Accuracy and self-validation of automated bone age determination

The BoneXpert method for automated determination of bone age from hand X-rays was introduced in 2009 and is currently running in over 200 hospitals. The aim of this work is to present version 3 of the method and validate its accuracy and self-validation mechanism that automatically rejects an image if it is at risk of being analysed incorrectly. The training set included 14,036 images from the 2017 Radiological Society of North America (RSNA) Bone Age Challenge, 1642 images of normal Dutch and Californian children, and 8250 images from Tübingen from patients with Short Stature, Congenital Adrenal Hyperplasia and Precocious Puberty. The study resulted in a cross-validated root mean square (RMS) error in the Tübingen images of 0.62 y, compared to 0.72 y in the previous version. The RMS error on the RSNA test set of 200 images was 0.45 y relative to the average of six manual ratings. The self-validation mechanism rejected 0.4% of the RSNA images. 121 outliers among the self-validated images of the Tübingen study were rerated, resulting in 6 cases where BoneXpert deviated more than 1.5 years from the average of the three re-ratings, compared to 72 such cases for the original manual ratings. The accuracy of BoneXpert is clearly better than the accuracy of a single manual rating. The self-validation mechanism rejected very few images, typically with abnormal anatomy, and among the accepted images, there were 12 times fewer severe bone age errors than in manual ratings, suggesting that BoneXpert could be safer than manual rating.

Monitoring Skull Base Abnormalities in Children with Osteogenesis Imperfecta - Review of Current Practice and a Suggested Clinical Pathway

OBJECTIVES

In the context of a lack of national consensus on the benefits of skull base imaging in children with osteogenesis imperfecta (OI), this study aims to analyse and correlate the clinical symptoms and radiological images of children with severe OI.

METHODS

A retrospective case notes and image analysis was carried out on children with complex OI between 2012 and 2018 at a specialist tertiary centre. Data were collected on patient demographic factors, clinical data, imaging findings (presence of Wormian bones, platybasia, basilar impression (McGregor's technique) and basilar invagination (McRae's technique)), and clinical features at the time of imaging.

RESULTS

Of the 127 patients in the OI database, 94 were included. A total of 321 radiographs, 21 CT scans and 39 MRI scans were analysed. Average frequency of radiographs was 8 per 10 years. Of the 94 patients, 58 (62%), 10 (11%), 1 (1%) demonstrated platybasia, basilar impression, and basilar invagination, respectively. Of the radiographs analysed, platybasia, basilar impression, basilar invagination, and the presence of Wormian bones, could not be evaluated in 71 (22.3%), 48 (15.2%), 61 (19.5%) and 28 (9.4%) radiographs respectively (due to poor positioning, anatomical abnormalities, and poor image quality). Of the 140 radiographs with platybasia, 17 (12%) also demonstrated basilar impression compared to only 3 (2.9%) out of the 99 without platybasia (p = 0.03). No significant associations were seen between the presence of Wormian bones and basilar impression. Of the 39 MRIs, additional information on CSF flow rate, spinal cord signal and cerebellar morphology was reported in 14 (36%). There was a lack of concordance between MRI and matched radiographs in 7.1% (1/14) and 36% (5/14) for platybasia and basilar impression respectively, with full concordance for basilar invagination. Fewer than 5% had positive clinical symptoms/signs at the time of imaging; 2% (7/321) had macrocephaly, 0.6% (2/321) headache, all other neurological features were absent). Clinical features were not documented in >85% of patients.

CONCLUSION

The apparent low prevalence of clinical symptoms and signs and of radiologically identified cranio-cervical abnormalities, suggests that current levels of serial imaging may be excessive. Until larger prospective studies clarify these issues, we suggest a clinical pathway for base of skull imaging which proposes a risk stratification approach to radiographic frequency and suggests parameters for proceeding to MRI.

The significance of internal calcifications on perinatal post-mortem radiographs

AIM

To determine whether the presence of internal calcifications on perinatal post-mortem skeletal surveys (PMSS) are associated with certain diagnoses of fetal loss.

