MISM: A Medical Image Segmentation Metric for Evaluation of Weak Labeled Data.

Performance measures are an important tool for assessing and comparing different medical image segmentation algorithms. Unfortunately, the current measures have their weaknesses when it comes to assessing certain edge cases. These limitations arise when images with a very small region of interest or without a region of interest at all are assessed. As a solution to these limitations, we propose a new medical image segmentation metric: MISm. This metric is a composition of the Dice similarity coefficient and the weighted specificity. MISm was investigated for definition gaps, an appropriate scoring gradient, and different weighting coefficients used to propose a constant value. Furthermore, an evaluation was performed by comparing the popular metrics in the medical image segmentation and MISm using images of magnet resonance tomography from several fictitious prediction scenarios. Our analysis shows that MISm can be applied in a general way and thus also covers the mentioned edge cases, which are not covered by other metrics, in a reasonable way. In order to allow easy access to MISm and therefore widespread application in the community, as well as reproducibility of experimental results, we included MISm in the publicly available evaluation framework MISeval.

Evaluation of the Availability of Nursing Quality Indicators in German FHIR Implementations.

Standardized nursing data sets facilitate data analysis and help to improve nursing research and quality management in Germany. Recently, governmental standardization approaches have favored the FHIR standard and helped to define it as the state of the art for healthcare interoperability and data exchange. In this study, we identify common data elements used for nursing quality research purposes by analyzing nursing quality data sets and databases. We then compare the results with current FHIR implementations in Germany to find most relevant data fields and overlaps. Our results show that most of the patient focused information has already been modelled in national standardization efforts and FHIR implementations. However, representation of data fields describing nursing staff related information, such as experience, workload or satisfaction, is missing or lacking.

The RCX Extension Hub: A Resource for Implementations Extending the R Adaption of the Cytoscape Exchange Format.

Public repositories provide access to biological networks for investigations, and subsequently serve to distribute the network encoded biomedical and even clinically relevant results. However, inclusion of complementary information requires data structures and implementations customized to the integrated data for network representation, usage in supporting application, and extending analysis functionality. Partitioning of this information into individual aspects of a network facilitates compatibility and reusability of the network-based results, but also requires support and accessibility of the extensions and their implementations. The RCX extension hub offers overview and access to extensions of the Cytoscape exchange format implemented in R. The hub supports the realization of self-customized extension through guides, example implementations, and a template for the creation of R extension packages.

Web-based Prototype for Graphical Exploration of FHIR® Questionnaire Responses.

The evaluation of clinical questionnaires is an important part of gaining knowledge in empirical research. The electronically captured responses are encoded in a standard format such as HL7 FHIR® that facilitates data exchange and systems interoperability. However, this also complicates access of the information to explore and interpret the results without appropriate tools. In this work, we present the design of a web-based graphical exploration tool for categorical questionnaire response data that can interact with FHIR-conformant HTTP endpoints. The web app enables non-technical users with simplified, direct visual access to highly structured FHIR questionnaire response data and preserves the applicability in arbitrary data exploration tasks. We describe the abstract feature design with the derived technical implementation to allow a universal, user-configurable data subselection mechanism to generate conditional one- and two-data-dimensional charts. The applicability of our developed prototype is demonstrated on synthetic FHIR data with the source code available at https://github.com/frankkramer-lab/FHIR-QR-Explorer.

Adaptation of HL7 FHIR for the Exchange of Patients' Gene Expression Profiles.

High-throughput technologies, especially gene expression analyses can accurately capture the molecular state in patients under different conditions. Thus, their application in clinical routine gains increasing relevance and fosters patient stratification towards individualized treatment decisions. Electronic health records already evolved to capture genomic data within clinical systems and standards like FHIR enable sharing within, and even between institutions. However, FHIR only provides profiles tailored to variations in the molecular sequence, while expression patterns are neglected although being equally important for decision making. Here we provide an exemplary implementation of gene expression profiles of a microarray analysis of patients with acute myeloid leukemia using an adaptation of the FHIR genomics extension. Our results demonstrate how FHIR resources can be facilitated in clinical systems and thereby pave the way for usage for the aggregation and exchange of transcriptomic data in multi-center studies.

