Convolutional neural network-based kidney volume estimation from low-dose unenhanced computed tomography scans

PURPOSE

Kidney volume is important in the management of renal diseases. Unfortunately, the currently available, semi-automated kidney volume determination is time-consuming and prone to errors. Recent advances in its automation are promising but mostly require contrast-enhanced computed tomography (CT) scans. This study aimed at establishing an automated estimation of kidney volume in non-contrast, low-dose CT scans of patients with suspected urolithiasis.

METHODS

The kidney segmentation process was automated with 2D Convolutional Neural Network (CNN) models trained on manually segmented 2D transverse images extracted from low-dose, unenhanced CT scans of 210 patients. The models' segmentation accuracy was assessed using Dice Similarity Coefficient (DSC), for the overlap with manually-generated masks on a set of images not used in the training. Next, the models were applied to 22 previously unseen cases to segment kidney regions. The volume of each kidney was calculated from the product of voxel number and their volume in each segmented mask. Kidney volume results were then validated against results semi-automatically obtained by radiologists.

RESULTS

The CNN-enabled kidney volume estimation took a mean of 32 s for both kidneys in a CT scan with an average of 1026 slices. The DSC was 0.91 and 0.86 and for left and right kidneys, respectively. Inter-rater variability had consistencies of ICC = 0.89 (right), 0.92 (left), and absolute agreements of ICC = 0.89 (right), 0.93 (left) between the CNN-enabled and semi-automated volume estimations.

CONCLUSION

In our work, we demonstrated that CNN-enabled kidney volume estimation is feasible and highly reproducible in low-dose, non-enhanced CT scans. Automatic segmentation can thereby quantitatively enhance radiological reports.

Comparison and Incorporation of Reasoning and Learning Approaches for Cancer Therapy Research

Representing knowledge in a comprehensible and maintainable way and transparently providing inferences thereof are important issues, especially in the context of applications related to artificial intelligence in medicine. This becomes even more obvious if the knowledge is dynamically growing and changing and when machine learning techniques are being involved. In this paper, we present an approach for representing knowledge about cancer therapies collected over two decades at St.-Johannes-Hospital in Dortmund, Germany. The presented approach makes use of InteKRator, a toolbox that combines knowledge representation and machine learning techniques, including the possibility of explaining inferences. An extended use of InteKRator's reasoning system will be introduced for being able to provide the required inferences. The presented approach is general enough to be transferred to other data, as well as to other domains. The approach will be evaluated, e. g., regarding comprehensibility, accuracy and reasoning efficiency.

cbpManager: a web application to streamline the integration of clinical and genomic data in cBioPortal to support the Molecular Tumor Board

Background

Extensive sequencing of tumor tissues has greatly improved our understanding of cancer biology over the past years. The integration of genomic and clinical data is increasingly used to select personalized therapies in dedicated tumor boards (Molecular Tumor Boards) or to identify patients for basket studies. Genomic alterations and clinical information can be stored, integrated and visualized in the open-access resource cBioPortal for Cancer Genomics. cBioPortal can be run as a local instance enabling storage and analysis of patient data in single institutions, in the respect of data privacy. However, uploading clinical input data and genetic aberrations requires the elaboration of multiple data files and specific data formats, which makes it difficult to integrate this system into clinical practice. To solve this problem, we developed cbpManager.

Results

cbpManager is an R package providing a web-based interactive graphical user interface intended to facilitate the maintenance of mutations data and clinical data, including patient and sample information, as well as timeline data. cbpManager enables a large spectrum of researchers and physicians, regardless of their informatics skills to intuitively create data files ready for upload in cBioPortal for Cancer Genomics on a daily basis or in batch. Due to its modular structure based on R Shiny, further data formats such as copy number and fusion data can be covered in future versions. Further, we provide cbpManager as a containerized solution, enabling a straightforward large-scale deployment in clinical systems and secure access in combination with ShinyProxy. cbpManager is freely available via the Bioconductor project at https://bioconductor.org/packages/cbpManager/ under the AGPL-3 license. It is already used at six University Hospitals in Germany (Mainz, Gießen, Lübeck, Halle, Freiburg, and Marburg).

