Faculty Bibliography

PURPOSE/OBJECTIVE:The aim of this study was to determine rates of and possible reasons for guideline-discordant ordering of CT pulmonary angiography for the evaluation of suspected pulmonary embolism (PE) in the emergency department. METHODS:A retrospective review was performed of 212 consecutive encounters (January 6, 2016, to February 25, 2016) with 208 unique patients in the emergency department that resulted in CT pulmonary angiography orders. For each encounter, the revised Geneva score and two versions of the Wells criteria were calculated. Each encounter was then classified using a two-tiered risk stratification method (PE unlikely versus PE likely). Finally, the rate of and possible explanations for guideline-discordant ordering were assessed via in-depth chart review. RESULTS:The frequency of guideline-discordant studies ranged from 53 (25%) to 79 (37%), depending on the scoring system used; 46 (22%) of which were guideline discordant under all three scoring systems. Of these, 18 (39%) had at least one patient-specific factor associated with increased risk for PE but not included in the risk stratification scores (eg, travel, thrombophilia). CONCLUSIONS:Many of the guideline-discordant orders were placed for patients who presented with evidence-based risk factors for PE that are not included in the risk stratification scores. Therefore, guideline-discordant ordering may indicate that in the presence of these factors, the assessment of risk made by current scoring systems may not align with clinical suspicion.

OBJECTIVE:The purpose of this study is to determine the most accurate imaging techniques to measure glenoid bone loss in anterior glenohumeral instability through a systematic review of existing literature. MATERIALS AND METHODS/METHODS:We performed a comprehensive literature search of five databases for original research measuring glenoid bone loss at radiography, CT, or MRI, using prospective or retrospective cohort, case-control, or cadaveric study designs up to January 2018. The Quality Assessment of Diagnostic Accuracy Studies-2 tool aided qualitative assessment of the methods. Data extraction included results, index test interobserver agreement, and accuracy analysis. RESULTS:Twenty-seven studies (evaluating 1425 shoulders) met inclusion criteria after full-text review by two independent readers. Glenoid bone loss was assessed, comparing several index tests to nonimaging (n = 18 studies) and imaging (n = 11) reference standards. Compared with arthroscopic or cadaveric measurements, 2D CT was accurate in six of seven studies (86%), 3D CT was accurate in eight of 10 studies (80%), 2D MRI was accurate in five of seven studies (71%), 3D MRI was accurate in four of four studies (100%), and radiographs were accurate in zero of four studies (0%). Best-fit circle methods (glenoid width or Pico surface area) were the most common and both were accurate (86-90% and 75-100%, respectively) using CT and MRI. Studies had good external validity (78%). Most risk for bias arose from patient selection and reference standards. Only two studies reported sensitivity and specificity, both comparing CT to arthroscopy using different bone loss thresholds (20% and 25%). CONCLUSION/CONCLUSIONS:CT and MRI (2D or 3D) accurately measure glenoid bone loss in anterior shoulder instability, but radiographs do not. Best-fit circle measurement techniques are reliable and accurate. Current literature about glenoid bone loss is heterogeneous, and future studies should focus on diagnosis of clinically relevant glenoid bone loss.

