Evaluating changes in diagnostic accuracy of ultrasound for appendicitis: does practice make perfect?
Austin-Page LR, Pham PK, and Elkhunovich M. Evaluating changes in diagnostic accuracy of ultrasound for appendicitis: does practice make perfect? Article in Press.
J Emerg Med 2020; 1–10
Question(s) What are the changes in diagnostic accuracy of ultrasound (US) for appendicitis over time after adopting an US-first protocol?
Design Retrospective review from 2009 to 2014
Setting Single center, urban, free-standing children’s hospital with Emergency Department (ED) volume >70,000 patient visits per year; University of California-San Diego/Rady Children’s Hospital
Participants 1,058 US exams were included and reviewed
Intervention US as initial imaging modality in children presenting to the ED with abdominal pain
Outcomes Primary outcomes were the accuracy rates of US over time based on either a conclusive report (fully visualized normal appendix or unequivocal appendicitis) or an equivocal report (non-visualized or partially visualized appendix without mention of appendicitis in the report). Accuracy rates were also analyzed in relation to the patient’s sex, body mass index (BMI), and the ultrasonographer’s experience. Secondary outcomes were sensitivity and specificity of US for appendicitis.
Main Results Over 5 years, the diagnostic accuracy of US for appendicitis improved significantly from 13.9% to 31.5% (P=0.001) with an overall accuracy of 24.5%. The overall sensitivity and specificity were 80.7% and 77.6%, respectively. Male children were significantly associated with conclusive reports (P<0.001), and higher BMI was significantly associated with equivocal reports (P<0.001). Ultrasonographer experience was not associated with conclusive reports (P=0.22).
Commentary US is the first-line imaging modality in many institutions for evaluation of appendicitis in children; however, many institutions still depend on CT as the initial modality. This study shows that US accuracy improves over time, which may encourage additional institutions to adopt US as the initial modality. Unfortunately, this study classifies a non-visualized or partially visualized appendix as equivocal. Most research defines a non-visualized appendix as normal. This difference in categorization significantly improves the reported accuracy of ultrasound in diagnosing appendicitis.
Abdominal lymph node size in children at computed tomography
Spijkers S, Staats J.M., Littooij A.S. et al. Abdominal lymph node size in children at computed tomography.
Pediatr Radiol 2020; 50, 1263–1270.
Question(s) What is the normal size of abdominal lymph nodes in children on CT?
Design Retrospective study over from 2012 to 2014
Setting Single center (University Medical Center Utrecht/Wilhelmina Children’s Hospital, Utrecht, Netherlands)
Participants 152 otherwise healthy children aged 1-17 years who underwent contrast enhanced CT after high energy trauma
Outcomes Lymph node measurements per age and station including maximum size, mean values, and upper limits of normal reference range. Adult RECIST criteria were used as a reference.
Main results For each of the 152 patients, at least 3 lymph node stations were measured in the axial and coronal planes for a total of 647 lymph nodes. The most commonly identified nodes were inguinal (100%), mesenteric (99%) and iliac (98%). The upper limits (determined by statistical bootstrapping) of the short axial axis of all lymph node stations were similar to the current adult guidelines (<10 mm) and ranged from 6.4-10 mm. The short coronal axis exceeded the reference interval for inguinal (12.4 mm) and mesenteric stations (11.2 mm). Using adult RECIST criteria, 19% of children would have had one or more enlarged lymph nodes (predominantly inguinal). There were positive statistically significant correlations between age and lymph node size for all stations (0.21-0.5, P<0.05).
Commentary This study provides normative data for abdominal and pelvic lymph node sizes in children. The authors provide values for both short and long axial and coronal measurements. Generally, the coronal long axis had the widest CI and are not applicable to the RECIST criteria. Thus, the short axis axial ranges represent a better reference option.
