Artificial intelligence improves resident detection of pediatric and young adult upper extremity fractures.

PURPOSE: We wished to evaluate if an open-source artificial intelligence (AI) algorithm ( https://www.childfx.com ) could improve performance of (1) subspecialized musculoskeletal radiologists, (2) radiology residents, and (3) pediatric residents in detecting pediatric and young adult upper extremity fractures. MATERIALS AND METHODS: A set of evaluation radiographs drawn from throughout the upper extremity (elbow, hand/finger, humerus/shoulder/clavicle, wrist/forearm, and clavicle) from 240 unique patients at a single hospital was constructed (mean age 11.3 years, range 0-22 years, 37.9% female). Two fellowship-trained musculoskeletal radiologists, three radiology residents, and two pediatric residents were recruited as readers. Each reader interpreted each case initially without and then subsequently 3-4 weeks later with AI assistance and recorded if/where fracture was present. RESULTS: Access to AI significantly improved area under the receiver operator curve (AUC) of radiology residents (0.768 [0.730-0.806] without AI to 0.876 [0.845-0.908] with AI, P < 0.001) and pediatric residents (0.706 [0.659-0.753] without AI to 0.844 [0.805-0.883] with AI, P < 0.001) in identifying fracture, respectively. There was no evidence of improvement for subspecialized musculoskeletal radiology attendings in identifying fracture (AUC 0.867 [0.832-0.902] to 0.890 [0.856-0.924], P = 0.093). There was no evidence of difference between overall resident AUC with AI and subspecialist AUC without AI (resident with AI 0.863, attending without AI AUC 0.867, P = 0.856). Overall physician radiograph interpretation time was significantly lower with AI (38.9 s with AI vs. 52.1 s without AI, P = 0.030). CONCLUSION: An openly accessible AI model significantly improved radiology and pediatric resident accuracy in detecting pediatric upper extremity fractures.

Factors influencing patient understanding of information on radiology examinations.

PURPOSE: To determine which factors influence patient understanding of information documents on radiology examinations. MATERIALS AND METHODS: This is a randomized prospective study with 361 consecutive patients. Documents with information on 9 radiology exams were obtained ( www.radiologyinfo.org ). Three versions of each of these were written at low (below 7th grade), middle (8-12th grade), and high (college) reading grades. Before their scheduled radiology exam, patients were randomized to read one document. Their subjective and objective understanding of the information was assessed. Statistics including logistic regression used to assess relationships between demographic factors and document grade level and understanding. RESULTS: Twenty-eight percent (100/361) of patients completed the study. More females vs. males (85% vs. 66%) read their entire document (p = 0.042). Document grade level was not associated with understanding (p > 0.05). Correlation between college degrees and subjective understanding was positive (r = 0.234, p = 0.019). More females (74% vs. 54%, p = 0.047) and patients with college degrees (72% vs. 48%, p = 0.034) had higher objective understanding. Controlling for document grade level and demographics, patients with college degrees were more likely to have subjective understanding of at least half of the document (OR 7.97, 95% CI [1.24, 51.34], p = 0.029) and females were more likely to have higher objective understanding (OR 2.65, 95% CI [1.06, 6.62], p = 0.037). CONCLUSION: Patients with college degrees understood more on information documents. Females read more of the documents than males and had a higher objective understanding. Reading grade level did not affect understanding.

Assessing the readability of patient-targeted online information on musculoskeletal radiology procedures.

OBJECTIVE: To assess the readability of patient-targeted online information on musculoskeletal radiology procedures. METHODS: Eleven common musculoskeletal radiology procedures were queried in three online search engines (Google, Yahoo!, Bing). All unique patient-targeted websites were identified (n = 384) from the first three pages of search results. The reading grade level of each website was calculated using 6 separate validated metrics for readability assessment. Analysis of word and sentence complexity was also performed. Results were compared between academic vs. non-academic websites and between websites found on different pages of the search results. Statistics were performed using a t test. RESULTS: The mean reading grade level across all procedures was 10th-14th grade. Webpages for nerve block were written at a higher reading grade level on non-academic websites (p = 0.025). There was no difference in reading grade levels between academic and non-academic sources for all other procedures. There was no difference in reading grade levels between websites found on the first page of search results compared with the second and third pages. Across all websites, 16-22% of the words used had 3+ syllables and 31-43% of the words used had 6+ characters (complex words); 13-24% of the sentences used had 22+ words (complex sentences). CONCLUSION: Patient-targeted online information on musculoskeletal radiology procedures are written at the 10th-14th grade reading level, which is well beyond the AMA and NIH recommendation. Readability can be lowered by decreasing text complexity through limitation of high-syllable words and reduction in word and sentence length.

