis. We chose to study the app in real world and pragmatic manner to ensure results will be generalizable. METHODS: This is a naturalistic empirical study of patients in a first episode of psychosis programme. The app was purpose-designed based on two previous studies, and it offers the following assessments: (a) three daily questions regarding anxiety, sadness and irritability; and (b) 18 weekly questions about medication adherence, medication side effects, medication attitudes and prodromal symptoms. The app offers preset alerts, reminders and the ability for patients to reach out to their clinicians. Data captured by the app are linked to the electronic medical record of the patient. Patients will use the app as part of their ongoing care for a maximum period of 5 years, and assessments will occur at baseline and at the end of the first, second and fifth years of app use. RESULTS: Recruitment started in October 2018 and is still ongoing. CONCLUSIONS: The ReMindCare app represents early real-world use of digital mental health tools that offer direct integration into clinical care. High retention and compliance rates are expected, and this will in turn lead to improved quality of assessments and communication between patients and clinicians.

ReMindCare App for Early Psychosis: Pragmatic Real World Intervention and Usability Study.

BACKGROUND: eHealth interventions are widely used in clinical trials and increasingly in care settings as well; however, their efficacy in real-world contexts remains unknown. ReMindCare is a smartphone app that has been systematically implemented in a first episode of psychosis program (FEPP) for patients with early psychosis since 2018. OBJECTIVE: The objective of this study was to assess the efficacy of ReMindCare after 19 months of use in the clinic and varying use by individual patients. METHODS: The integration of the ReMindCare app into the FEPP started in October 2018. Patients with early psychosis self-selected to the app (ReMindCare group) or treatment as usual (TAU group). The outcome variables considered were adherence to the intervention and number of relapses, hospital admissions, and visits to urgent care units. Data from 90 patients with early psychosis were analyzed: 59 in the ReMindCare group and 31 in the TAU group. The mean age of the sample was 32.8 (SD 9.4) years, 73% (66/90) were males, 91% (83/90) were White, and 81% (74/90) were single. RESULTS: Significant differences between the ReMindCare and TAU groups were found in the number of relapses, hospitalizations, and visits to urgent care units, with each showing benefits for the app. Only 20% (12/59) of patients from the ReMindCare group had a relapse, while 58% (18/31) of the TAU patients had one or more relapses (χ2=13.7, P=.001). Moreover, ReMindCare patients had fewer visits to urgent care units (χ2=7.4, P=.006) and fewer hospitalizations than TAU patients (χ2=4.6, P=.03). The mean of days using the app was 352.2 (SD 191.2; min/max: 18-594), and the mean of engagement was 84.5 (SD 16.04). CONCLUSIONS: To our knowledge, this is the first eHealth intervention that has preliminarily proven its benefits in the real-world treatment of patients with early psychosis. INTERNATIONAL REGISTERED REPORT IDENTIFIER (IRRID): RR2-10.1111/eip.12960.

PRIMAGE project: predictive in silico multiscale analytics to support childhood cancer personalised evaluation empowered by imaging biomarkers.

PRIMAGE is one of the largest and more ambitious research projects dealing with medical imaging, artificial intelligence and cancer treatment in children. It is a 4-year European Commission-financed project that has 16 European partners in the consortium, including the European Society for Paediatric Oncology, two imaging biobanks, and three prominent European paediatric oncology units. The project is constructed as an observational in silico study involving high-quality anonymised datasets (imaging, clinical, molecular, and genetics) for the training and validation of machine learning and multiscale algorithms. The open cloud-based platform will offer precise clinical assistance for phenotyping (diagnosis), treatment allocation (prediction), and patient endpoints (prognosis), based on the use of imaging biomarkers, tumour growth simulation, advanced visualisation of confidence scores, and machine-learning approaches. The decision support prototype will be constructed and validated on two paediatric cancers: neuroblastoma and diffuse intrinsic pontine glioma. External validation will be performed on data recruited from independent collaborative centres. Final results will be available for the scientific community at the end of the project, and ready for translation to other malignant solid tumours.

