Real-time execution of machine learning (ML) pipelines on radiology images is difficult due to limited computing resources in clinical environments, whereas running them in research clusters requires efficient data transfer capabilities. We developed Niffler, an open-source Digital Imaging and Communications in Medicine (DICOM) framework that enables ML and processing pipelines in research clusters by efficiently retrieving images from the hospitals’ PACS and extracting the metadata from the images. We deployed Niffler at our institution (Emory Healthcare, the largest healthcare network in the state of Georgia) and retrieved data from 715 scanners spanning 12 sites, up to 350 GB/day continuously in real-time as a DICOM data stream over the past 2 years. We also used Niffler to retrieve images bulk on-demand based on user-provided filters to facilitate several research projects. This paper presents the architecture and three such use cases of Niffler. First, we executed an IVC filter detection and segmentation pipeline on abdominal radiographs in real-time, which was able to classify 989 test images with an accuracy of 96.0%. Second, we applied the Niffler Metadata Extractor to understand the operational efficiency of individual MRI systems based on calculated metrics. We benchmarked the accuracy of the calculated exam time windows by comparing Niffler against the Clinical Data Warehouse (CDW). Niffler accurately identified the scanners’ examination timeframes and idling times, whereas CDW falsely depicted several exam overlaps due to human errors. Third, with metadata extracted from the images by Niffler, we identified scanners with misconfigured time and reconfigured five scanners. Our evaluations highlight how Niffler enables real-time ML and processing pipelines in a research cluster.
Purpose:
To compare T1 values of blood and myocardium at 1.5T and 3T before and after administration of Gd-DTPA-BMA in normal volunteers, and to evaluate the distribution of contrast media between myocardium and blood during steady state.
Materials and Methods:
Ten normal subjects were imaged with either 0.1 mmol/kg (N = 5) or 0.2 mmol/kg (N = 5) of Gd-DTPA-BMA contrast agent at 1.5T and 3T.T1 measurements of blood and myocardium were performed prior to contrast injection and every five minutes for 35 minutes following contrast injection at both field strengths. Measurements of biodistribution were calculated from the ratio of ΔR1 (ΔR1myo/ΔR1blood).
Results:
Precontrast blood T1 values (mean ± SD, N = 10) did not significantly differ between 1.5T and 3T (1.58 ± .13 sec, and 1.66 ± .06 sec, respectively; P > 0.05), but myocardium T1 values were significantly different (1.07 ± .03 sec and 1.22 ± .07 sec, respectively; P < 0.05). The field-dependent difference in myocardium T 1 postinjection (T1@3T - T1@1.5T) decreased by approximately 72% relative to precontrast T1 values, while the field-dependent difference of blood T1 decreased only 30% postcontrast. Measurements of ΔR1myo/ΔR1blood were constant for 35 minutes postcontrast, but changed between 1.5T and 3T (0.46 ± .06 vs. 0.54 ± .06, P < 0.10).
Conclusion:
T1 is significantly longer for myocardium (but not blood) at 3T compared to 1.5T. The differences in T1 due to field strength are reduced following contrast administration, which may be attributed to changes in ΔR 1myo/ΔR1blood with field strength.
Understanding system performance metrics ensures better utilization of the radiology resources with more targeted interventions. The images produced by radiology scanners typically follow the DICOM (Digital Imaging and Communications in Medicine) standard format. The DICOM images consist of textual metadata that can be used to calculate key timing parameters, such as the exact study durations and scanner utilization. However, hospital networks lack the resources and capabilities to extract the metadata from the images quickly and automatically compute the scanner utilization properties. Thus, they resort to using data records from the Radiology Information Systems (RIS). However, data acquired from RIS are prone to human errors, rendering many derived key performance metrics inadequate and inaccurate. Hence, there is motivation to establish a real-time image transfer from the Picture Archiving and Communication Systems (PACS) to receive the DICOM images from the scanners to research clusters to conduct such metadata processing to evaluate scanner utilization metrics efficiently and quickly. This paper analyzes the scanners' utilization by developing a real-time monitoring framework that retrieves radiology images into a research cluster using the DICOM networking protocol and then extracts and processes the metadata from the images. Our proposed approach facilitates a better understanding of scanner utilization across a vast healthcare network by observing properties such as study duration, the interval between the encounters, and the series count of studies. Benchmarks against using the RIS data indicate that our proposed framework based on real-time PACS data estimates the scanner utilization more accurately. Furthermore, our framework has been running stable and performing its computation for more than two years on our extensive healthcare network in pseudo real-time.