Accurately distinguishing pediatric ileocolic intussusception from small-bowel intussusception using ultrasonography
Zhang M, Zhou X, Hu Q, et al. Accurately distinguishing pediatric ileocolic intussusception from small-bowel intussusception using ultrasonography.
J Pediatr Surg 2020.
Question(s) Can ultrasound distinguish between ileocolic and small bowel intussusception?
Design Retrospective study from 2018 to 2020
Setting Single center (Changsha Central Hospital, Hunan, China)
Participants 183 patients with intussusception, 123 cases of small bowel intussusception and 60 cases of ileocolic intussusception
Outcomes Comparison ultrasound measurements including lesion diameter, fat core thickness, outer wall thickness, length of intussusception, the presence of lymph nodes or a lead point and the appearance of the ileocecal region.
Main result For ileocolic intussusception, the mean diameter was 28.9 mm, fat core thickness was 11.4 mm, outer wall thickness was 6.5 mm, lesion length was 53.4 mm, and 56/60 had visible lymph nodes. There were no cases with a normal appearing ileocecal region and only 5/60 (8%) had a normal appearing ascending colon. For small bowel intussusception, the mean diameter was 15.8 mm, fat core thickness was 2.5 mm, outer wall thickness was 3.8 mm, lesion length was 27.6 mm, and only 1 case had lymph nodes visible. There was a normal appearing ileocecal region in 104/123 cases (85%) and a normal ascending colon in 120/123 cases (98%). These differences were all statistically significant with P<0.001.
Commentary Ileocolic and small bowel intussusception are the most common types of intussusception in children but treatment is different. Ileocolic intussusception can lead to ischemia, necrosis, and perforation and therefore must be promptly reduced. Most small bowel intussusceptions are transient and can be managed conservatively. This article shows that ultrasound can distinguish between small bowel and ileocolic intussusception based on the examined parameters. A normal US appearance of the ileocecal region and ascending colon may be the best finding to differentiate ileocolic and small bowel intussusception.
Coronavirus disease 2019 (COVID-19) in children: a systematic review of imaging findings.
Shelmerdine SC, Lovrenski J, Caro-Dominguez P et al. Coronavirus disease 2019 (COVID-19) in children: a systematic review of imaging findings. Published online: 18 June 2020.
Question(s) What are the imaging findings in pediatric cases of COVID-19?
Design Systematic review of 4 databases (Medline, Embase, Cochrane, Google Scholar) over 5 years
Setting China, South Korea, and Iran
Articles 22 articles included (chest imaging findings in 431 children)
Intervention Chest imaging (chest CT with or without IV contrast, chest radiographs)
Outcomes Initial imaging findings, follow up imaging findings (3-15 days later)
Main Results 34% of patients who underwent chest CT had normal findings initially. Abnormalities were most commonly seen in the lower lobes and unilaterally. The most common imaging pattern was ground-glass opacities (62.4%). Pleural effusions were rare and no cases showed lymphadenopathy. On follow up imaging, 29% showed improvement, 25% remained normal and 9% showed progression.
Commentary This is a timely study that illustrates the common chest imaging findings encountered in pediatric patients with COVID-19. The authors acknowledge several limitations, including gaps in evaluation of immunocompromised children and the fact that the cohort was primarily from China. Although COVID-19 has been known for much less than 5 years, the authors performed a literature search over a span of 5 years, presumably for completeness. Their diligence is also noted with 2 additional literature searches after the initial search and inclusion of grey papers, as well. This manuscript does not address the extra-thoracic imaging findings of COVID-19.
E-cigarette or vaping product use-associated lung injury in the pediatric population: imaging features at presentation and short-term follow-up
Wang KY, Jadhav SP, Yenduri NJS et al. E-cigarette or vaping product use-associated lung injury in the pediatric population: imaging features at presentation and short-term follow-up.
Pediatr Radiol 2020; 50, 1231–1239.
Question(s) What are the most common imaging findings of e-cigarette or vaping product use-associated lung injury (EVALI) on CT?
