Published online before print August 9, 2007, 10.1148/radiol.2443051785
(Radiology 2007;244:838-844.)
© RSNA, 2007
The Brain in Children: Is Contrast Enhancement Really Needed after Obtaining Normal Unenhanced CT Results?1
Helen M. Branson, BSc, MBBS, FRACR,
Andrea S. Doria, MD, MSc, PhD,
Rahim Moineddin, PhD, and
Manohar M. Shroff, MD
1 From the Department of Diagnostic Imaging, The Hospital for Sick Children, 555 University Ave, Toronto, ON, Canada M5G 1X8 (H.M.B., A.S.D., M.M.S.); and Department of Public Health, Family and Community Medicine, University of Toronto, Ontario, Canada (R.M.). Received November 8, 2005; revision requested January 4, 2006; revision received March 24; accepted April 20; final version accepted January 8, 2007. Supported in part by a research award from the Department of Medical Imaging, University of Toronto.
Address correspondence to M.M.S. (e-mail: manohar.shroff{at}sickkids.ca).
 |
ABSTRACT
|
|---|
Purpose: To retrospectively determine and compare the sensitivity and specificity of unenhanced and contrast material–enhanced computed tomography (CT) (reference standard) in the diagnosis of brain abnormalities and to evaluate any change in diagnosis that resulted from the contrast-enhanced study.
Materials and Methods: This study was approved by the local research ethics board; the requirement for informed consent was waived. The authors reviewed the unenhanced and contrast-enhanced CT scans of the brain obtained in 353 children for indications other than trauma. There were 196 boys and 157 girls aged 0 months to 17.8 years. Scans were read independently by two pediatric neuroradiologists who were blinded to clinical information. The diagnosis for each scan was recorded according to the anatomic section (supratentorial, infratentorial, ventricles, and skull). The final diagnosis was classified as normal, abnormal, or equivocal. 
Statistics, with 95% confidence intervals, were reported, and the sensitivity, specificity, positive predictive value, and negative predictive value were calculated.
Results: Interreader agreement for different anatomic regions varied between good ( coefficient, 0.63) and very good ( coefficient, 0.88) for unenhanced and contrast-enhanced scans. Sensitivity, specificity, positive predictive value, and negative predictive value for unenhanced scans were 97%, 89%, 87%, and 97%, respectively. The use of contrast material led to a change in the original normal or equivocal diagnosis to an abnormal diagnosis for only five (2.7%) of the 183 normal unenhanced scans.
Conclusion: Unenhanced CT of developing brains has high sensitivity and specificity in the diagnosis of pathologic findings. The use of intravenous contrast material after unenhanced CT of the brain in children did not change the diagnosis frequently.
© RSNA, 2007
|
INTRODUCTION
|
|---|
Studies in adults (1–4) previously have been performed to examine the usefulness of and indications for contrast material–enhanced computed tomography (CT) of the brain, revealing that contrast enhancement is only helpful after the acquisition of a normal unenhanced scan and when there are persistent focal signs and symptoms suggestive of an intracranial abnormality. To our knowledge, no prior investigations have been conducted in children in such situations to address the usefulness of contrast-enhanced CT in the diagnosis of abnormalities of the brain.
We hypothesized that the sensitivity of unenhanced CT scans in patients with low risk for an inflammatory or neoplastic process would be high compared with that of corresponding contrast-enhanced scans. Thus, the purpose of our study was to retrospectively determine and compare the sensitivity and specificity of unenhanced and contrast-enhanced CT (reference standard) in the diagnosis of brain abnormalities and to evaluate any change in diagnosis that resulted from the contrast-enhanced study.
|
MATERIALS AND METHODS
|
|---|
Sample Size Calculation
This study was approved by the research ethics board of The Hospital for Sick Children; the requirement for informed consent was waived. A total of 352 examinations were required for a power of 90% and an 
error of .05 (two-tailed test level) on the basis of a 5% difference in the false-negative rate of unenhanced and contrast-enhanced CT scans obtained from studies in adults (2–4).