Perspective on Code Submission and Automated Evaluation Platforms for University Teaching.

We present a perspective on platforms for code submission and automated evaluation in the context of university teaching. Due to the COVID-19 pandemic, such platforms have become an essential asset for remote courses and a reasonable standard for structured code submission concerning increasing numbers of students in computer sciences. Utilizing automated code evaluation techniques exhibits notable positive impacts for both students and teachers in terms of quality and scalability. We identified relevant technical and non-technical requirements for such platforms in terms of practical applicability and secure code submission environments. Furthermore, a survey among students was conducted to obtain empirical data on general perception. We conclude that submission and automated evaluation involves continuous maintenance yet lowers the required workload for teachers and provides better evaluation transparency for students.

Data-dependent visualization of biological networks in the web-browser with NDExEdit.

Networks are a common methodology used to capture increasingly complex associations between biological entities. They serve as a resource of biological knowledge for bioinformatics analyses, and also comprise the subsequent results. However, the interpretation of biological networks is challenging and requires suitable visualizations dependent on the contained information. The most prominent software in the field for the visualization of biological networks is Cytoscape, a desktop modeling environment also including many features for analysis. A further challenge when working with networks is their distribution. Within a typical collaborative workflow, even slight changes of the network data force one to repeat the visualization step as well. Also, just minor adjustments to the visual representation not only need the networks to be transferred back and forth. Collaboration on the same resources requires specific infrastructure to avoid redundancies, or worse, the corruption of the data. A well-established solution is provided by the NDEx platform where users can upload a network, share it with selected colleagues or make it publicly available. NDExEdit is a web-based application where simple changes can be made to biological networks within the browser, and which does not require installation. With our tool, plain networks can be enhanced easily for further usage in presentations and publications. Since the network data is only stored locally within the web browser, users can edit their private networks without concerns of unintentional publication. The web tool is designed to conform to the Cytoscape Exchange (CX) format as a data model, which is used for the data transmission by both tools, Cytoscape and NDEx. Therefore the modified network can be directly exported to the NDEx platform or saved as a compatible CX file, additionally to standard image formats like PNG and JPEG.

Implementation of Gene Expression Profiles in the HL7 FHIR Standard.

Gene expression profiles can capture significant molecular differences paving the way toward precision medicine. However, clinical standards like FHIR only provide encoding of molecular sequence variations, even so, expression patterns are equally important for decision making. Here we provide an exemplary implementation of gene expression profiles of a microarray analysis using an adaption of the FHIR Genomics extension. Our results demonstrate how FHIR resources can be facilitated in bioinformatics-based decision support systems or used for the aggregation of molecular genetics data in multi-center clinical trials.

MISeval: A Metric Library for Medical Image Segmentation Evaluation.

Correct performance assessment is crucial for evaluating modern artificial intelligence algorithms in medicine like deep-learning based medical image segmentation models. However, there is no universal metric library in Python for standardized and reproducible evaluation. Thus, we propose our open-source publicly available Python package MISeval: a metric library for Medical Image Segmentation Evaluation. The implemented metrics can be intuitively used and easily integrated into any performance assessment pipeline. The package utilizes modern DevOps strategies to ensure functionality and stability. MISeval is available from PyPI (miseval) and GitHub: https://github.com/frankkramer-lab/miseval.

RCX-an R package adapting the Cytoscape Exchange format for biological networks.

Motivation: The Cytoscape Exchange (CX) format is a JSON-based data structure designed for the transmission of biological networks using standard web technologies. It was developed by the network data exchange, which itself serves as online commons to share and collaborate on biological networks. Furthermore, the Cytoscape software for the analysis and visualization of biological networks contributes structure elements to capture the visual layout within the CX format. However, there is a fundamental difference between data handling in web standards and R. A manual conversion requires detailed knowledge of the CX format to reproduce and work with the networks. Results: Here, we present a software package to create, handle, validate, visualize and convert networks in CX format to standard data types and objects within R. Networks in this format can serve as a source for biological knowledge and also capture the results of the analysis of those while preserving the visual layout across all platforms. The RCX package connects the R environment for statistical computing with outside platforms for storage and collaboration, as well as further analysis and visualization of biological networks. Availability: RCX is a free and open-source R package, available on Bioconductor from release 3.15 (https://bioconductor.org/packages/RCX) and via GitHub (https://github.com/frankkramer-lab/RCX). Supplementary information: Supplementary data are available at Bioinformatics Advances online.

