Synoptic reporting for cervical spine trauma-common data elements towards developing a deep learning model for AO classification of injury.

Abstract

Learning Objectives: 1. To describe the common data elements for structured reporting of Cervical spine trauma per the AO classification 2. Identify the minimum data set required to develop a deep learning model for AO classification of C-spine trauma Background: Cervical spine traumatic injuries have historically had limited consensus around management, partly from the lack of any universally accepted classification. Recent literature supports the use of the AOSpine Trauma classification to improve communication between clinicians.1,5,6,7 While the AOSpine classification can be cumbersome to report, the use of synoptic reporting can assist, using standardised nomenclature, universally required findings and a consistent report structure for the purpose of developing robust deep learning models.4,8 Imaging Findings: C-spine trauma is classified into upper C-spine consisting of C0-C3 and lower C-spine (or subaxial spine) which consists of C3-7. 3 items are included to explain the injury severity and prognosis-1. Injury morphology (type A-C bony injuries; additionally, type F and BL for subaxial) 2. Neurological status (N0-4) 3. Clinical modifiers (M1-4) Imaging is used to classify injury morphology-with fractures classified as type A, B or C. In the Upper C-Spine, the first step is to identify the fractured bone and joint (C0-C2), followed by the type. Type A injuries include isolated bony injuries only (usually stable injuries); type B include ligamentous injuries (potentially unstable injuries); type C include any displacement, translation or instability-suggesting unstable injuries.1,2 Subaxial injuries are classified as type A for compression bony injuries, type B for tension band injuries, type C consist of any displacement or translation. Additionally, the subaxial classification incorporates type F for facet injuries; while type BL is used for 'Bilateral injuries'.1,3 If multiple injuries are present, the most severe injury's grading is given. The neurological status is graded from N0-4 depending on the severity of neurological deficit. Modifiers are clinical factors that account for key features of spinal trauma such as risk of instability, vascular structure involvement, patient comorbidities. We discuss in detail the data points required including the AO classification (morphology, neurology and modifiers), injury mechanism, patient age and comorbidities.4 Conclusion(s): Multidisciplinary consensus is difficult to obtain in time critical settings, and clinicians should be aware of the AOSpine classification for its surgical implications. Synoptic reporting using the AOSpine classification improves communication between clinicians; assists with surgical decision making and provides robust data for the potential development of artificial intelligence assisted point of care clinical decision-making tools.