An Automated Deep Learning Method for Tile AO/OTA Pelvic Fracture Severity Grading from Trauma whole-Body CT

Admission trauma whole-body CT is routinely employed as a first-line diagnostic tool for characterizing pelvic fracture severity. Tile AO/OTA grade based on the presence or absence of rotational and translational instability corresponds with need for interventions including massive transfusion and angioembolization. An automated method could be highly beneficial for point of care triage in this critical time-sensitive setting. A dataset of 373 trauma whole-body CTs collected from two busy level 1 trauma centers with consensus Tile AO/OTA grading by three trauma radiologists was used to train and test a triplanar parallel concatenated network incorporating orthogonal full-thickness multiplanar reformat (MPR) views as input with a ResNeXt-50 backbone. Input pelvic images were first derived using an automated registration and cropping technique. Performance of the network for classification of rotational and translational instability was compared with that of (1) an analogous triplanar architecture incorporating an LSTM RNN network, (2) a previously described 3D autoencoder-based method, and (3) grading by a fourth independent blinded radiologist with trauma expertise. Confusion matrix results were derived, anchored to peak Matthews correlation coefficient (MCC). Associations with clinical outcomes were determined using Fisher's exact test. The triplanar parallel concatenated method had the highest accuracies for discriminating translational and rotational instability (85% and 74%, respectively), with specificity, recall, and F1 score of 93.4%, 56.5%, and 0.63 for translational instability and 71.7%, 75.7%, and 0.77 for rotational instability. Accuracy of this method was equivalent to the single radiologist read for rotational instability (74.0% versus 76.7%, p = 0.40), but significantly higher for translational instability (85.0% versus 75.1, p = 0.0007). Mean inference time was < 0.1 s per test image. Translational instability determined with this method was associated with need for angioembolization and massive transfusion (p = 0.002-0.008). Saliency maps demonstrated that the network focused on the sacroiliac complex and pubic symphysis, in keeping with the AO/OTA grading paradigm. A multiview concatenated deep network leveraging 3D information from orthogonal thick-MPR images predicted rotationally and translationally unstable pelvic fractures with accuracy comparable to an independent reader with trauma radiology expertise. Model output demonstrated significant association with key clinical outcomes.

Young–Burgess classification: Inter-observer and inter-method agreement between pelvic radiograph and computed tomography in emergency polytrauma management

Background:

While there are intrinsic differences in the pros and cons between plain radiographs and computer tomography, the role of pelvic radiographs in polytrauma management is diminishing as computer tomography scans are becoming more accessible. Previous studies found varying results in the inter-observer agreement in pelvic radiograph interpretations.

Objectives:

To evaluate inter-observer agreement of classifying pelvic fractures in major trauma patients by emergency physicians, and the inter-method agreement between plain radiographs and computer tomography scans.

Methods:

Three hundred sixty-nine patients with pelvic trauma were recruited from the trauma registries of four designated trauma centres in Hong Kong, each having one set of anteroposterior pelvic radiographs and pelvic computer tomography scans. Pelvic radiographs were classified by two emergency physicians using Young–Burgess classification, and pelvic computer tomography scans classified by an experienced radiologist. Disagreed pelvic radiographs were evaluated by a senior emergency physician to make a final decision before comparing with computer tomography scans. Cohen’s kappa was used to measure the inter-observer and the inter-method agreements, in the groups ‘mechanism of injury’, ‘stable versus unstable fractures’ and ‘complete classification’.

Results:

Inter-observer agreements of plain radiograph classification for ‘mechanism of injury’, ‘stable versus unstable fractures’ and ‘complete classification’ were moderate to substantial (κ = 0.72, 0.60 and 0.55, respectively). Inter-method agreement for the three groups between plain radiographs and computer tomography were fair to moderate (κ = 0.42, 0.59 and 0.38, respectively).

Conclusion:

The inter-method agreement between plain pelvic radiographs and computer tomography scans was fair in classifying pelvic fractures, and moderate in detection of unstable pelvic fractures. If the patient is haemodynamically unstable or when computer tomography is unavailable, it is reasonable to obtain plain radiographs to screen for unstable pelvic fractures to expedite early intervention. A review in the education approach and material of Young–Burgess classification may improve inter-observer agreement.

Trends in the management of pelvic fractures, 2008-2010

BACKGROUND

Bleeding from pelvic fractures can be lethal. Angioembolization (AE) and external fixation (EXFIX) are common treatments to control bleeding, but it is not known how frequently they are used. We hypothesized that AE would be increasingly more common compared with EXFIX over time.

METHODS

The National Trauma Data Bank for the years from 2008-2010 were used. Patients were included in the study if they had an International Classification of Diseases, ninth edition, Clinical Modification codes for pelvic fractures and were aged ≥18 y. Patients were excluded if they had isolated acetabular fractures, were not admitted, or had minor injuries. Outcomes included receiving a procedure and in-hospital mortality.

RESULTS

A total of 22,568 patients met study criteria. AE and EXFIX were performed in 746 (3.3%) and 663 (2.9%) patients, respectively. AE was performed more often as the study period progressed (2.5% in 2007 to 3.7% in 2010; P < 0.001). This remained significant in adjusted analysis (odds ratio per year 1.15; P = 0.008). Having a procedure was associated with higher mortality in unadjusted analyses compared with those with no procedure (11.0% for no procedure versus 20.5% and 13.4% for AE or EXFIX, respectively; P < 0.001). In adjusted analyses, only AE remained associated with higher mortality (odds ratio 1.63; P < 0.001).

CONCLUSIONS

AE in severely injured pelvic fracture patients is increasing. AE is associated with higher mortality, which may reflect the fact that it is used for patients at higher risk of death. The role of AE for bleeding should be examined in future studies.

The emergency radiology of pelvic trauma

The imaging of pelvic trauma is complex and may involve different radiological techniques depending on the severity and type of injury. Following high-energy blunt trauma, computed tomography (CT) is the investigation of choice as it can identify life-threatening findings such as arterial extravasation as well as bony and soft tissue injuries, in particular that of the urological system. In this overview of pelvic imaging in trauma, the role of CT, plain radiography and focussed assessment with sonography in trauma (FAST) are considered, as well as the role of interventional radiology for pelvic haemorrhage.