Purpose: To explore variation in downstream costs associated with cartilage lesions incidentally detected on radiographs. Materials andMethods: The cohort was composed of 120 patients with incidental, not previously diagnosed, cartilage lesions seen on appendicular plain radiographs. The population was divided into three subgroups based on the interpreting radiologist's description: enchondroma, lowgrade cartilage lesion, and chondrosarcoma. Downstream events (follow-up imaging, office visits, biopsy, tumor resection) associated with the lesions were identified from the electronic medical record. American College of Radiology (ACR) Appropriateness Criteria were used to classify radiologists' recommendations. NationalMedicare rates were used to estimate costs of downstream events. Average cost per lesion was stratified, and cost ratios were computed among subgroups.
Result(s): Average downstream cost per lesion was $75.56. Costs were 4.6 times greater in patients under the age of 65 than over. Costs were 13.2 and 13.7 times higher when radiologists characterized lesions as chondrosarcoma versus low-grade cartilage lesion and enchondroma, respectively. There was no statistically significant difference in costs between the subgroups when accounting for size and location of lesions. Compared to when follow-up imaging was neither recommended nor obtained, costs rose from $0 to $26.03 per patient when follow-up imaging was recommended and obtained, and $62.21 per patient when followup imaging was obtained despite not being recommended. Costs rose from $0 to $14.83 per patient when radiologists' recommendations for follow-up were adherent to the ACR guidelines for management of incidental bone lesions. Costs were 2.3 times greater when ordering physicians overmanaged compared with radiologists' recommendations. No malignancy was pathologically proven in the cohort.
Conclusion(s): Costs for incidental cartilage lesions vary. Size and location of lesions do not have a significant effect on downstream costs; however, radiologists' characterization and recommendation have an impact. Therefore, it is imperative that radiologists accurately characterize such lesions and recommendations reflect the best value for patient care

Purpose: To present an automatic humeral head and glenoid segmentation method based on two-dimensional deep convolutional neural networks (CNNs).
Material(s) and Method(s): The study received institutional review board approval. A retrospective dataset of volumetric structural MR images of the shoulder from 100 subjects, including 73 normal cases and 27 cases with a Hill-Sachs lesion and/or anterior glenoid bone loss in the setting of anterior shoulder instability, were manually segmented by experts. A 2D CNN architecture was trained with multiple initial feature maps and layers. Its segmentation performance was then tested against the gold standard of manual segmentation using four-fold cross-validation. The time needed to manually segment each shoulder MRI was documented for each case.
Result(s): Automatic segmentation of the humeral head achieved a mean average precision for object detection of 0.99, a dice similarity score of 0.95, a segmentation precision of 0.95, and recall of 0.95. The Hausdorff distance was 26.9mm, the mean square distance of 0.5mm, and the residual mean square distance of 1.5mm. For the glenoid, automatic segmentation achieved a mean average precision for object detection of 0.92, a dice similarity score of 0.86, a segmentation precision of 0.88, and recall of 0.86. The Hausdorff distance was 20.7mm, themean square distance of 0.8mm, and the residual mean square distance of 1.8mm. On average, the time for manual segmentation ranged between 90 to 120 minutes per imaging study.
Conclusion(s): Using CNNs, we were able to accurately segment the humeral head and glenoid on MRI. Our results serve as an important initial step towards the automatic diagnosis and quantification of Hill-Sachs lesions and glenoid bone loss and determination of on/off track status. This, in turn, has the potential to provide consistently accurate imaging information that can be used to guide the selection of the most appropriate initial treatment for the anterior shoulder instability patient population

Purpose: Preoperative cartilage loss is a predictor of poor outcomes after arthroscopic partial meniscectomy (APM). Previous studies have used time-intensive MRI grading systems, such as MOAKS (MRI OsteoArthritis Knee Score), which are not amenable for routine clinical use. This study's purpose was to test whether cartilage loss graded using MOAKS provides better prediction of outcomes than a simpler clinically used grading system.
Material(s) and Method(s): 80 cases were selected meeting the following criteria: 1. Preoperative knee MRI performed within 6 months of APM surgery 2. Outcomes measured at the time of surgery and 1 year after surgery. Surgical failure was defined as a less than 10 point improvement in the Knee Osteoarthritis Pain Score (KOOSpain). Cases were independently evaluated by 2 musculoskeletal (MSK) radiologists and 1 radiology fellow using MOAKS and a modified Outerbridge grading system used clinically. Accuracy of each system in discriminating success and failure was estimated using area under the ROC (AUC) with 95% confidence intervals (CI).
Result(s): 78 patients (38 females) with mean age of 56.6 years (range of 45-77) were studied. 32 patients (41%) were surgical failures. At least Grade 2 (< 50% cartilage thickness loss) ranged from 8% (lateral tibial plateau) to 26% (medial femoral condyle) of the observations. AUC values of the clinical grading system (range 0.585 - 0.625) were very similar to MOAKS (range 0.553 to 0.667) for all cartilage surfaces and non-inferior to MOAKS in the medial femoral condyle (p = 0.015) and trochlea (p =0.031). The lateral femoral condyle was the only surface where MOAKS (0.667) was significantly higher than the clinical grading system (0.614).
Conclusion(s): Cartilage loss graded usingMOAKS and a simpler clinically used system have similar ability in predicting outcomes after APM. This suggests that it is feasible to use routine clinical grading of cartilage to develop models to predict outcomes after APM