Selection Criteria for Patients
We searched for scans of patients who were referred to our CT scan service for clinical indications other than trauma through the emergency department of our institution, inpatient wards, or outpatient clinics. Research assistants searched for and collected scans by using our CT logbook database, excluding scans of children with a previous diagnosis of ventricular shunt and incomplete scans. We retrieved 400 pediatric brain CT scans obtained before and after the intravenous administration of contrast material during a 9-month period (June 2003 to March 2004). Three hundred seventy-eight patients met the criteria for inclusion in our study, and their images were reviewed. Twenty-five patients with contrast-enhanced scans that were equivocal for the presence of abnormalities (eg, patients with a diagnosis of prominent subarachnoid spaces but no available information about the head circumference and those with prominent and/or asymmetric ventricles) were excluded from the study. Thus, the data for a total of 353 patients, each with a scan obtained before and after the intravenous administration of contrast material, were included.
There were 196 boys and 157 girls aged 0 months to 17.8 years (mean age, 6.7 years). The clinical diagnoses (as recorded on the CT request forms) were congenital abnormalities (n = 19), infectious and inflammatory processes (n = 47), seizures (n = 49), stroke (n = 59), intracranial space occupation and tumors (n = 95), vascular abnormalities (n = 6), and miscellaneous (n = 78).
Image Acquisition
All scans were obtained with a helical eight-section unit (LightSpeed Ultra; GE Medical Systems, Milwaukee, Wis). Transverse scans were obtained from the skull base to the top of the vertex with 120 kVp and an acquisition time of 2.0 seconds per revolution. Transverse scans were obtained with 5-mm-thick sections for the infratentorium and the supratentorium in all patients. Depending on the age of the patient, the amperage varied from 80 to 120 mA for the posterior fossa and from 70 to 110 mA through the supratentorium. In each patient, however, the milliampere-second setting was the same for both unenhanced and contrast-enhanced scans. Contrast-enhanced scans were obtained after manual or mechanical (power injector) intravenous administration of 2.5 mL per kilogram of body weight iohexol (Omnipaque-300; Amersham Health, Princeton, NJ).
Imaging Analysis
One pediatric neuroradiologist with more than 10 years experience (M.M.S.) and one pediatric neuroradiology fellow with 2 years experience (H.M.B.) independently reviewed the scans while blinded to the patients' clinical information, which was electronically stripped from all scans. A random number code was assigned to each unenhanced and contrast-enhanced scan, and the order of the scans was randomly assigned. The unenhanced and contrast-enhanced scans were read independently by each radiologist. Images were viewed at a standard soft-tissue algorithm with a window width of 54 HU and a window level of 35 HU. This is the standard setting for viewing the brain on the picture archiving and communication system used in our institution (Centricity; GE Medical Systems). In addition, as is usual practice, images were viewed at several different window levels and widths. Bone windows and/or bone algorithms were reviewed when available. The diagnosis for each unenhanced and contrast-enhanced scan (normal, abnormal, or equivocal) was recorded according to the following anatomic sections of the brain: supratentorium, infratentorium, ventricles, subarachnoid spaces, and skull. For scans with interreader disagreement (72 unenhanced and 50 contrast-enhanced scans), consensus was reached after the reviewers re-read the scans.
Reference Standard, Change in Diagnosis, and Statistical Analysis
The contrast-enhanced scan was considered the reference standard. With regard to each anatomic compartment evaluated, scans that showed abnormalities at both unenhanced and contrast-enhanced CT were considered true-positive. Scans with normal findings at both unenhanced and contrast-enhanced CT were considered true-negative. Scans that were interpreted as showing abnormalities at unenhanced CT and normal findings at contrast-enhanced CT were considered false-positive. Scans interpreted as normal at unenhanced CT but as showing abnormalities at contrast-enhanced CT were considered false-positive. Equivocal results were considered false-positive or false-negative findings and were incorporated into the final analysis. An equivocal scan in a patient with a corresponding normal contrast-enhanced scan was considered to be a false-positive result. An equivocal scan in a patient with a corresponding abnormal contrast-enhanced scan was considered to be a false-negative result. Equivocal contrast-enhanced scans were excluded from the final analysis. The unenhanced equivocal scans, however, were incorporated into the final statistical analysis, as described earlier.