Comprehensive Analysis of Chemical Structures That Have Been Tested as CFTR Activating Substances in a Publicly Available Database CandActCFTR.

Background: Cystic fibrosis (CF) is a genetic disease caused by mutations in CFTR, which encodes a chloride and bicarbonate transporter expressed in exocrine epithelia throughout the body. Recently, some therapeutics became available that directly target dysfunctional CFTR, yet research for more effective substances is ongoing. The database CandActCFTR aims to provide detailed and comprehensive information on candidate therapeutics for the activation of CFTR-mediated ion conductance aiding systems-biology approaches to identify substances that will synergistically activate CFTR-mediated ion conductance based on published data. Results: Until 10/2020, we derived data from 108 publications on 3,109 CFTR-relevant substances via the literature database PubMed and further 666 substances via ChEMBL; only 19 substances were shared between these sources. One hundred and forty-five molecules do not have a corresponding entry in PubChem or ChemSpider, which indicates that there currently is no single comprehensive database on chemical substances in the public domain. Apart from basic data on all compounds, we have visualized the chemical space derived from their chemical descriptors via a principal component analysis annotated for CFTR-relevant biological categories. Our online query tools enable the search for most similar compounds and provide the relevant annotations in a structured way. The integration of the KNIME software environment in the back-end facilitates a fast and user-friendly maintenance of the provided data sets and a quick extension with new functionalities, e.g., new analysis routines. CandActBase automatically integrates information from other online sources, such as synonyms from PubChem and provides links to other resources like ChEMBL or the source publications. Conclusion: CandActCFTR aims to establish a database model of candidate cystic fibrosis therapeutics for the activation of CFTR-mediated ion conductance to merge data from publicly available sources. Using CandActBase, our strategy to represent data from several internet resources in a merged and organized form can also be applied to other use cases. For substances tested as CFTR activating compounds, the search function allows users to check if a specific compound or a closely related substance was already tested in the CF field. The acquired information on tested substances will assist in the identification of the most promising candidates for future therapeutics.

Explaining decisions of graph convolutional neural networks: patient-specific molecular subnetworks responsible for metastasis prediction in breast cancer.

BACKGROUND: Contemporary deep learning approaches show cutting-edge performance in a variety of complex prediction tasks. Nonetheless, the application of deep learning in healthcare remains limited since deep learning methods are often considered as non-interpretable black-box models. However, the machine learning community made recent elaborations on interpretability methods explaining data point-specific decisions of deep learning techniques. We believe that such explanations can assist the need in personalized precision medicine decisions via explaining patient-specific predictions. METHODS: Layer-wise Relevance Propagation (LRP) is a technique to explain decisions of deep learning methods. It is widely used to interpret Convolutional Neural Networks (CNNs) applied on image data. Recently, CNNs started to extend towards non-Euclidean domains like graphs. Molecular networks are commonly represented as graphs detailing interactions between molecules. Gene expression data can be assigned to the vertices of these graphs. In other words, gene expression data can be structured by utilizing molecular network information as prior knowledge. Graph-CNNs can be applied to structured gene expression data, for example, to predict metastatic events in breast cancer. Therefore, there is a need for explanations showing which part of a molecular network is relevant for predicting an event, e.g., distant metastasis in cancer, for each individual patient. RESULTS: We extended the procedure of LRP to make it available for Graph-CNN and tested its applicability on a large breast cancer dataset. We present Graph Layer-wise Relevance Propagation (GLRP) as a new method to explain the decisions made by Graph-CNNs. We demonstrate a sanity check of the developed GLRP on a hand-written digits dataset and then apply the method on gene expression data. We show that GLRP provides patient-specific molecular subnetworks that largely agree with clinical knowledge and identify common as well as novel, and potentially druggable, drivers of tumor progression. CONCLUSIONS: The developed method could be potentially highly useful on interpreting classification results in the context of different omics data and prior knowledge molecular networks on the individual patient level, as for example in precision medicine approaches or a molecular tumor board.

ndexr-an R package to interface with the network data exchange.