Differences in the Use and Opinions About New eHealth Technologies Among Patients With Psychosis: Structured Questionnaire.

BACKGROUND: Despite a growing interest in the use of technology in order to support the treatment of psychotic disorders, limited knowledge exists about the viability and acceptability of these eHealth interventions in relation to the clinical characteristics of patients. OBJECTIVE: The objective of this study was to assess the access and use of, as well as experiences and interest in, new technologies using a survey of patients diagnosed with early psychosis compared with a survey of patients diagnosed with chronic psychotic disorders. METHODS: We designed a structured questionnaire. This questionnaire was divided into five parts: (1) clinical and demographic information, (2) access and use of the internet, (3) use of the internet in relation to mental health, (4) experiences with technology, and (5) patients' interest in eHealth services. In total, 105 patients were recruited from early psychosis units (n=65) and recovery units (n=40). RESULTS: In this study, 84.8% (89/105) of the patients had access to the internet and 88.6% (93/105) owned an electronic internet device. In total, 71.3% (57/80) of patients who owned a mobile phone were interested in eHealth systems and 38.2% (37/97) reported negative experiences related to the internet usage. We observed differences between the groups in terms of device ownership (P=.02), the frequency of internet access (P<.001), the use of social media (P=.01), and seeking health information (P=.04); the differences were found to be higher in the early psychosis group. No differences were found between the groups in terms of the use of internet in relation to mental health, experiences and opinions about the internet, or interest in eHealth interventions (P=.43). CONCLUSIONS: The availability and use of technology for the participants in our survey were equivalent to those for the general population. The differences found between the groups in relation to the access or use of technology seemed to due to age-related factors. The use of technology involving mental health and the interest in eHealth interventions were mainly positive and equivalent between the groups. Accordingly, this group of patients is a potential target for the emerging eHealth interventions, regardless of their clinical status. However, 28.7% (23/80) of the studied patients rejected the use of internet interventions and 38.2% (37/97) had unpleasant experiences related to its usage; thus, more in-depth studies are needed to better define the profile of patients with psychosis who may benefit from eHealth treatments.

Use of mobile technologies in patients with psychosis: A systematic review. / Utilización de tecnologías móviles en pacientes con psicosis: una revisión sistemática.

There is a growing interest in mobile Health interventions (m-Health) in patients with psychosis. The aim of this study is to conduct a systematic review in order to analysethe current state of research in this area. The search of articles was carried out following the PRISMA criteria, focusing on those studies that used mobile technologies in patients with psychosis during the period from 1990 to 2016. A total of 20 articles were selected from the 431 studies found. Three types of studies are distinguished: 1) Analysis of quality and usability, 2) Improving treatment adherence and reducing hospital admissions, and 3) Analysisof patient symptoms. CONCLUSIONS: m-Health interventions are feasible, and are easy to use for patients with psychosis. They evaluate the evolution of psychotic symptoms more efficiently, and improve adherence to treatment, as well as symptoms and hospital admissions. However, a particular strategy does not stand out over the rest, because differences in methodology make them difficult to compare.

Increasing the Efficiency on Producing Radiology Reports for Breast Cancer Diagnosis by Means of Structured Reports. A Comparative Study.