The usefulness of contrast-enhanced CT was classified on the basis of the following criteria: (a) No further information was gained (normal to normal findings equal true-negative results), (b) contrast-enhanced CT helped to confirm the diagnosis made at unenhanced CT (abnormal to abnormal findings equal true-positive results), and (c) the original diagnosis was changed owing to the findings at contrast-enhanced CT (normal or equivocal to abnormal findings equal false-negative results, abnormal or equivocal to normal findings equal false-positive results).
The sensitivity, specificity, and positive and negative predictive values of unenhanced CT compared with those of contrast-enhanced CT were calculated on the basis of the true-positive, true-negative, false-positive, and false-negative rates. A software package (SAS, version 9.1, 2003; SAS Institute, Cary, NC) was used for statistical calculations. Ninety-five percent confidence intervals (CIs) were calculated by using binomial distribution. Sensitivities of anatomic regions were calculated by using the log-linear model and generalized estimation equation method.
The level of agreement between readers was determined by using values, which were measures of interreader concordance (adjusted for chance agreement). Values of less than 0.40 were indicative of poor concordance; values of 0.40–0.60, of moderate concordance; values of 0.60–0.80, of good concordance; and values greater than 0.80, of very good agreement (5). For each statistic, the 95% CI is reported. P < .05 was considered indicative of a statistically significant difference.
|
RESULTS
|
|---|
Sensitivity and Specificity
The sensitivity, specificity, and positive and negative predictive values for unenhanced CT scans were 97% (143 of 148; 95% CI: 92%, 99%), 89% (183 of 205; 95% CI: 85%, 94%), 87% (143 of 165; 95% CI: 81%, 92%), and 97% (183 of 188; 95% CI: 93%, 99%), respectively (Table 1). The sensitivity of unenhanced scans was slightly higher for the diagnosis of abnormalities in the supratentorium (90%) than for the diagnosis of abnormalities in the infratentorium (69%) (P = .046), ventricles (75%) (P = .016), and skull (71%) (P = .044). The specificity of unenhanced scans in the diagnosis of infratentorial (96%) and skull (96%) abnormalities was higher than that for the diagnosis of supratentorial (91%) (P = .007 for comparison with infratentorial data) and ventricular (86%) (P < .001 for comparison with infratentorial data) abnormalities (Table 2).
View this table:
[in this window]
[in a new window]
|
Table 2. Sensitivity, Specificity, and Positive and Negative Predictive Values for Unenhanced CT according to Anatomic Compartment
|
|
Changes in Diagnosis
Of the 353 unenhanced scans, 183 (51.8%) were classified as normal. The use of contrast material changed the initial normal or equivocal diagnosis to abnormal in only five of the 183 scans (2.7%). In three of the 183 patients (1.6%), the diagnosis for the scan changed from normal to abnormal (Table 1; Figs 1, 2).
View larger version (92K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1a: Change in the clinical diagnosis for brain CT scan obtained in 1-year-old girl from normal to abnormal after administration of contrast material. (a) Unenhanced transverse CT scan reported as normal. (b) Corresponding contrast-enhanced transverse CT scan demonstrates right cerebellar developmental venous anomaly (arrow). (c) Unenhanced transverse CT scan obtained near the vertex reported as normal; however, in retrospect, the patient's superior sagittal sinus (arrow) had low attenuation. (d) Corresponding contrast-enhanced transverse CT scan demonstrates filling defect in posterior sagittal sinus (arrow), which most likely represents an incidental arachnoid granulation.
|
|
View larger version (76K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1b: Change in the clinical diagnosis for brain CT scan obtained in 1-year-old girl from normal to abnormal after administration of contrast material. (a) Unenhanced transverse CT scan reported as normal. (b) Corresponding contrast-enhanced transverse CT scan demonstrates right cerebellar developmental venous anomaly (arrow). (c) Unenhanced transverse CT scan obtained near the vertex reported as normal; however, in retrospect, the patient's superior sagittal sinus (arrow) had low attenuation. (d) Corresponding contrast-enhanced transverse CT scan demonstrates filling defect in posterior sagittal sinus (arrow), which most likely represents an incidental arachnoid granulation.