Purpose: To better understand alterations in repaired supraspinatus tendons using a multimodality approach including MRI, assessment of tendon vascularity by power Doppler (PD), and tendon mechanical properties using shear wave elastography (SWE). To investigate whether SWE and PD can provide quantitative assessment of tendon healing following rotator cuff repair.
Material(s) and Method(s): This HIPAA compliant prospective study was approved by the institutional review board with informed consent. Between 9/2013 and 6/2016, twelve patients (7 males, 5 females; mean age 61 years) with unilateral full-thickness supraspinatus tendon tears underwent MRI and ultrasound pre-operatively, 3-months and 6-months post-surgery. The supraspinatus tendon MRI signal intensity, PD and SWE properties were measured. Repaired and asymptomatic shoulders were compared over time within and between modalities.
Result(s): No significant association was seen between mean SWE and MRI signal intensity (non-insertional portion -0.25, p=0.467, insertional portion -0.18, p=0.593), or between PD and MRI signal intensity (non-insertional portion -0.19, p=0.599, insertional portion 0.22, p=0.533) within the supraspinatus tendon. MRI signal intensity and PD within the supraspinatus tendon, both increased and then decreased postoperatively. Shear wave velocities increased throughout the postoperative period in the tendon footprint, while increasing and then decreasing in the distal tendon.
Conclusion(s): MRI and ultrasound parameters did not achieve statistically significant correlation; however, their respective trend behavior suggests that a temporal relationship exists between modali ties. We postul a te that a more detai l ed multiparametric imaging approach and/or comparison with a more selective MR measure, such as T2* values, may be required to evaluate rotator cuff repair

PURPOSE/OBJECTIVE:To correlate the histopathologic appearances of resected long head of the biceps tendon (LHBT) specimens following biceps tenodesis, with pre-operative MRI and arthroscopic findings, with attention to the radiologically normal biceps. MATERIAL AND METHODS/METHODS:Retrospective analysis of patients who had undergone preoperative MRI, subsequent arthroscopic subpectoral tenodesis for SLAP tears and histopathologic inspection of the excised sample between 2013 and 16. Those with a normal MRI appearance or mildly increased intrasubstance signal were independently analyzed by 2 blinded radiologists. A blinded orthopedic surgeon and pathologist reviewed all operative imaging and pathologic slides, respectively. RESULTS:Twenty-three LHBT resected samples were identified on MRI as either normal (Reader 1 n = 15; Reader 2 n = 14) or demonstrating low-grade increased signal (Reader 1 n = 8; Reader 2 n = 9). Of these, 86.9% demonstrated a histopathological abnormality. 50% of samples with histopathological abnormality demonstrated normal appearance on MRI. The most common reported histopathology finding was myxoid degeneration (73.9%) and fibrosis (52.2%). The most common arthroscopic abnormality was fraying (18.2%) and erythema (13.6%). Utilizing histopathology as the gold standard, the two radiologists demonstrated a sensitivity of 35.0% v 42.9%, specificity of 66.7% v 100%, PPV of 87.5% v 100%, and NPV of 13.3% v 14.3%. Corresponding arthroscopic inspection demonstrated a sensitivity of 31.6%, specificity of 66.6%, PPV 85.7% and NPV of 13.3%. There was moderate agreement between the two radiologists, κ = 0.534 (95% CI, 0.177 to 0.891), p = 0.01. CONCLUSION/CONCLUSIONS:Histopathological features of low grade tendinosis including mainly myxoid degeneration and fibrosis are frequently occult on MR imaging.