BACKGROUND: Radiology reports are commonly written on free-text using voice recognition devices. Structured reports (SR) have a high potential but they are usually considered more difficult to fill-in so their adoption in clinical practice leads to a lower efficiency. However, some studies have demonstrated that in some cases, producing SRs may require shorter time than plain-text ones. This work focuses on the definition and demonstration of a methodology to evaluate the productivity of software tools for producing radiology reports. A set of SRs for breast cancer diagnosis based on BI-RADS have been developed using this method. An analysis of their efficiency with respect to free-text reports has been performed. MATERIAL AND METHODS: The methodology proposed compares the Elapsed Time (ET) on a set of radiological reports. Free-text reports are produced with the speech recognition devices used in the clinical practice. Structured reports are generated using a web application generated with TRENCADIS framework. A team of six radiologists with three different levels of experience in the breast cancer diagnosis was recruited. These radiologists performed the evaluation, each one introducing 50 reports for mammography, 50 for ultrasound scan and 50 for MRI using both approaches. Also, the Relative Efficiency (REF) was computed for each report, dividing the ET of both methods. We applied the T-Student (T-S) test to compare the ETs and the ANOVA test to compare the REFs. Both tests were computed using the SPSS software. RESULTS: The study produced three DICOM-SR templates for Breast Cancer Diagnosis on mammography, ultrasound and MRI, using RADLEX terms based on BIRADs 5th edition. The T-S test on radiologists with high or intermediate profile, showed that the difference between the ET was only statistically significant for mammography and ultrasound. The ANOVA test performed grouping the REF by modalities, indicated that there were no significant differences between mammograms and ultrasound scans, but both have significant statistical differences with MRI. The ANOVA test of the REF for each modality, indicated that there were only significant differences in Mammography (ANOVA p = 0.024) and Ultrasound (ANOVA p = 0.008). The ANOVA test for each radiologist profile, indicated that there were significant differences on the high profile (ANOVA p = 0.028) and medium (ANOVA p = 0.045). CONCLUSIONS: In this work, we have defined and demonstrated a methodology to evaluate the productivity of software tools for producing radiology reports in Breast Cancer. We have evaluated that adopting Structured Reporting in mammography and ultrasound studies in breast cancer diagnosis improves the performance in producing reports.

MuffinInfo: HTML5-Based Statistics Extractor from Next-Generation Sequencing Data.

Usually, the information known a priori about a newly sequenced organism is limited. Even resequencing the same organism can generate unpredictable output. We introduce MuffinInfo, a FastQ/Fasta/SAM information extractor implemented in HTML5 capable of offering insights into next-generation sequencing (NGS) data. Our new tool can run on any software or hardware environment, in command line or graphically, and in browser or standalone. It presents information such as average length, base distribution, quality scores distribution, k-mer histogram, and homopolymers analysis. MuffinInfo improves upon the existing extractors by adding the ability to save and then reload the results obtained after a run as a navigable file (also supporting saving pictures of the charts), by supporting custom statistics implemented by the user, and by offering user-adjustable parameters involved in the processing, all in one software. At the moment, the extractor works with all base space technologies such as Illumina, Roche, Ion Torrent, Pacific Biosciences, and Oxford Nanopore. Owing to HTML5, our software demonstrates the readiness of web technologies for mild intensive tasks encountered in bioinformatics.

Fast inexact mapping using advanced tree exploration on backward search methods.

BACKGROUND: Short sequence mapping methods for Next Generation Sequencing consist on a combination of seeding techniques followed by local alignment based on dynamic programming approaches. Most seeding algorithms are based on backward search alignment, using the Burrows Wheeler Transform, the Ferragina and Manzini Index or Suffix Arrays. All these backward search algorithms have excellent performance, but their computational cost highly increases when allowing errors. In this paper, we discuss an inexact mapping algorithm based on pruning strategies for search tree exploration over genomic data. RESULTS: The proposed algorithm achieves a 13x speed-up over similar algorithms when allowing 6 base errors, including insertions, deletions and mismatches. This algorithm can deal with 400 bps reads with up to 9 errors in a high quality Illumina dataset. In this example, the algorithm works as a preprocessor that reduces by 55% the number of reads to be aligned. Depending on the aligner the overall execution time is reduced between 20-40%. CONCLUSIONS: Although not intended as a complete sequence mapping tool, the proposed algorithm could be used as a preprocessing step to modern sequence mappers. This step significantly reduces the number reads to be aligned, accelerating overall alignment time. Furthermore, this algorithm could be used for accelerating the seeding step of already available sequence mappers. In addition, an out-of-core index has been implemented for working with large genomes on systems without expensive memory configurations.