|
|
View larger version (75K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1c: Change in the clinical diagnosis for brain CT scan obtained in 1-year-old girl from normal to abnormal after administration of contrast material. (a) Unenhanced transverse CT scan reported as normal. (b) Corresponding contrast-enhanced transverse CT scan demonstrates right cerebellar developmental venous anomaly (arrow). (c) Unenhanced transverse CT scan obtained near the vertex reported as normal; however, in retrospect, the patient's superior sagittal sinus (arrow) had low attenuation. (d) Corresponding contrast-enhanced transverse CT scan demonstrates filling defect in posterior sagittal sinus (arrow), which most likely represents an incidental arachnoid granulation.
|
|
View larger version (85K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 1d: Change in the clinical diagnosis for brain CT scan obtained in 1-year-old girl from normal to abnormal after administration of contrast material. (a) Unenhanced transverse CT scan reported as normal. (b) Corresponding contrast-enhanced transverse CT scan demonstrates right cerebellar developmental venous anomaly (arrow). (c) Unenhanced transverse CT scan obtained near the vertex reported as normal; however, in retrospect, the patient's superior sagittal sinus (arrow) had low attenuation. (d) Corresponding contrast-enhanced transverse CT scan demonstrates filling defect in posterior sagittal sinus (arrow), which most likely represents an incidental arachnoid granulation.
|
|
View larger version (110K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2a: Transverse CT images illustrate changes in the clinical diagnosis for a brain CT scan from normal to abnormal. (a) Unenhanced scan obtained in 6-day-old girl reported as normal. (b) Corresponding contrast-enhanced scan demonstrates white matter edema (arrows).
|
|
View larger version (46K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 2b: Transverse CT images illustrate changes in the clinical diagnosis for a brain CT scan from normal to abnormal. (a) Unenhanced scan obtained in 6-day-old girl reported as normal. (b) Corresponding contrast-enhanced scan demonstrates white matter edema (arrows).
|
|
The use of contrast material changed the diagnosis from abnormal or equivocal on the unenhanced scan to normal on the contrast-enhanced scan for 22 patients (Table 1). These abnormalities included left mesial temporal low attenuation (n = 2), left periventricular low attenuation (n = 1), left frontal white matter low attenuation (n = 2), frontal low attenuation and possible pineal cyst (n = 1), small infarcts (n = 1), loss of gray-white matter differentiation (n = 1), subarachnoid cyst (n = 1), asymmetric ventricles (n = 3), prominent ventricles with or without subarachnoid spaces (n = 9) (Fig 3), and calcified falx (n = 1).
View larger version (99K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3a: (a) Unenhanced transverse CT scan in 2-month-old girl shows slightly enlarged ventricles and subarachnoid spaces (arrows), which were reported as abnormal. (b) Corresponding contrast-enhanced CT scan shows "filling-in" of these spaces with normal vessels (arrows), a finding that prompted a change in the diagnosis to normal.
|
|
View larger version (97K):
[in this window]
[in a new window]
[Download PPT slide]
|
Figure 3b: (a) Unenhanced transverse CT scan in 2-month-old girl shows slightly enlarged ventricles and subarachnoid spaces (arrows), which were reported as abnormal. (b) Corresponding contrast-enhanced CT scan shows "filling-in" of these spaces with normal vessels (arrows), a finding that prompted a change in the diagnosis to normal.
|
|
Interreader Agreement
Interreader agreement for different anatomic regions ranged from good for the ventricles ( coefficients: 0.63 [95% CI: 0.54, 0.72] for unenhanced CT, 0.70 [95% CI: 0.60, 0.79] for contrast-enhanced CT) to very good for the supratentorial region ( coefficients: 0.84 [95% CI: 0.78, 0.91] for unenhanced CT, 0.88 [95% CI: 0.83, 0.94] for contrast-enhanced CT) for both unenhanced and contrast-enhanced scans (Table 3).
|
DISCUSSION
|
|---|
The main results of our study were that the sensitivity of unenhanced CT scans was high (97%) compared with that of corresponding contrast-enhanced scans and that the intravenous administration of contrast material does not lead to a change in the clinical diagnosis for a substantial number of patients. A change in diagnosis after the intravenous administration of contrast material—from normal or equivocal to abnormal—occurred in five of 183 patients (2.7%). Therefore, the performance of a second contrast-enhanced CT examination should be considered with regard to the potential risks related to radiation (6,7), sedation, and allergic reactions to contrast material.