MRI following medial patellofemoral ligament reconstruction: assessment of imaging features found with post-operative pain, arthritis, and graft failure.

OBJECTIVE: To assess MR features following MPFL reconstruction and determine their influence on post-operative pain, progressive arthritis, or graft failure. MATERIALS AND METHODS: Retrospective study on 38 patients with MPFL reconstruction and a post-operative MRI between January 2010 and June 2019. Two radiologists assessed MPFL graft signal, graft thickness, femoral screw, femoral tunnel widening, and patellofemoral cartilage damage. The third performed patellofemoral instability measurements. All three assessed femoral tunnel position with final result determined by majority consensus. Imaging findings were evaluated in the setting of post-operative pain, patellofemoral arthritis, and MPFL graft failure including need for MPFL revision. Statistics included chi-square, Fisher's exact test, t test, and kappa. RESULTS: Mean graft thickness was 6.0 ± 1.8 mm; 24% of the grafts were diffusely hypointense. Mean femoral tunnel widening was 2.5 ± 1.8 mm; 34% of the femoral screws were broken or extruded. Fifty-two percent of the patients had no interval cartilage change. Non-anatomic femoral tunnels were found in 66% of patients, including in all 9 patients requiring revision MPFL reconstruction (p = 0.013). Revised MPFL grafts had more abnormal femoral screws compared to those that did not (67% vs. 24%) (p = 0.019). Other MR features did not significantly influence the evaluated outcomes. CONCLUSION: The need for revision MPFL reconstruction occurs more frequently when there is a non-anatomic femoral tunnel and broken or extruded femoral screws. The appearance of the MPFL graft itself is not an influencing factor for post-operative pain, progression of patellofemoral arthritis, or graft failure.

The prevalence of athletic pubalgia imaging findings on MRI in patients with femoroacetabular impingement.

OBJECTIVE: To determine the prevalence of athletic pubalgia imaging findings on MRI in patients with femoroacetabular impingement and assess for correlative risk factors. MATERIALS AND METHODS: A retrospective search identified 156 hips with femoroacetabular impingement and a control group of 113 without femoroacetabular impingement that had an MRI performed between January 1, 2015, and January 1, 2018. Two fellowship-trained musculoskeletal radiologists reviewed studies for the presence of acute osteitis pubis, chronic osteitis pubis, adductor tendinosis, and tendon tear; rectus abdominis tendinosis and tendon tear; and aponeurotic plate tear. Findings were correlated with various clinical and imaging risk factors. Univariate and multivariate statistical analyses were performed. RESULTS: Imaging findings of adductor tendinosis (p = 0.02) and chronic osteitis pubis (p = 0.01) were more prevalent in FAI patients than controls. Univariate analyses in FAI patients showed that an alpha angle ≥ 60° had a higher prevalence of aponeurotic plate tears (p = 0.02) and adductor tendinosis (p = 0.049). Multivariate analyses showed that an alpha angle ≥ 60° had a higher prevalence of chronic osteitis pubis (OR = 2.27, p = 0.031), sports participation had a higher prevalence of adductor tendon tears (OR = 4.69, p = 0.013) and chronic osteitis pubis (OR = 2.61, p = 0.0058), and males had a higher prevalence of acute osteitis pubis (OR = 5.17, p = 0.032). CONCLUSION: Sports participation, alpha angle ≥ 60°, and male sex predict a higher prevalence of athletic pubalgia imaging findings in patients with femoroacetabular impingement.

MR Enterographic Findings as Biomarkers of Mucosal Healing in Young Patients With Crohn Disease.