Design Retrospective review
Setting Single center (Baylor College of Medicine/Texas Children’s Hospital, Houston, TX)
Participants 11 teenagers with confirmed or probable EVALI
Outcomes Normal or abnormal chest X-rays; CT imaging findings, specifically peri-bronchovascular sparing, subpleural sparing, lobular sparing, gravitationally dependent gradient of opacities, consolidation, ground-glass opacities, nodules, interlobular septal thickening, crazy-paving pattern, cysts, architectural distortion or honeycombing, bronchial wall thickening, bronchiectasis, pleural effusion, pneumothorax, pericardial effusion, pneumomediastinum, and hilar or mediastinal lymphadenopathy. Interobserver agreement was also performed.
Main results The median age at presentation was 15.7 years old. There were 9 boys and 2 girls. All patients had fever and nausea and/or vomiting. Shortness of breath was only a symptom in 5/11 cases and cough in 6/11. All patients had chest radiographs at presentation, 10/11 of which were abnormal. Chest CT was obtained in 9 cases, all of which were abnormal. The most common findings on CT were ground-glass opacities (9/9), interlobular septal line thickening (8/9), subpleural sparring (8/9), crazy paving pattern (8/9), lymphadenopathy (7/9), and consolidations (6/9). Complete or near-complete resolution in the imaging findings was noted in 5/6 cases at short-term follow-up (median 114 days).
Commentary Reported cases EVALI peaked in 2019 with 2,807 cases and 68 deaths reported by February 2020. This study reports that common CT findings of EVALI in teenagers include ground glass opacities, interlobular septal line thickening, and crazy paving pattern with subpleural sparring and lymphadenopathy. Because these findings do not fit one specific pattern of lung injury a clinical history of recent vaping use and/or the lack other explanation for the findings are key to the diagnosis. Notably, fever and gastrointestinal symptoms were more common that respiratory symptoms. While these findings are consistent with the literature for EVALI findings in adults, the small sample size and lack of pathologic correlation are limitations.
Validating 3D indexes in the non-surgical pectus excavatum patient.
Fuentes S, Pradillos-Serna JM, Berlioz M, et al. Validating 3D indexes in the non-surgical pectus excavatum patient. Article In Press.
J Ped Surg 2020.
Question(s) What is the accuracy and reliability of the measures obtained by a portable 3D scanner in patients with pectus excavatum (PE) of different severity grades (mild, moderate, severe)?
Design Prospective, cross-sectional study
Setting Single-center; Servicio de Cirugía Pediátrica, Complejo Asistencial Universitario de León, Spain
Participants 28 children (21 children with PE and 7 children without PE as controls)
Intervention 3D surface imaging via portable hand-held infrared scanner vs. limited MRI (control)
Outcomes 3D Haller Index (3DHI), 3D Correction Index (3DCI), traditional HI and traditional CI via MRI
Main Results Statistically significant correlation was found between the 3DHI and traditional HI (0.653, P<0.05) and between 3DCI and traditional CI (0.724, P<0.01) in the control group. Statistically significant correlation was also seen between the 3DHI and traditional HI (0.576, P<0.05) and between 3DCI and traditional CI (0.764, P<0.01) in the PE group. Mean values of traditional HI and CI on MRI and 3DCI differed, depending on severity (P<0.001). However, differences among 3DHI values were not statistically significant among the severity groups and controls.
Commentary Despite its small sample size, this study shows that surface 3D imaging technology is accurate and correlates well with traditional cross-sectional imaging, regardless of the severity of the PE deformity. The ease of accessibility and short scanning duration are favorable features of the technology and future studies could be targeted to study dynamics of the deformity during breathing and clinical correlation with symptoms. The major disadvantage of 3D surface scanning compared to MRI is the inability to assess cardiac function.
Microbiology and radiographic features of osteomyelitis in children and adolescents with sickle cell disease
Kao CM, Yee ME, Maillis A, et al. Microbiology and radiographic features of osteomyelitis in children and adolescents with sickle cell disease.