The sensitivity, specificity, and positive and negative predictive values for unenhanced scans in a pediatric population were similar to those obtained in a study of adults in which the diagnostic performance of unenhanced CT of the head was evaluated (1). In that study, the sensitivity, specificity, and positive and negative predictive values of unenhanced CT in adults were 94%, 100%, 100%, and 98%, respectively (1). Overall, there were substantially more false-positive findings than false-negative findings in our study in contrast to the results of studies with adults. Demaerel et al (1) had no false-positive findings in their cohort; Cowan and MacDonald (2) stated that 19 of 400 abnormal scans had a changed diagnosis after contrast material administration, but they did not state the nature of these changes. In our study, the overall false-positive rate was greater than the false-negative rate for the unenhanced scans, yielding a sensitivity that was marginally greater than the specificity for unenhanced CT. Specific training of radiologists to interpret potential false-positive findings on unenhanced scans of the head could help to improve the specificity of the method for the diagnosis of brain abnormalities in children.
The proportion of patients in whom the use of contrast material led to a change in the original normal diagnosis to an abnormal diagnosis at head CT in adults, which was 5% in two examinations (n = 400 [2] and n = 300 [3]) and 9.7% (n = 547) (4) in another examination, was slightly higher than or similar to that in our study. These findings suggest that the proportion of vascularized brain structures that necessitate contrast material enhancement for proper visualization is similar or even smaller in the pediatric population. Previous studies of head CT in adults have shown that contrast material contributed to the diagnosis only when there was a suspicious abnormality on the unenhanced scan (1) or when there were persistent focal clinical signs and symptoms suggestive of an intracranial abnormality (3). In our study, only one of the five patients in whom the diagnosis changed from normal or equivocal to abnormal had a diagnosis of cerebral edema that might have had some implication for patient treatment. We would therefore argue that the other changes in diagnosis in our study would not substantially alter patient treatment. If we were to consider that the three positive cases found after contrast material administration showed developmental venous anomalies as an incidental finding with no clinical relevance, this would further strengthen our conclusion that the use of contrast-enhanced CT is not justified given the added risks related to contrast material administration and radiation.
With regard to the effect of contrast material injection on the interpretation of findings in different anatomic compartments, the false-negative rate of unenhanced scans was lower for the supratentorial region than for the other regions in this study. Conversely, the false-positive rate was lower for the infratentorial region and skull than for the ventricles and supratentorial region. This was likely due to the inherent subjectivity in the assessment of normal- versus abnormal-size ventricles and subarachnoid spaces according to the child's age. The abnormalities found only on the contrast-enhanced scan in other studies (1,2,4) include developmental venous anomalies, ectatic cerebral arteries, and small stable meningioma.
For both unenhanced and contrast-enhanced scans, the highest and lowest interrater agreement was noted for the supratentorial and ventricular systems, respectively. The subjectivity inherent in the interpretation of normal morphologic variants for the ventricles and in the evaluation of the normal range of size for the ventricles may have contributed to the lower interreader concordance with regard to the presence or absence of findings in the ventricular system. The development of an atlas of normal CT findings in children of different ages may help to improve the objectivity in the assessment of normal anatomic structures in the developing brain. To our knowledge, no prior study investigators had attempted to evaluate interreader agreement rates in the evaluation of findings according to the anatomic compartments of the brain.
Our study had limitations. An important shortcoming was the increased number of false-positive unenhanced scan results. The patients in whom the diagnosis changed from abnormal to normal were likely overrepresented because of asymmetric or slightly prominent ventricles and subarachnoid spaces, which were overdiagnosed on the unenhanced scans. The number of false-positive results might have been lower if the patients' head circumference and age were known, because this information would have enabled more accurate assessment—especially in patients with extraventricular obstructive hydrocephalus (also known as benign external hydrocephalus) or atrophy. Of the 11 scans classified as abnormal with unenhanced CT and normal with contrast-enhanced CT, three demonstrated a "fogging" effect (8)—that is, low-attenuation areas due to ischemia (on the basis of clinical and follow-up findings)—which disappeared on the contrast-enhanced scans. None of the 11 equivocal scans demonstrated the fogging effect.