Motivation: Seamless exchange of biological network data enables bioinformatic algorithms to integrate networks as prior knowledge input as well as to document resulting network output. However, the interoperability between pathway databases and various methods and platforms for analysis is currently lacking. The Network Data Exchange (NDEx) is an open-source data commons that facilitates the user-centered sharing and publication of networks of many types and formats. Results: Here, we present a software package that allows users to programmatically connect to and interface with NDEx servers from within R. The network repository can be searched and networks can be retrieved and converted into igraph-compatible objects. These networks can be modified and extended within R and uploaded back to the NDEx servers. Availability and implementation: ndexr is a free and open-source R package, available via GitHub (https://github.com/frankkramer-lab/ndexr) and Bioconductor (http://bioconductor.org/packages/ndexr/). Contact: florian.auer@med.uni-goettingen.de. Supplementary information: Supplementary data are available at Bioinformatics online.

Small vessel disease, but neither amyloid load nor metabolic deficit, is dependent on age at onset in Alzheimer's disease.

BACKGROUND: There is controversy concerning whether Alzheimer's disease (AD) with early onset is distinct from AD with late onset with regard to amyloid pathology and neuronal metabolic deficit. We hypothesized that compared with patients with early-onset AD, patients with late-onset AD have more comorbid small vessel disease (SVD) contributing to clinical severity, whereas there are no differences in amyloid pathology and neuronal metabolic deficit. METHODS: The study included two groups of patients with probable AD dementia with evidence of the AD pathophysiologic process: 24 patients with age at onset <60 years old and 36 patients with age at onset >70 years old. Amyloid deposition was assessed using carbon-11-labeled Pittsburgh compound B positron emission tomography, comorbid SVD was assessed using magnetic resonance imaging, and neuronal metabolic deficit was assessed using fluorodeoxyglucose positron emission tomography. Group differences of global and regional distribution of pathology were explored using region of interest and voxel-based analyses, respectively, carefully controlling for the influence of dementia severity, apolipoprotein E genotype, and in particular SVD. The pattern of cognitive impairment was determined using z scores of the subtests of the Consortium to Establish a Registry for Alzheimer's Disease Neuropsychological Assessment Battery. RESULTS: Patients with late-onset AD showed a significantly greater amount of SVD. No statistically significant differences in global or regional amyloid deposition or neuronal metabolic deficit between the two groups were revealed. However, when not controlling for SVD, subtle differences in fluorodeoxyglucose uptake between early-onset AD and late-onset AD groups were detectable. There were no significant differences regarding cognitive functioning. CONCLUSIONS: Age at onset does not influence amyloid deposition or neuronal metabolic deficit in AD. The greater extent of SVD in late-onset AD influences the association between neuronal metabolic deficit and clinical symptoms.

Global proteome analysis of the NCI-60 cell line panel.

The NCI-60 cell line collection is a very widely used panel for the study of cellular mechanisms of cancer in general and in vitro drug action in particular. It is a model system for the tissue types and genetic diversity of human cancers and has been extensively molecularly characterized. Here, we present a quantitative proteome and kinome profile of the NCI-60 panel covering, in total, 10,350 proteins (including 375 protein kinases) and including a core cancer proteome of 5,578 proteins that were consistently quantified across all tissue types. Bioinformatic analysis revealed strong cell line clusters according to tissue type and disclosed hundreds of differentially regulated proteins representing potential biomarkers for numerous tumor properties. Integration with public transcriptome data showed considerable similarity between mRNA and protein expression. Modeling of proteome and drug-response profiles for 108 FDA-approved drugs identified known and potential protein markers for drug sensitivity and resistance. To enable community access to this unique resource, we incorporated it into a public database for comparative and integrative analysis (http://wzw.tum.de/proteomics/nci60).

Homology-based inference sets the bar high for protein function prediction.