Pediatr Blood Cancer 2020; e28517.
Question(s) What is the diagnostic probability of determining osteomyelitis in children with sickle cell anemia based on imaging findings?
Design Retrospective study from 2010 to 2019
Setting Single center (Emory University/Children’s Hospital of Atlanta, Atlanta, GA)
Participants 3553 patients with sickle cell anemia, 20 total cases of osteomyelitis in 19 patients (one participant had 2 episodes of osteomyelitis over 2 years apart), only 19 of the 20 cases had imaging.
Outcomes Imaging evidence of osteomyelitis and culture confirmed infection
Main Results 18 MRI’s were performed, 1 CT. MR evidence of osteomyelitis was retrospectively read as definitive in 4/19 (21%), probable in 10/19 (53%), and suspected in 5/19 (26%). In the 9 culture confirmed cases (either blood or surgical cultures), the MRs were read as definitive in 2 cases (22%), probable in 4 (44%), and suspected in 3 (33%). Other imaging characteristics included abscess or fluid collection in 58% of cases, adjacent myositis or muscle edema in 79%, joint effusion in 63%, and osteonecrosis in 47%.
Commentary Osteomyelitis is cause of musculoskeletal pain in children with sickle cell anemia. The grading of the likelihood of osteomyelitis on MRI in this study highlights the limitations of MRI alone in diagnosing osteomyelitis. Only 22% of patients with positive cultures were found to have MRI findings deemed definitive for osteomyelitis. Therefore, there is a strong need for radiologic, microbiologic, and clinical collaboration when making diagnosis and treatment decisions.
Utility of ultrasound for evaluating masses in the pediatric population.
LeMoine B and Samet JD. Utility of ultrasound for evaluating masses in the pediatric population. Article In Press.
Adv Clin Rad 2020.
Question(s) What is the accuracy of US in evaluation of pediatric musculoskeletal soft tissue masses and frequency of additional imaging and interventions?
Design Retrospective study from 2007 to 2011
Setting Single center; Lurie Children’s Hospital of Chicago/Northwestern University
Participants 456 soft tissue, nonvascular US exams and 505 masses (some patients had multiple masses)
Intervention US evaluation compared to pathologic diagnosis when available. Masses were presumed to be ‘Benign’ after 2 years of follow up with no mention of malignancy and were considered ‘Indeterminate’ if the patient had < 2 years of clinical encounters and no pathologic analysis.
Outcomes Radiologic impression: no mass, benign, probably benign, indeterminate, malignant; frequency of additional imaging, and rate of intervention based on radiologic impression
Main Results No mass was seen in 12.3% of cases. 41.7% of cases were dictated as ‘Benign,’ 21.5% were dictated as ‘Probably Benign,’ 24.1% were ‘Indeterminate,’ and 0.4% were read as ‘Malignant.’ Interventions rates on ranged from 7.1% (‘No Mass’) to 100% (‘Malignant’). Additional imaging was obtained in a total of 9.4% cases, with MRI being the most frequent.
Commentary The authors show that most pediatric soft tissue masses that present on US are benign and relatively few underwent additional imaging. However, it is possible that more concerning masses may bypass US and MRI may be the initial modality. The intervention rates of ‘Probably Benign’ and ‘Indeterminate’ were similar (32.7% vs. 34.5%, respectively), yet there was no documentation of malignancy for any of the ‘Probably Benign’ masses. This suggests that clinical and imaging surveillance may be acceptable for ‘Probably Benign’ masses rather than surgical intervention
Automatic machine learning to differentiate pediatric posterior fossa tumors on routine MR imaging.
Zhou H, Hu R, Tang O et al. Automatic machine learning to differentiate pediatric posterior fossa tumors on routine MR imaging.
AJNR 2020; 41: 1279–85.
Question(s) How does an automatic model of machine learning compare to a human selected model or expert Neuroradiologists in regards to differentiating pediatric posterior fossa tumors?