OBJECTIVE:To evaluate change in patient-reported outcomes following image-guided intra-articular therapeutic steroid hip injections for pain and assess correlations of outcomes with patient- and injection-specific factors. MATERIALS AND METHODS/METHODS:We retrospectively reviewed consecutive patients treated for hip pain who completed outcomes assessments from October 2011 to September 2017 at an outpatient orthopedic surgery clinic. Only patients with radiographic hip osteoarthritis (Tönnis grade ≥ 1) who underwent steroid hip injections were included. Outcomes assessments included EuroQol-5 domain (EQ5D), EQ5D-visual analog scale (VAS), and hip disability and osteoarthritis outcome score (HOOS), obtained before and within 1-6 months post-injection. Among 113 patients who completed surveys, the mean age was 59 years (±13.7 years), including 77 women (68%) and 36 men (32%). Time to repeat injection or arthroplasty was recorded. Exact Wilcoxon signed rank test assessed score differences and Spearman correlation, Kruskal-Wallis, and Mann-Whitney tests assessed correlations. RESULTS:Of 113 patients, 34 had outcomes measured at <8 weeks and 79 at ≥8 weeks. There was no significant change among any of the patients, short- or long-term follow-up subgroups in EQ5D (p = 0.450, 0.770, 0.493 respectively), EQ5D-VAS (p = 0.581, 0.915, 0.455), average-HOOS (p = 0.478, 0.696, 0.443) or total-HOOS (p = 0.380, 0.517, 0.423) scores. Forty-nine patients underwent hip arthroplasty within 1 year. Positive correlation was found between days from injection to surgery and change in EQ5D (r = 0.29, p = 0.025), average-HOOS (r = 0.33, p = 0.019), and total-HOOS (r = 0.37, p = 0.008). CONCLUSION/CONCLUSIONS:We demonstrated no significant change in patient-reported outcomes measured at short- and long-term intervals up to 6 months after therapeutic steroid hip injections.

OBJECTIVE:A major factor that impacts the long-term outcome and complication rates of total shoulder arthroplasty is the preoperative posterior glenoid bone loss quantified by glenoid retroversion. The purpose of this study was to assess if glenoid retroversion varies significantly at different glenoid heights in Walch B2 and B3 glenoids. MATERIALS AND METHODS/METHODS:Patients with B2 and B3 glenoid types were included following retrospective review of 386 consecutive CT shoulder studies performed for arthroplasty preoperative planning. True axial CT reconstructions were created using a validated technique. Two readers independently measured the glenoid retroversion angles according to the Friedman method using the "intermediate" glenoid at three glenoid heights: 75% (upper), 50% (equator), and 25% (lower). The variances between the three levels for a given patient were calculated. RESULTS:Twenty-nine B2 and 8 B3 glenoid types were included. There was no significant difference in variance of glenoid version among the three levels in B2 or B3 groups. The mean variance in retroversion degree between equator-lower, upper-equator, and upper-lower glenoid was - 0.4, 0.3, and - 0.2 for B2; and - 0.2, 1.9, and 1.9 for B3 glenoid, respectively. The level of inter-reader agreement was fair to good for variance at equator-lower, and good to excellent for upper-equator and upper-lower glenoid. CONCLUSIONS:Glenoid version can be accurately measured at any level between 25 and 75% of glenoid height for Walch B2 and B3. We recommend that the glenoid equator be used as the reference to assure consistent and reliable version measurements in this group of patients.