Conclusion

In summary, our package cbpManager is currently a unique software solution in the workflow with cBioPortal for Cancer Genomics, to assist the user in the interactive generation and management of study files suited for the later upload in cBioPortal.

[ADAPTHERA-Statewide cross-sectoral care network for patients with early rheumatoid arthritis shows sustained remission in standard care]

BACKGROUND/OBJECTIVE

The majority of patients in Germany miss out on the necessity of early diagnosis and initiation of therapy for rheumatoid arthritis (RA) caused by considerable structural deficits in the health care system. The challenge is to reconcile the individual demand for the best possible therapy result with a sustainable expenditure of resources.

METHODS

The cross-sectoral regional care network ADAPTHERA aims to improve early RA diagnosis and treatment in Rhineland-Palatinate. The retrospective triage analyses of suspected early onset RA patients was performed by tracing the selection process of all available enquiries (n = 1045). For analysis of the clinical course of the disease, a subset comprising 143 patients with a minimum observation time of 12 months (5 consecutive visits) was available. Clinical and laboratory parameters were collected quarter yearly, self-administered questionnaires were filled out and the treatment was adapted if necessary.

RESULTS

A total of 454 patients were included. The mean waiting time was 23.9 (SD = 18) days. The mean observation period in the subcohort was 29.2 (SD = 12.7) months, with about 50% of the patients presenting within 3 months. Almost 75% of the patients were in remission after 2 years. A sustained remission could be described for 74.8% (6 months) and 53.5% (12 months), respectively. Especially patients with rapid remission induction benefited in terms of longer remissions (p = 0.03). A very early stage of the disease (VERA) was associated with a rarely necessary biologic therapy (p = 0.022).

DISCUSSION

The approach of a supply network is not a panacea, but it might improve healthcare for patients with early onset RA. In order to minimize resource utilization, a pinpoint referral and accurate triage of potential cases are crucial.

[The regional network ADAPTHERA : Rheumatology care through coordinated cooperation: comprehensive, trans-sectoral, covering all health insurance. Initial results]

The aim of the rheumatology network ADAPTHERA ("risk-adapted rheumatology therapy") is to achieve a comprehensive improvement in rheumatology care by coordinating treatment in a regional, trans-sectoral network. Accompanying biomedical research projects, training concepts, and the construction of a rheumatology register (gathering data and biomaterials) should furthermore ensure the stable and sustainable optimisation of care. In the pilot phase (2012-2015) the focus of the ADAPTHERA network, required as a "regional key project" within the framework of the Initiative on Health Economy of Rheinland-Palatinate (RL-P), Germany, was placed on the optimisation of the early diagnosis of rheumatoid arthritis, where it is well-known that there is a significant care deficit.Through the intensive, stable, and coordinated cooperation of all health care partners in the field of rheumatology (registered general practitioners and orthopaedic specialists, registered core rheumatologists as well as the Association of Rheumatology of RL-P) a unique regional, comprehensive offer with verifiable care optimisation has been established in RL-P. The network is supported by outstanding collaboration with the Association of Statutory Health Insurance Physicians and the self-help organisation Rheumatology League.The aims that were established at the start of the project will be achieved by the end of the pilot phase:- significant improvement in the early diagnosis of rheumatoid arthritis (an average of 23.7 days until diagnosis by rheumatologists)- access covering all health insurance (regardless of the particular scheme the patients belong to)- comprehensive (verifiable participation of general practitioners from all over RL-P)- data and biomaterials collection, established as a basis for biomarker research, and a rheumatology register for RL-P.

TeamTreat--a communication platform for concerted cancer treatment

TeamTreat is an internet platform providing a case record for cancer patients across and inside the primary and secondary health care sector. Due to the slow progress of cross-institutional integration in healthcare IT, we created an alternative low-level approach to this problem and put special emphasis on an easy access for healthcare professionals regardless of the specific IT infrastructure they use. Physicians use the platform to share and collect information to achieve a collaborative treatment of cancer. Furthermore, the data in the case record is searchable for clinical researchers to find suitable patients for inclusion in clinical trials. Reading access for patients is also possible.