Faculty Bibliography

Magnetic resonance imaging (MRI) has been proposed as a complimentary method to measure bone quality and assess fracture risk. However, manual segmentation of MR images of bone is time-consuming, limiting the use of MRI measurements in the clinical practice. The purpose of this paper is to present an automatic proximal femur segmentation method that is based on deep convolutional neural networks (CNNs). This study had institutional review board approval and written informed consent was obtained from all subjects. A dataset of volumetric structural MR images of the proximal femur from 86 subjects were manually-segmented by an expert. We performed experiments by training two different CNN architectures with multiple number of initial feature maps, layers and dilation rates, and tested their segmentation performance against the gold standard of manual segmentations using four-fold cross-validation. Automatic segmentation of the proximal femur using CNNs achieved a high dice similarity score of 0.95 ± 0.02 with precision = 0.95 ± 0.02, and recall = 0.95 ± 0.03. The high segmentation accuracy provided by CNNs has the potential to help bring the use of structural MRI measurements of bone quality into clinical practice for management of osteoporosis.

OBJECTIVES/OBJECTIVE:To determine whether use of a stylus user input device (UID) would be superior to a mouse for CMR segmentation. METHODS:Twenty-five consecutive clinical cardiac magnetic resonance (CMR) examinations were selected. Image analysis was independently performed by four observers. Manual tracing of left (LV) and right (RV) ventricular endocardial contours was performed twice in 10 randomly assigned sessions, each session using only one UID. Segmentation time and the ventricular function variables were recorded. The mean segmentation time and time reduction were calculated for each method. Intraclass correlation coefficients (ICC) and Bland-Altman plots of function variables were used to assess intra- and interobserver variability and agreement between methods. Observers completed a Likert-type questionnaire. RESULTS:The mean segmentation time (in seconds) was significantly less with the stylus compared to the mouse, averaging 206±108 versus 308±125 (p<0.001) and 225±140 versus 353±162 (p<0.001) for LV and RV segmentation, respectively. The intra- and interobserver agreement rates were excellent (ICC≥0.75) regardless of the UID. There was an excellent agreement between measurements derived from manual segmentation using different UIDs (ICC≥0.75), with few exceptions. Observers preferred the stylus. CONCLUSION/CONCLUSIONS:The study shows a significant reduction in segmentation time using the stylus, a subjective preference, and excellent agreement between the methods. KEY POINTS/CONCLUSIONS:• Using a stylus for MRI ventricular segmentation is faster compared to mouse • A stylus is easier to use and results in less fatigue • There is excellent agreement between stylus and mouse UIDs.

Rapidly enlarging, painful plexiform neurofibromas (PN) in neurofibromatosis type 1 (NF1) patients are at higher risk for harboring a malignant peripheral nerve sheath tumor (MPNST). Fludeoxyglucose (FDG) positron emission tomography/computed tomography (PET/CT) has been used to support more invasive diagnostic and therapeutic interventions. However, PET/CT imparts an untoward radiation hazard to this population with tumor suppressor gene impairment. The use of FDG PET coupled with magnetic resonance imaging (MRI) rather than CT is a safer alternative but its relative diagnostic sensitivity requires verification. Ten patients (6 females, 4 males, mean age 27 years, range 8-54) with NF1 and progressive PN were accrued from our institutional NF Clinic. Indications for PET scanning included increasing pain and/or progressive disability associated with an enlarging PN on serial MRIs. Following a clinically indicated whole-body FDG PET/CT, a contemporaneous PET/MRI was obtained using residual FDG activity with an average time interval of 3-4 h FDG-avid lesions were assessed for both maximum standardized uptake value (SUVmax) from PET/CT and SUVmax from PET/MR and correlation was made between the two parameters. 26 FDG avid lesions were detected on both PET/CT and PET/MR with an accuracy of 100%. SUVmax values ranged from 1.4-10.8 for PET/CT and from 0.2-5.9 for PET/MRI. SUVmax values from both modalities demonstrated positive correlation (r = 0.45, P < 0.001). PET/MRI radiation dose was significantly lower (53.35% ± 14.37% [P = 0.006]). In conclusion, PET/MRI is a feasible alternative to PET/CT in patients with NF1 when screening for the potential occurrence of MPNST. Reduction in radiation exposure approaches 50% compared to PET/CT.