In our study, the rate of false-positive findings with which the unenhanced scan was considered equivocal was 3.1% (11 of 353 scans), and the rate of false-negative findings with which the unenhanced scan was considered equivocal was 0.56% (two of 353 scans). Most studies performed to evaluate the diagnostic performance of unenhanced and contrast-enhanced CT of the head in adults (1,2,4) do not include reports of the rate of equivocal findings. Investigators in another study of gastrointestinal diseases, such as appendicitis, reported an indeterminate case rate of up to 15% at abdominal CT (9), resulting in a mean false-positive rate of 1.9% ± 0.41 (standard deviation) and a mean false-negative rate of 0.50% ± 0.41 for equivocal findings. In our study, 11 of the 353 patients had false-positive findings (mean rate, 3.1% ± 3.1 for equivocal findings) and three had false-negative findings (mean rate, 0.85% ± 0.56 for equivocal findings). A higher rate of equivocal findings in false-positive cases may be related to differences in the type of tissue being assessed. Another important limitation of this study was the lack of final pathologic diagnoses available for review; contrast-enhanced scans, which are not ideal reference standards, were used for comparisons with unenhanced scans.
In conclusion, unenhanced scans of developing brains have high sensitivity for the diagnosis of pathologic findings. The use of intravenous contrast material after obtaining a normal unenhanced scan of the brain in children did not change the diagnosis in most patients. Unless an abnormality is suspected on the unenhanced CT scan, we believe that the use of contrast material and repeat scanning, with their potential increased risks, may not be necessary.
|
ADVANCE IN KNOWLEDGE
|
|---|
- In the pediatric population, use of intravenous contrast material does not frequently lead to a change in the diagnosis for a normal unenhanced brain CT scan.
|
ACKNOWLEDGMENTS
|
|---|
We thank Manju Asdhir, BSc, and Nancy Padfield, MRTR, for retrieving cases from the radiology CT database and Pheroze Bharucha, MBBS, for assistance with CT review and data entry.
|
FOOTNOTES
|
|---|
Abbreviations: CI = confidence interval
Authors stated no financial relationship to disclose.
Author contributions:Guarantors of integrity of entire study, A.S.D., M.M.S.; study concepts/study design or data acquisition or data analysis/interpretation, all authors; manuscript drafting or manuscript revision for important intellectual content, all authors; manuscript final version approval, all authors; literature research, H.M.B., M.M.S.; clinical studies, H.M.B., M.M.S.; statistical analysis, H.M.B., A.S.D., R.M.; and manuscript editing, all authors
|
References
|
|---|
- Demaerel P, Buelens C, Wilms G, Baert AL. Cranial CT revisited: do we really need contrast enhancement? Eur Radiol 1998;8:1447–1451.[CrossRef][Medline]
- Cowan I, MacDonald S. How useful is contrast enhancement after a normal unenhanced computed tomography brain scan? Australas Radiol 1999;43:448–450.[CrossRef][Medline]
- Bernard MS, Hourihanj MD, Adams H. Computed tomography of the brain: does contrast enhancement really help? Clin Radiol 1991;44:161–164.[CrossRef][Medline]
- Chishti FA, Al Saeed OM, Al-Khawari H, Shaikh M. Contrast-enhanced cranial computed tomography in magnetic resonance imaging era. Med Princ Pract 2003;12:248–251.[CrossRef][Medline]
- Altman DG, ed. Practical statistics for medical research. London, England: Chapman & Hall, 1991; 404–408.
- Hall P, Adami HO, Trichopoulos D, et al. Effect of low doses of ionizing radiation in infancy on cognitive function in adulthood: Swedish population based cohort study. BMJ 2004;328:19–24.[Abstract/Free Full Text]
- Brenner D, Elliston C, Hall E, Berdon W. Estimated risks of radiation-induced cancer from pediatric CT. AJR Am J Roentgenol 2001;176(2):289–296.[Abstract/Free Full Text]
- Becker H, Desch H, Hacker H, Pencs A. CT fogging effect with ischemic cerebral infarcts. Neuroradiology 1979;18(4):185–192.
- Taylor GA, Callahan MJ, Rodriguez D, Smink DS. CT for suspected appendicitis in children: an analysis of diagnostic errors. Pediatr Radiol 2006;36(4):331–337.
TOP
ABSTRACT
INTRODUCTION