BACKGROUND: Any method that de novo predicts protein function should do better than random. More challenging, it also ought to outperform simple homology-based inference. METHODS: Here, we describe a few methods that predict protein function exclusively through homology. Together, they set the bar or lower limit for future improvements. RESULTS AND CONCLUSIONS: During the development of these methods, we faced two surprises. Firstly, our most successful implementation for the baseline ranked very high at CAFA1. In fact, our best combination of homology-based methods fared only slightly worse than the top-of-the-line prediction method from the Jones group. Secondly, although the concept of homology-based inference is simple, this work revealed that the precise details of the implementation are crucial: not only did the methods span from top to bottom performers at CAFA, but also the reasons for these differences were unexpected. In this work, we also propose a new rigorous measure to compare predicted and experimental annotations. It puts more emphasis on the details of protein function than the other measures employed by CAFA and may best reflect the expectations of users. Clearly, the definition of proper goals remains one major objective for CAFA.

White matter hyperintensities predict amyloid increase in Alzheimer's disease.

Impaired amyloid clearance probably contributes to increased amyloid deposition in sporadic Alzheimer's disease (AD). Diminished perivascular drainage due to cerebral small-vessel disease (CSVD) has been proposed as a cause of reduced amyloid clearance. White matter hyperintensities (WMHs) are considered to reflect CSVD and can be measured using fluid-attenuated inversion recovery (FLAIR) magnetic resonance imaging (MRI). Amyloid deposition can be determined in vivo using Pittsburgh compound B ([11C]PiB) positron emission tomography (PET). We aimed to elucidate the association between WMH and the progression of amyloid deposition in patients with AD. Twenty-two patients with probable AD underwent FLAIR-MRI and [11C]PiB-PET examinations at baseline (BL) and after a mean follow-up (FU) interval of 28 months. The relationship between BL-WMH and the progression of cerebral amyloid between BL and FU was examined using a regions-of-interest (ROI) approach. The region-specific variability of this relationship was analyzed using a voxel-based method. The longitudinal analysis revealed a statistically significant association between the amount of BL-WMH and the progression of amyloid load between BL and FU (p = 0.006, adjusted R2 = 0.375, standardized coefficient ß = 0.384). The association was particularly observed in parieto-occipital regions and tended to be closer in regions adjacent to the brain surface. According to our knowledge, this is the first in vivo study in human being supporting the hypothesis that impaired amyloid clearance along perivascular drainage pathways may contribute to ß-amyloid deposition in sporadic AD. The extent of WMH might be a relevant factor to be assessed in antiamyloid drug trials.

Accuracy of robot-assisted placement of lumbar and sacral pedicle screws: a prospective randomized comparison to conventional freehand screw implantation.

STUDY DESIGN: Single-center prospective randomized controlled study. OBJECTIVE: To evaluate the accuracy of robot-assisted (RO) implantation of lumbar/sacral pedicle screws in comparison with the freehand (FH) conventional technique. SUMMARY OF BACKGROUND DATA: SpineAssist is a miniature robot for the implantation of thoracic, lumbar, and sacral pedicle screws. The system, studied in cadaver and cohort studies, revealed a high accuracy, so far. A direct comparison of the robot assistance with the FH technique is missing. METHODS: Patients requiring mono- or bisegmental lumbar or lumbosacral stabilization were randomized in a 1:1 ratio to FH or RO pedicle screw implantation. Instrumentation was performed using fluoroscopic guidance (FH) or robot assistance. The primary end point screw position was assessed by a postoperative computed tomography, and screw position was classified (A: no cortical violation; B: cortical breach <2 mm; C: ≥2 mm to <4 mm; D: ≥4 mm to <6 mm; E: ≥6 mm). Secondary end points as radiation exposure, duration of surgery/planning, and hospital stay were assessed. RESULTS: A total of 298 pedicle screws were implanted in 60 patients (FH, 152; RO, 146). Ninety-three percent had good positions (A or B) in FH, and 85% in RO. Preparation time in the operating room (OR), overall OR time, and intraoperative radiation time were not different for both groups. Surgical time for screw placement was significantly shorter for FH (84 minutes) than for RO (95 minutes). Ten RO screws required an intraoperative conversion to the FH. One FH screw needed a secondary revision. CONCLUSION: In this study, the accuracy of the conventional FH technique was superior to the RO technique. Most malpositioned screws of the RO group showed a lateral deviation. Attachment of the robot to the spine seems a vulnerable aspect potentially leading to screw malposition as well as slipping of the implantation cannula at the screw entrance point.