RATIONALE AND OBJECTIVE/OBJECTIVE:The objective of this study was to evaluate breast lesion outcomes in patients after canceled MRI-guided breast biopsy due to lesion nonvisualization. MATERIALS AND METHODS/METHODS:Electronic medical records (January 2007-December 2014) were searched for patients with canceled magnetic resonance imaging (MRI)-guided breast biopsies due to lesion nonvisualization. A total of 1403 MRI-detected lesions were scheduled for MRI-guided biopsy and 89 were canceled because of nonvisualization. Imaging studies and medical records were reviewed for patient demographics, lesion characteristics, and subsequent malignancy. Patients without adequate MRI follow-up imaging were excluded. Statistical analysis was employed to determine if patient demographics or lesion characteristics were predictive of lesion resolution or lesion biopsy after subsequent follow-up. RESULTS:Eighty-nine (6.3% [89/1403]; 95% confidence interval, 5.2%-7.7%) biopsies in 89 women were canceled because of nonvisualization. Follow-up MRIs greater than 5.5 months were available for 60.7% (54/89) of women. In 74.1% (40/54) of these patients, the lesions completely resolved on follow-up. In 25.9% (14/54) of the patients, the lesion persisted on follow-up; 42.9% (6/14) of these patients underwent biopsy. One case (1.9% [1/54]) yielded ductal carcinoma in situ with microinvasion at the 6-month follow-up. No patient demographics or lesion features were associated with lesion resolution or lesion biopsy. CONCLUSIONS:The majority of canceled MRI-guided biopsy lesions resolved on later follow-up; however, because of the small possibility of a missed malignancy, follow-up MRI imaging at 6 months is recommended.

PURPOSE/OBJECTIVE:To conduct a meta-analysis of studies investigating discrepancy rates and clinical impact of imaging secondary interpretations and to identify factors influencing these rates. METHODS:EMBASE and PubMed databases were searched for original research investigations reporting discrepancy rates for secondary interpretations performed by radiologists for imaging examinations initially interpreted at other institutions. Two reviewers extracted study information and assessed study quality. Meta-analysis was performed. RESULTS:Twenty-nine studies representing a total of 12,676 imaging secondary interpretations met inclusion criteria; 19 of these studies provided data specifically for oncologic imaging examinations. Primary risks of bias included availability of initial interpretations, other clinical information, and reference standard before the secondary interpretation. The overall discrepancy rate of secondary interpretations compared with primary interpretations was 32.2%, including a 20.4% discrepancy rate for major findings. Secondary interpretations were management changing in 18.6% of cases. Among discrepant interpretations with an available reference standard, the secondary interpretation accuracy rate was 90.5%. The overall discrepancy rates by examination types were 28.3% for CT, 31.2% for MRI, 32.7% for oncologic imaging, 43.8% for body imaging, 39.9% for breast imaging, 34.0% for musculoskeletal imaging, 23.8% for neuroradiologic imaging, 35.5% for pediatric imaging, and 19.7% for trauma imaging. CONCLUSION/CONCLUSIONS:Most widely studied in the context of oncology, imaging secondary interpretations commonly result in discrepant interpretations that are management changing and more accurate than initial interpretations. Policymakers should consider these findings as they consider the value of, and payment for, secondary imaging interpretations.