METHODS AND MATERIALS

A 6-month retrospective, single-centre, cohort study was conducted on PMSS performed for perinatal death assessment. One reader re-reviewed all PMSS images for the presence and location of internal calcifications, and noted whether these were included within the original radiology report. Findings at autopsy were then reviewed independently by a second researcher and cause of fetal loss or main diagnosis recorded. Chi-squared tests were conducted to identify differences between those with and without internal calcifications at PMSS.

RESULTS

Two hundred and thirty perinatal deaths (mean gestational age 18 weeks; average 12–35 weeks) were included in the study, of which 42 (18.3%) demonstrated intra-abdominal calcifications, and 16/42 (38.1%) were mentioned in the radiology reports. Most calcifications were found to be within the lumen of the gastrointestinal tract, and in the left upper quadrant of the abdomen. There was no statistical difference between identifiable causes for fetal loss at autopsy in cases with and without calcification at PMSS (59.5% versus 58.5% respectively, p=0.904). Nevertheless, where calcification and a cause for fetal loss were found, the aetiology was more likely to be due a fetal rather than placental issue.

CONCLUSION

The presence of internal calcifications on PMSS was not associated with an increased likelihood of explainable fetal loss or particular diagnosis at autopsy.

•

Fetal calcifications on post-mortem skeletal surveys are not significantly associated with causes of fetal loss.

•

When present, internal calcifications were usually seen in lower gestational aged fetuses, intra-abdominal in location, with the majority being intraluminal.

•

The majority of radiologists do not report internal fetal calcifications on radiographs, without significant consequences for final outcome at autopsy.

Role of routine computed tomography in paediatric pleural empyema

BACKGROUND

The incidence of empyema in children is increasing worldwide. While there are emerging data for the best treatment options, there is little evidence to support the imaging modalities used to guide treatment, particularly with regard to the role of routine CT scanning. The aims of this study were to develop a radiological scoring system for paediatric empyema and to assess the utility of routine CT scanning in this disease.

METHODS

Children with empyema were prospectively enrolled over a 3-year period into a randomised clinical trial of video-assisted thoracoscopic surgery versus percutaneous chest drain insertion and urokinase. All children received a preoperative chest radiograph (CXR), pleural ultrasound scan (USS) and chest CT scan. In the urokinase arm the clinician inserted the drain with USS evidence only and did not have access to the CT scan at the time of insertion to reflect clinical practice. A scoring system was developed for each individual radiological modality and used to compare imaging characteristics of the pleural fluid collection and underlying parenchyma and to assess the utility of USS and CT to predict length of stay after the intervention.

RESULTS

Of the 60 subjects recruited, 46 had USS images available for review, 36 had a CT scan which met the inclusion criteria and 31 had all three radiological measurements (CT, USS and CXR) available for analysis. There was substantial interobserver agreement for USS grades (kappa = 0.709) and moderate agreement for total CT scores (kappa = 0.520). There were weak correlations between USS grade and total CT score as well as CT loculation and density scores. Of the 25 CXRs showing simple opacification of the underlying parenchyma only, CT demonstrated simple consolidation (n = 14), necrotising pneumonia (n = 7), cavitary necrosis (n = 3) and pneumatoceles (n = 1). No abnormality was detected on CT scanning which directly altered clinical management. Neither the USS score nor the CT score, nor a combination of the two, were able to predict length of hospital stay.

CONCLUSIONS

CT scanning detects more parenchymal abnormalities than chest radiography. However, the additional information does not alter management and is unable to predict clinical outcome. This suggests that there is no role for the routine use of CT scanning in children if treated with urokinase and percutaneous chest drain. The omission of routine CT scanning in empyema will reduce the exposure of children to unnecessary radiation and reduce costs.

TRIAL REGISTRATION NUMBER

The trial is fully registered with clinicaltrials.gov (ID: NCT00144950).