DICOM Imaging Router: An Open Deep Learning Framework for Classification of Body Parts from DICOM X-ray Scans

X-ray imaging in Digital Imaging and Communications in Medicine (DICOM) format is the most commonly used imaging modality in clinical practice, resulting in vast, non-normalized databases. This leads to an obstacle in deploying artificial intelligence (AI) solutions for analyzing medical images, which often requires identifying the right body part before feeding the image into a specified AI model. This challenge raises the need for an automated and efficient approach to classifying body parts from X-ray scans. Unfortunately, to the best of our knowledge, there is no open tool or framework for this task to date. To fill this lack, we introduce a DICOM Imaging Router that deploys deep convolutional neural networks (CNNs) for categorizing unknown DICOM X-ray images into five anatomical groups: abdominal, adult chest, pediatric chest, spine, and others. To this end, a large-scale X-ray dataset consisting of 16,093 images has been collected and manually classified. We then trained a set of state-of-the-art deep CNNs using a training set of 11,263 images. These networks were then evaluated on an independent test set of 2,419 images and showed superior performance in classifying the body parts. Specifically, our best performing model (i.e., MobileNet-V1) achieved a recall of 0.982 (95% CI, 0.977– 0.988), a precision of 0.985 (95% CI, 0.975–0.989) and a F1-score of 0.981 (95% CI, 0.976–0.987), whilst requiring less computation for inference (0.0295 second per image). Our external validity on 1,000 X-ray images shows the robustness of the proposed approach across hospitals. These remarkable performances indicate that deep CNNs can accurately and effectively differentiate human body parts from X-ray scans, thereby providing potential benefits for a wide range of applications in clinical settings. The dataset, codes, and trained deep learning models from this study will be made publicly available on our project website at https://vindr.ai/datasets/bodypartxr.

Sonography of Magnetically Controlled Growing Rods: A Quality Initiative in the Creation of a Multidisciplinary Clinic

Objective: The MAGEC (Magnetic Expansion Control) rods were introduced to a medical institution in 2015. The rod expansion procedures were initially evaluated with radiographs. The staff undertook a quality initiative to reduce radiation exposure by utilizing sonography. Material and Methods: The radiation dose for a typical visit was measured by examining DICOM imaging data, stored in PACS. Imaging visit time was determined from the difference between times of first radiograph/sonogram before distraction to last radiograph/sonogram after distraction. Results: The 21 patients (8 male, 13 female) were an average age of 11.4 ± 2.82 years (age at implant = 7.5 ± 1.94) when evaluated. The average length of time for a radiographic visit was 40.7 ± 20.7 minutes, whereas a sonography visit was 10.7 ± 3.7 minutes. Radiation dose per study visit prior to the introduction of the MAGEC clinic was 0.42 ± 0.39 mSv. Given an ideal patient schedule, the MAGEC clinic could reduce radiation dose by 1.3 to 2.5 mSv annually, with 95% confidence. Conclusion: This quality improvement study demonstrated a reduction in radiation exposure and imaging time. The added benefits were providing a successful multidisciplinary imaging clinic and creation of a new exam that aligned with the “ultrasound first” initiative.

Secure provision of unconsented radiological imaging for research

ABSTRACT Objectives Design and implement an architecture for managing unconsented DICOM imaging Maintain sufficient data to define research cohorts when data quality is unknown Perform project-level linkage and extraction into a Safe Haven (SH) environment Extract large image volumes for multiple projects with limited storage constraints Provide applications for an imaging research workflow within the SH environment Serve as a prototype for the Farr/NHS Scotland project to create a research dataset from Scotland’s national PACS ApproachThe software architecture builds on the Research Data Management Platform (RDMP) developed at Dundee’s Health Informatics Centre (HIC) within Farr@Dundee. The RDMP provides core services common to loading any dataset, with configuration and extensibility points for dataset-specific implementations. This architecture augments the RDMP with scalable micro-services performing peripheral functions. Images are sourced from the local PACS server in Ninewells Hospital and cached securely within HIC using an implementation for the RDMP with a custom server to query/retrieve data. Data stored in the catalogue should be anonymous, according to the Scottish SH model. The imaging dataset is poorly understood, with several potentially identifiable free-text fields which may contain information required for defining suitable research cohorts. The load process only permits verified metadata fields into the anonymised catalogue; a Mongo database stores other data for later analysis, should a field subsequently be required for cohort definition. A DICOM extraction implementation is provided, using DICOM Confidential for anonymisation and a project-specific remapping of DICOM GUIDs. Two provisioning methods have been designed. A basic copy when sufficient storage is available, and a more sophisticated method using a custom filesystem to provide separate project-specific views onto shared image files. ResultsA full end-to-end solution has been developed, from initial caching through to provisioning anonymised images. Two imaging cohorts have been loaded, one with over 5000 studies. NHS Tayside CT and MR data since 2008 is currently being loaded. Two projects have had anonymised extracts released using the ‘copy’ method. The custom filesystem method has been developed and tested with limited amounts of data. This work has highlighted anonymisation, cohort creation and SH issues which require further exploration. ConclusionA production system for securely providing linked DICOM imaging to researchers has been implemented, serving as a testbed for a national system which will provide a unique population-level resource for researchers.