RATIONALE AND OBJECTIVES/OBJECTIVE:This study aims to determine the optimal photon energy for image quality of the pulmonary arteries (PAs) on dual-energy computed tomography (CT) pulmonary angiography (CTPA) utilizing low volumes of iodinated contrast. MATERIALS AND METHODS/METHODS:The study received institutional review board exemption and was Health Insurance Portability and Accountability Act compliant. Adults (n = 56) who underwent dual-energy CTPA with 50-60 cc of iodinated contrast on a third-generation dual-source multidetector CT were retrospectively and consecutively identified. Twelve virtual monoenergetic kiloelectron volt (keV) image data sets (40-150 keV, 10-keV increments) were generated with a second-generation noise-reducing algorithm. Standard regions of interest were placed on main, right, left, and right interlobar pulmonary arteries; pectoralis muscle; and extrathoracic air. Attenuation [mean CT number (Hounsfield unit, HU)], noise [standard deviation (HU)], signal to noise (SNR), and contrast to noise ratio were evaluated. Three blinded chest radiologists rated (from 1 to 5, with 5 being the best) randomized monoenergetic and weighted-average images for attenuation and noise. P <.05 was considered significant. RESULTS:Region of interest mean CT number increased as keV decreased, with 40 keV having the highest value (P < .001). Mean SNR was highest for 40-60 keV (P <.05) (14.5-14.7) and was higher (P <.05) than all remaining energies (90-150 keV) for all vessel regions combined. Contrast to noise ratio was highest for 40 keV (P <.001) and decreased as keV increased. SNR was highest at 60 and 70 keV, only slightly higher than 40-50 keV (P <.05). Reader scores for 40-50 keV were greater than other energies and weighted-average images (P <.05). CONCLUSIONS:Kiloelectron volt images of 40-50 keV from the second-generation algorithm optimize attenuation on dual-energy CTPA and can potentially aid in interpretation and avoiding nondiagnostic examinations.

BACKGROUND:Diffusion-weighted imaging (DWI) provides insight into the pathophysiology underlying renal dysfunction. Variants of DWI include intravoxel incoherent motion (IVIM), which differentiates between microstructural diffusion and vascular or tubular flow, and diffusion tensor imaging (DTI), which quantifies diffusion directionality. PURPOSE/OBJECTIVE:To investigate the reproducibility of joint IVIM-DTI and compare controls to presurgical renal mass patients. STUDY TYPE/METHODS:Prospective cross-sectional. SUBJECTS/METHODS:Thirteen healthy controls and ten presurgical renal mass patients were scanned. Ten controls were scanned twice to investigate reproducibility. FIELD STRENGTH/SEQUENCE/UNASSIGNED:Subjects were scanned on a 3T system using 10 b-values and 20 diffusion directions for IVIM-DTI in a study approved by the local Institutional Review Board. ASSESSMENT/RESULTS:Retrospective coregistration and measurement of joint IVIM-DTI parameters were performed. STATISTICAL ANALYSIS/METHODS:Parameter reproducibility was defined as intraclass correlation coefficient (ICC) >0.7 and coefficient of variation (CV) <30%. Patient data were stratified by lesion side (contralateral/ipsilateral) for comparison with controls. Corticomedullary differentiation was evaluated. RESULTS:In controls, the reproducible subset of REnal Flow and Microstructure AnisotroPy (REFMAP) parameters had average ICC = 0.82 and CV = 7.5%. In renal mass patients, medullary fractional anisotropy (FA) was significantly lower than in controls (0.227 ± 0.072 vs. 0.291 ± 0.044, P = 0.016 for the kidney contralateral to the mass and 0.228 ± 0.070 vs. 0.291 ± 0.044, P = 0.018 for the kidney ipsilateral). In the kidney ipsilateral to the mass, cortical Dp,radial was significantly higher than in controls (P = 0.012). Conversely, medullary Dp,axial was significantly lower in contralateral than ipsilateral kidneys (P = 0.027) and normal controls (P = 0.044). DATA CONCLUSION/UNASSIGNED:REFMAP-MRI parameters provide unique information regarding renal dysfunction. In presurgical renal mass patients, directional flow changes were noted that were not identified with IVIM analysis alone. Both contralateral and ipsilateral kidneys in patients show reductions in structural diffusivities and anisotropy, while flow metrics showed opposing changes in contralateral vs. ipsilateral kidneys. LEVEL OF EVIDENCE/METHODS:2 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2018.