NeuPAT: an intranet database supporting translational research in neuroblastic tumors.

Translational research in oncology is directed mainly towards establishing a better risk stratification and searching for appropriate therapeutic targets. This research generates a tremendous amount of complex clinical and biological data needing speedy and effective management. The authors describe the design, implementation and early experiences of a computer-aided system for the integration and management of data for neuroblastoma patients. NeuPAT facilitates clinical and translational research, minimizes the workload in consolidating the information, reduces errors and increases correlation of data through extensive coding. This design can also be applied to other tumor types.

A service-based BLAST command tool supported by cloud infrastructures.

Notwithstanding the benefits of distributed-computing infrastructures for empowering bioinformatics analysis tools with the needed computing and storage capability, the actual use of these infrastructures is still low. Learning curves and deployment difficulties have reduced the impact on the wide research community. This article presents a porting strategy of BLAST based on a multiplatform client and a service that provides the same interface as sequential BLAST, thus reducing learning curve and with minimal impact on their integration on existing workflows. The porting has been done using the execution and data access components from the EC project Venus-C and the Windows Azure infrastructure provided in this project. The results obtained demonstrate a low overhead on the global execution framework and reasonable speed-up and cost-efficiency with respect to a sequential version.

Integrating TRENCADIS components in gLite to share DICOM medical images and structured reports.

The problem of sharing medical information among different centres has been tackled by many projects. Several of them target the specific problem of sharing DICOM images and structured reports (DICOM-SR), such as the TRENCADIS project. In this paper we propose sharing and organizing DICOM data and DICOM-SR metadata benefiting from the existent deployed Grid infrastructures compliant with gLite such as EGEE or the Spanish NGI. These infrastructures contribute with a large amount of storage resources for creating knowledge databases and also provide metadata storage resources (such as AMGA) to semantically organize reports in a tree-structure. First, in this paper, we present the extension of TRENCADIS architecture to use gLite components (LFC, AMGA, SE) on the shake of increasing interoperability. Using the metadata from DICOM-SR, and maintaining its tree structure, enables federating different but compatible diagnostic structures and simplifies the definition of complex queries. This article describes how to do this in AMGA and it shows an approach to efficiently code radiology reports to enable the multi-centre federation of data resources.

Using grid-enabled distributed metadata database to index DICOM-SR.

Integrating medical data at inter-centre level implies many challenges that are being tackled from many disciplines and technologies. Medical informatics have applied an important effort on describing and standardizing Electronic Health Records, and specially standardisation has achieved an important extent on Medical Imaging. Grid technologies have been extensively used to deal with multi-domain authorisation issues and to provide single access points for accessing DICOM Medical Images, enabling the access and processing to large repositories of data. However, this approach introduces the challenge of efficiently organising data according to their relevance and interest, in which the medical report is a key factor. The present work shows an approach to efficiently code radiology reports to enable the multi-centre federation of data resources. This approach follows the tree-like structure of DICOM-SR reports in a self-organising metadata catalogue based on AMGA. This approach enables federating different but compatible distributed repositories, automatically reconfiguring the database structure, and preserving the autonomy of each centre in defining the template. Tools developed so far and some performance results are provided to prove the effectiveness of the approach.

SHARE road map for HealthGrids: methodology.