OBJECTIVE: The purpose of this study was to investigate the MR enterographic findings that best correlate with mucosal healing assessed with ileocolonoscopy. MATERIALS AND METHODS: Patients with Crohn disease who underwent two ileocolonoscopic examinations and also underwent MR enterography close in time to the second endoscopic examination were included in a retrospective study. Two pediatric gastroenterologists blinded to the imaging findings reviewed the endoscopic examinations to assess for mucosal healing, defined as resolution of inflammation within a bowel segment at subsequent ileocolonoscopy. Two radiologists blinded to endoscopic and clinical data interpreted the MR enterographic images. Sensitivity, specificity, and accuracy for mucosal healing were calculated for several imaging features. RESULTS: A total of 30 patients (15 female patients, 15 male patients; age range, 8-24 years; mean, 17.2 ± 3.2 years) with pediatric-onset Crohn disease were examined. The average time between MR enterography and the second ileocolonoscopic examination was 12.7 ± 7.9 days. A total of 202 bowel segments from the terminal ileum to rectum were evaluated in the 60 ileocolonoscopic examinations. Forty-four bowel segments exhibited mucosal healing, and 37 segments exhibited persistent inflammation. At imaging, the MR index of activity score in mucosal healing segments was 6.6 ± 3.4, compared with 13.7 ± 9.7 in segments without mucosal healing (p = 0.0001). The average bowel wall thickness in healing segments was 2.7 ± 0.9 mm compared with 4.7 ± 3.1 mm in persistently inflamed segments (p = 0.0004). An MR index of activity score less than 8 had the highest accuracy for mucosal healing (accuracy, 74%; sensitivity, 84%; specificity, 62%; p < 0.0001). Mucosal hyperenhancement (72%, 98%, 41%), mesenteric hypervascularity (72%, 98%, 41%), bowel wall edema (72%, 93%, 46%), and bowel wall thickness less than 4 mm (72%, 84%, 57%) were also strongly associated with mucosal healing (p < 0.0003). CONCLUSION: In this study MR enterography was accurate for assessing mucosal healing, an important therapeutic endpoint in pediatric patients with Crohn disease.

The Role of MR Enterography in Assessing Crohn's Disease Activity and Treatment Response.

MR enterography (MRE) has become the primary imaging modality in the assessment of Crohn's disease (CD) in both children and adults at many institutions in the United States and worldwide, primarily due to its noninvasiveness, superior soft tissue contrast, and lack of ionizing radiation. MRE technique includes distention of the small bowel with oral contrast media with the acquisition of T2-weighted, balanced steady-state free precession, and multiphase T1-weighted fat suppressed gadolinium contrast-enhanced sequences. With the introduction of molecule-targeted biologic agents into the clinical setting for CD and their potential to reverse the inflammatory process, MRE is increasingly utilized to evaluate disease activity and response to therapy as an imaging complement to clinical indices or optical endoscopy. New and emerging MRE techniques, such as diffusion-weighted imaging (DWI), magnetization transfer, ultrasmall superparamagnetic iron oxide- (USPIO-) enhanced MRI, and PET-MR, offer the potential for an expanded role of MRI in detecting occult disease activity, evaluating early treatment response/resistance, and differentiating inflammatory from fibrotic strictures. Familiarity with MR enterography is essential for radiologists and gastroenterologists as the technique evolves and is further incorporated into the clinical management of CD.

A new, simple method for estimating pleural effusion size on CT scans.

BACKGROUND: There is no standardized system to grade pleural effusion size on CT scans. A validated, systematic grading system would improve communication of findings and may help determine the need for imaging guidance for thoracentesis. METHODS: CT scans of 34 patients demonstrating a wide range of pleural effusion sizes were measured with a volume segmentation tool and reviewed for qualitative and simple quantitative features related to size. A classification rule was developed using the features that best predicted size and distinguished among small, moderate, and large effusions. Inter-reader agreement for effusion size was assessed on the CT scans for three groups of physicians (radiology residents, pulmonologists, and cardiothoracic radiologists) before and after implementation of the classification rule. RESULTS: The CT imaging features found to best classify effusions as small, moderate, or large were anteroposterior (AP) quartile and maximum AP depth measured at the midclavicular line. According to the decision rule, first AP-quartile effusions are small, second AP-quartile effusions are moderate, and third or fourth AP-quartile effusions are large. In borderline cases, AP depth is measured with 3-cm and 10-cm thresholds for the upper limit of small and moderate, respectively. Use of the rule improved interobserver agreement from κ = 0.56 to 0.79 for all physicians, 0.59 to 0.73 for radiology residents, 0.54 to 0.76 for pulmonologists, and 0.74 to 0.85 for cardiothoracic radiologists. CONCLUSIONS: A simple, two-step decision rule for sizing pleural effusions on CT scans improves interobserver agreement from moderate to substantial levels.