BACKGROUND: Multiparametric magnetic resonance imaging (mpMRI) combining different techniques such as MR elastography (MRE) has emerged as a noninvasive approach to diagnose and stage liver fibrosis with high accuracy allowing for anatomical and functional information. PURPOSE: To assess the diagnostic performance of mpMRI including qualitative and quantitative assessment of MRE, liver surface nodularity (LSN) measurement, hepatic enhancement ratios postgadoxetic acid, and serum markers (APRI, FIB-4) for the detection of liver fibrosis. STUDY TYPE: IRB-approved retrospective. SUBJECTS: Eighty-three adult patients. FIELD STRENGTH/SEQUENCE: 1.5T and 3.0T MR systems. MRE and T1 -weighted postgadoxetic acid sequences. ASSESSMENT: Two independent observers analyzed qualitative color-coded MRE maps on a scale of 0-3. Regions of interest were drawn to measure liver stiffness on MRE stiffness maps and on pre- and postcontrast T1 -weighted images to measure hepatic enhancement ratios. Software was used to generate LSN measurements. Histopathology was used as the reference standard for diagnosis of liver fibrosis in all patients. STATISTICAL TESTS: A multivariable logistic analysis was performed to identify independent predictors of liver fibrosis. Receiver operating characteristic (ROC) analysis evaluated the performance of each imaging technique for detection of fibrosis, in comparison with serum markers. RESULTS: Liver stiffness measured with MRE provided the strongest correlation with histopathologic fibrosis stage (r = 0.74, P < 0.001), and the highest diagnostic performance for detection of stages F2-F4, F3-F4, and F4 (areas under the curve [AUCs] of 0.87, 0.91, and 0.89, respectively, P < 0.001) compared to other methods. Qualitative assessment of MRE maps showed fair to good accuracy for detection of fibrosis (AUC range 0.76-0.84). Multivariable logistic analysis identified liver stiffness and FIB-4 as independent predictors of fibrosis with AUCs of 0.90 (F2-F4), 0.93 (F3-F4) and 0.92 (F4) when combined. DATA CONCLUSION: Liver stiffness measured with MRE showed the best performance for detection of liver fibrosis compared to LSN and gadoxetic acid uptake, with slight improvement when combined with FIB-4. LEVEL OF EVIDENCE: 3 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2017.

Fluorodeoxyglucose (FDG) positron emission tomography-magnetic resonance (PET/MR) is useful for the evaluation of cognitively-impaired patients. This study aims to assess two different attenuation correction (AC) methods (Dixon-MR and atlas-based) versus index-standard computed tomography (CT) AC for the visual interpretation of regional hypometabolism in patients with cognitive impairment. Two board-certified nuclear medicine physicians blindly scored brain region FDG hypometabolism as normal versus hypometabolic using two-dimensional (2D) and 3D FDG PET/MR images generated by MIM software. Regions were quantitatively assessed as normal versus mildly, moderately, or severely hypometabolic. Hypometabolism scores obtained using the different methods of AC were compared, and interreader, as well as intra-reader agreement, was assessed. Regional hypometabolism versus normal metabolism was correctly classified in 16 patients on atlas-based and Dixon-based AC map PET reconstructions (vs. CT reference AC) for 94% (90%-96% confidence interval [CI]) and 93% (89%-96% CI) of scored regions, respectively. The averaged sensitivity/specificity for detection of any regional hypometabolism was 95%/94% (P = 0.669) and 90%/91% (P = 0.937) for atlas-based and Dixon-based AC maps. Interreader agreement for detection of regional hypometabolism was high, with similar outcome assessments when using atlas- and Dixon-corrected PET data in 93% (Κ =0.82) and 93% (Κ =0.84) of regions, respectively. Intrareader agreement for detection of regional hypometabolism was high, with concordant outcome assessments when using atlas- and Dixon-corrected data in 93%/92% (Κ =0.79) and 92/93% (Κ =0.78). Despite the quantitative advantages of atlas-based AC in brain PET/MR, routine clinical Dixon AC yields comparable visual ratings of regional hypometabolism in the evaluation of cognitively impaired patients undergoing brain PET/MR and is similar in performance to CT-based AC. Therefore, Dixon AC is acceptable for the routine clinical evaluation of dementia syndromes.