The SHARE(1) project (http://www.eu-share.org) was asked to identify the key developments needed to achieve wide adoption and deployment of HealthGrids throughout Europe. The project was asked to organise these as milestones on a road map, so that all technical advances, social actions, economic investments and ethical or legal initiatives necessary for HealthGrids would be seen together in a single coherent document. The full road map includes an extensive analysis of several case studies exploring their technical requirements, full discussion of the ethical, legal, social and economic issues which may impede early deployment, and concludes with an attempt to reconcile the tensions between technological developments and regulatory frameworks. This paper has been restricted to the technical aspects of the project. SHARE built on the work of the 'HealthGrid' initiative so we begin by, reviewing work carried out in various European HealthGrid projects and report on joint work with numerous European collaborators. Following many successful HealthGrid projects, HealthGrid published a 'White Paper' which establishes the foundations, potential scope and prospects of an approach to health informatics based on a grid infrastructure. The White Paper demonstrates the ways in which the HealthGrid approach supports many modern trends in medicine and healthcare, such as evidence-based practice, integration across levels, from molecules and cells, through tissues and organs to the whole person and community, and the promise of individualised healthcare. SHARE was funded by the European Commission to define a research roadmap for a 'HealthGrid for Europe', to be seen as the preferred infrastructure for biomedical and healthcare projects in the European Research Area.

A highly optimized grid deployment: the metagenomic analysis example.

Computational resources and computationally expensive processes are two topics that are not growing at the same ratio. The availability of large amounts of computing resources in Grid infrastructures does not mean that efficiency is not an important issue. It is necessary to analyze the whole process to improve partitioning and submission schemas, especially in the most critical experiments. This is the case of metagenomic analysis, and this text shows the work done in order to optimize a Grid deployment, which has led to a reduction of the response time and the failure rates. Metagenomic studies aim at processing samples of multiple specimens to extract the genes and proteins that belong to the different species. In many cases, the sequencing of the DNA of many microorganisms is hindered by the impossibility of growing significant samples of isolated specimens. Many bacteria cannot survive alone, and require the interaction with other organisms. In such cases, the information of the DNA available belongs to different kinds of organisms. One important stage in Metagenomic analysis consists on the extraction of fragments followed by the comparison and analysis of their function stage. By the comparison to existing chains, whose function is well known, fragments can be classified. This process is computationally intensive and requires of several iterations of alignment and phylogeny classification steps. Source samples reach several millions of sequences, which could reach up to thousands of nucleotides each. These sequences are compared to a selected part of the "Non-redundant" database which only implies the information from eukaryotic species. From this first analysis, a refining process is performed and alignment analysis is restarted from the results. This process implies several CPU years. The article describes and analyzes the difficulties to fragment, automate and check the above operations in current Grid production environments. This environment has been tuned-up from an experimental study which has tested the most efficient and reliable resources, the optimal job size, and the data transference and database reindexation overhead. The environment should re-submit faulty jobs, detect endless tasks and ensure that the results are correctly retrieved and workflow synchronised. The paper will give an outline on the structure of the system, and the preparation steps performed to deal with this experiment.

SHARE, from vision to road map: technical steps.

We present the 'HealthGrid' initiative and briefly review work carried out in various European healthgrid projects. We report on joint work with numerous European collaborators. Since the European Commission's Information Society Technologies programme funded the first gridbased health and medical projects, the HealthGrid movement has flourished in Europe. Many projects have now been completed and 'HealthGrid' consulted a number of experts to compile and publish a 'White Paper' which establishes the foundations, potential scope and prospects of an approach to health informatics based on a grid infrastructure. With a second generation of projects now funded, the EC has commissioned the SHARE Project, a study to define a research roadmap for a 'healthgrid for Europe' as the preferred infrastructure for medical and health care projects in the European Research Area. The project explores the ways in which the healthgrid approach supports modern trends both in research in biomedicine and in healthcare, such as evidence-based practice and information integration.

Advances in the biomedical applications of the EELA Project.

In the last years an increasing demand for Grid Infrastructures has resulted in several international collaborations. This is the case of the EELA Project, which has brought together collaborating groups of Latin America and Europe. One year ago we presented this e-infrastructure used, among others, by the biomedical groups for the studies of oncological analysis, neglected diseases, sequence alignments and computational phylogenetics. After this period, the achieved advances are summarised in this paper.