Purpose: To investigate the accuracy and reproducibility of hepatic lipid measurements using 1H magnetic resonance spectroscopy (MRS) with T2 relaxation correction, compared to measurements without correction. Materials and Methods: Experiments were conducted in phantoms of varying lipid and iron-induced susceptibility to simulate fatty liver with variable T2. Single-voxel 1H MRS was conducted with multiple TE values, and percent lipid content (lipid%) was determined at each TE to assess accuracy and TE dependency. Concurrently, T2 and equilibrium values of water and lipid were determined separately,and T2 effects on the lipid% were corrected. A similar procedure was conducted in 12 human subjects to determine susceptibility effects on water and lipid MRS signals andlipid%. Multiple measurements were used to test reproducibility. Results: The use of T2-correction was found to be more accurate than uncorrected lipid% in phantom samples (<10% error). Uncorrected lipid% error increased with increasing TE (>20% when TE > 24 msec) and with increasing susceptibility effect. In humans, while measurement repeatability was high for both corrected and uncorrected MRS, uncorrected lipid% was sensitive to acquisition TE,with 83.6% of all measurements significantly different than T2-corrected measures (P < 0.05). Conclusion: Separate T2-correction of water and lipid 1HMRS signals provides more accurate and consistent measurements of lipid%, in comparison to uncorrected estimations.
Purpose: To retrospectively determine the incidence of nephrogenic systemic fibrosis (NSF) in patients on dialysis administered either a lower dose high-relaxivity linear gadolinium-chelate, gadobenate dimeglumine (Multi-Hance, MH), compared to a standard dose linear gadolinium chelate, gadodiamide (Omniscan, OM). Materials and Methods: This study was Health Insurance Portability and Accountability Act (HIPAA)-compliant and Institutional Review Board (IRB)-approved. As per institution standardized contrast-enhanced magnetic resonance imaging (MRI) protocols, patients on dialysis were imaged using either MH, between 2/2007 to 9/2008, or OM between 10/2003 and 1/2007. Rates of NSF were compared using 95% score-based confidence intervals (CI). The Wilcoxon rank sum test was used to test similarity/ difference between contrast doses given to each patient group. Results: Overall, 312 patients on dialysis received OM and eight (2.6%) developed NSF (95% CI: 1.30%-4.98%). In all, 784 patients on dialysis received MH at a mean cumulative dose of 0.11 mmol/kg (0.05-0.75 mmol/kg) and no cases of NSF were identified (upper 95% confidence bound of 0.45%). The mean cumulative dose of OM was 0.16 mmol/kg (0.1-0.9 mmol/kg) for all patients and 0.28 mmol/kg (0.1-0.8 mmol/kg) for the patients with NSF. The median OM dose was greater in patients who developed NSF (P = 0.03), and was greater than the median MH dose (P < 0.005).
Noninvasive diagnosis of giant cell arteritis (GCA) remains challenging, particularly with regard to evaluation of extracranial arterial disease. The objective of the study was to retrospectively review extracranial involvement in patients with GCA and/or polymyalgia rheumatica (PMR), evaluated with magnetic resonance imaging (MRI), especially 3-dimensional contrast-enhanced magnetic resonance angiography images of the aortic arch and its branches. Clinical information, biopsy status, and MRI examinations of 28 patients with GCA/PMR were reviewed. Patient images were mixed randomly with 20 normal control images and were independently reviewed by 2 radiologists. Interobserver agreement for detection of arterial stenosis was determined by the k coefficient. Both readers described vascular alterations in keeping with extracranial GCA in 19 of 28 patients (67%) with good interobserver agreement (k = 0.73) and with even higher agreement on diagnosing nonocclusive versus occlusive disease (k = 1.00). The most common lesions were bilateral axillary stenosis or obstructions, observed by both readers in 8 patients (28%). Among the 19 patients with magnetic resonance angiography lesions in the subclavian/axillary arteries, 12 (75%) had biopsy-proven GCA, but only 5 (41%) of these patients had clinical features of large artery disease. In our series review, MRI could provide accurate information on involvement of the aortic arch and its branches in extracranial GCA, depicting different degrees of stenosis. Our analysis also illustrates that occult large artery vasculitis should be considered in patients without biopsy-proven GCA, patients with classic GCA but without clinical signs of large artery disease, and in patients initially diagnosed as having PMR.