Design Retrospective study
Setting Single institution
Participants 288 patients with pediatric posterior fossa tumors
Intervention Automatic machine learning via the Tree-Based Pipeline Optimization Tool (TPOT) , an open-source form of machine learning that chooses the most optimal machine learning pipeline without human intervention, vs. models generated by standard manual optimization by a human machine learning expert (Chi-Square score/Generalized Linear Model) vs. qualitative expert MR imaging review (2 expert Neuroradiologists)
Outcomes Accuracy of TPOT, model chosen by machine learning expert, and qualitative expert MR reviewers in 3-way classifications (medulloblastoma vs. ependymoma vs. pilocytic astrocytoma) and binary classification (medulloblastoma vs. non, ependymoma vs. non, pilocytic astrocytoma vs. non)
Main Results For the 3-way classification, TPOT achieved an AUC of 0.91 with accuracy of 0.83, while the Chi-Square score/Generalized Linear Model had an AUC of 0.92 with accuracy was 0.74. TPOT achieved a significantly higher accuracy than the average qualitative expert MR imaging review (0.83 vs. 0.54, P<0.001). No statistical significance was observed between the two machine learning models (P=0.16). For binary classification, TPOT achieved an AUC of 0.94 with accuracy of 0.85 for medulloblastoma vs. non, AUC of 0.84 with accuracy of 0.8 for ependymoma vs. non, and AUC of 0.94 with accuracy of 0.88 for pilocytic astrocytoma vs. non.
Commentary An automatic machine learning model can perform similarly well to a machine learning model based on manual optimization by a human machine learning expert and better than expert human radiologists. This is not necessarily surprising as pattern recognition is a strength of machine learning. Despite these results, machine learning does not obviate the need for the reference standard of histopathologic diagnosis.
QUALITY & SAFETY
Trends in use of advanced imaging in pediatric emergency departments, 2009-2018
Marin JR, Rodean J, Hall M, et al. Trends in Use of Advanced Imaging in Pediatric Emergency Departments, 2009-2018.
JAMA Pediatr. Published online August 03, 2020.
Question(s) How has the use of advanced imaging changed during the past 10 years across pediatric emergency departments in the US?
Design Cross sectional study for 2009 to 2018
Setting Multicenter study using the Pediatric Health Information System, a database that includes data from 52 tertiary care children’s hospitals in the USA
Participants 32 emergency departments, over 26 million pediatric emergency room visits
Outcomes Change in the use of CT, ultrasonography, and MR imaging over the course of 10 years. Secondary outcomes included ED length of stay, hospitalizations, and 3-day ED revisit rates.
Main results There were a total of 26,082,062 emergency room visits by 9,868,406 children. Advanced imaging (CT, US, or MR) was performed in 1,919,283 with an increase from 6.4% of visits in 2009 to 8.7% in 2018. The use of CT dropped by 1% (3.9% to 2.9%) but ultrasound increased by 3.3% (2.5% to 5.8%) and MR by 0.3% (from 0.3% to 0.6%). All changes were found to be statistically significant with P<0.001. Overall, hospitalizations and 3-day ED revisit rates decreased.
Commentary Over the last 10 years, there has been an overall increase in the use of advanced imaging for pediatric emergency room visits. The use of CT has decreased while the use of nonionizing modalities such as US and MRI has increased. This may be due to campaigns such as Image Gently and Choosing Wisely as well as the increasing use of ultrasound in diagnosing abdominal pathology. Faster, shorter MR protocols can now be used to evaluate shunts without sedation. The overall increase in US use is not matched by the decrease in CT, raising the possibility that US is now being overused or being applied to a broader range of indications. Limitations of this study include the lack of assessment of indications and that only tertiary care children’s hospitals were included so these results may not be generalizable to other settings. Some non-tertiary care hospitals may not have the technology or sonographer training/staffing to choose MRI or US over CT.References