Tumor shrinkage assessed by volumetric MRI in the long-term follow-up after stereotactic radiotherapy of meningiomas.

PURPOSE: To evaluate tumor volume reduction in the follow-up of meningiomas after fractionated stereotactic radiotherapy (FSRT) or linac radiosurgery (RS) by using magnetic resonance imaging (MRI). PATIENTS AND METHODS: In 59 patients with skull base meningiomas, gross tumor volume (GTV) was outlined on contrast-enhanced MRI before and median 50 months (range 11-92 months) after stereotactic radiotherapy. MRI was performed as an axial three-dimensional gradient-echo T1-weighted sequence at 1.6 mm slice thickness without gap (3D-MRI). Results were compared to the reports of diagnostic findings. RESULTS: Mean tumor size of all 59 meningiomas was 13.9 ml (0.8-62.9 ml) before treatment. There was shrinkage of the treated meningiomas in all but one patient. Within a median volumetric follow-up of 50 months (11-95 months), an absolute mean volume reduction of 4 ml (0-18 ml) was seen. The mean relative size reduction compared to the volume before radiotherapy was 27% (0-73%). Shrinkage measured by 3D-MRI was greater at longer time intervals after radiotherapy. The mean size reduction was 17%, 23%, and 30% (at < 24 months, 24-48 months, and 48-72 months). CONCLUSION: By using 3D-MRI in almost all patients undergoing radiotherapy of a meningioma, tumor shrinkage is detected. The data presented here demonstrate that volumetric assessment from 3D-MRI provides additional information to routinely used radiologic response measurements. After FSRT or RS, a mean size reduction of 25-45% can be expected within 4 years.

Computer-assisted detection of pulmonary embolism: performance evaluation in consensus with experienced and inexperienced chest radiologists.

The value of a computer-aided detection tool (CAD) as second reader in combination with experienced and inexperienced radiologists for the diagnosis of acute pulmonary embolism (PE) was assessed prospectively. Computed tomographic angiography (CTA) scans (64 x 0.6 mm collimation; 61.4 mm/rot table feed) of 56 patients (31 women, 34-89 years, mean = 66 years) with suspected PE were analysed by two experienced (R1, R2) and two inexperienced (R3, R4) radiologists for the presence and distribution of emboli using a five-point confidence rating, and by CAD. Informed consent was obtained from all patients. Results were compared with an independent reference standard. Inter-observer agreement was calculated by kappa, confidence assessed by ROC analysis. A total of 1,116 emboli [within mediastinal (n = 72), lobar (n = 133), segmental (n = 465) and subsegmental arteries (n = 455)] were included. CAD detected 343 emboli (sensitivity = 30.74%, correct-positive rate = 6.13/patient; false-positive rate = 4.1/patient). Inter-observer agreement was good (R1, R2: kappa = 0.84, 95% CI = 0.81-0.87; R3, R4: kappa = 0.79, 95% CI = 0.76-0.81). Extended inter-observer agreement was higher in mediastinal and lobar than in segmental and subsegmental arteries (kappa = 0.84-0.86 and kappa = 0.51-0.58 for mediastinal/lobar and segmental/subsegmental arteries, respectively P < 0.05). Agreement between experienced and inexperienced readers was improved by CAD (kappa = 0.60-0.62 and kappa = 0.69-0.72 before and after CAD consensus, respectively P < 0.05). The experienced outperformed the inexperienced readers (Az = 0.95, 0.93, 0.89 and 0.86 for R1-4, respectively, P < 0.05). CAD significantly improved overall performances of readers 3 and 4 (Az = 0.86 for R3, R4 and Az = 0.89 for R3, R4 with CAD, P < 0.05), by enhancing sensitivities in segmental/subsegmental arteries. CAD improved experienced readers' sensitivities in segmental/subsegmental arteries (sens. = 0.93 and 0.90 for R1, R2 before and 0.97 and 0.94 for R1, R2 after CAD consensus, P < 0.05), without significant improvement of their overall performances (P > 0.05). Particularly inexperienced readers benefit from consensus with CAD data, greatly improving detection of segmental and subsegmental emboli. This system is advocated as a second reader.