Computer-Aided Detection of Cervical Spine Injuries: A Feasibility Project

The cervical spine (the neck region) is a highly flexible part of the spine and is particularly vulnerable to trauma, such as injuries resulting from automobile accidents, falls, and shallow water dives. Dislocation or fracture of the cervical vertebrae has the potential for long-term and life-changing disabilities. Patients suspected of cervical spine injury (CSI) are often assessed using x-ray imaging. However, detection of CSI on the images presents a major challenge for emergency physicians, combining images that can be extremely difficult to interpret, particularly to the less experienced eye, with clinical scenarios that may result in death or serious disability with a failure to establish the correct diagnosis. Unfortunately, up to 20% of cases have a delayed or incorrect diagnosis - this has a significant personal and socioeconomic impact on family members, as well as a large financial burden on the NHS and social services. A paralysed person will incur between £1M and £3M in lifetime medical expenses.

Therefore, early and accurate detection of a CSI is critical to plan appropriate care and prevent further injury. We propose novel computer-aided detection (CAD) software to aid in detection of dislocations and fractures to the cervical spine. To our knowledge, there is no existing CAD software designed for cervical spine injuries imaged with x-ray. Our research will be the first to produce alignment curves so that the physician can more accurately determine if a vertebra is displaced, as well as prompts designed to direct the physician's attention to a region of the image suspected of injury. The physician would use the CAD as a "spell-checker" for radiological images of the cervical spine. In addition to research and development of the software, the project includes a clinical validation, to determine the effect the CAD has on the ability of human readers to detect CSIs.

We have formed an interdisciplinary (computer science, radiology, emergency medicine) team of researchers from three institutions to address this problem, and have the right set of skills to successfully deliver the project. The PI is a Senior Lecturer in computer science and an expert in the field of medical image analysis, and has over ten years' experience in commercial research and development, including CAD software for medical images. The Co-I is a director of education for medical imaging and a diagnostic radiographer with experience in accident and emergency. In addition, the team includes two practicing emergency medicine physicians, who are also research active. Furthermore, an SME working in the field of spine analysis software has expressed an interest in the project and has made a valuable contribution of software and staff time.

The primary beneficiary of the CAD software will be emergency department physicians. We ultimately envision CAD as being part of the routine clinical workflow when treating patients with suspected cervical spine injury. The physicians will benefit from more accurate detection of dislocations and fractures, leading to improved healthcare outcomes for their patients. Wider beneficiaries include patients, the healthcare managers, regulators, policy-makers, UK industry, and academics. The project falls within scope of the EPSRC Healthcare Technologies Challenge Theme and involves the healthcare technology industry, where the UK is recognised as a leading international innovator.

The project will last 12 months and is feasible given the expertise of the team. Expected outputs include prototype CAD software as well as results of a validation study demonstrating the effectiveness of the CAD in assisting junior radiographers in detected cervical spine injuries. Other expected outputs include clinical and scientific publications as well as raised awareness through presentations at conferences, training of individuals, a strong online presence, community and policy engagement, and commercial exploitation.

There are SIX main beneficiaries from the research on computer-aided detection (CAD) software for detecting cervical spine injury (CSI).

1. Emergency department (ED) physicians.
This is the primary beneficiary group. We envision CAD as being part of the routine clinical workflow when treating patients with suspected CSI. Given the potentially life-changing consequences of a delayed or missed diagnosis, ED physicians are under great pressure to quickly and robustly identify any fractures or dislocations from the cervical spine series. However, given the subtle appearance of CSI this is a difficult challenge. This issue is compounded by the fact junior physicians working on a rotational basis in the ED have far less experience in reading radiologic images. In other parts of the body such as the colon, it has been shown that CAD software can significantly improve the diagnostic performance of less experienced physicians, compared to experts (Baker et al., 2012).

2. Patients.
Approximately 40,000 people in the UK are living with paralysis resulting from CSI (Spinal Research Facts and Figures, 2012). The global figure is estimated at 2.5M people. Despite active and promising research for novel therapies such as those involving stem cells, there are currently no effective cures for CSI. Severing the contact between the brain and other organs can lead to a host of secondary medical issues, including autonomic dysreflexia, bladder and bower problems, loss of sexual function, spasticity, pressure sores, and inability to control body temperature and blood pressure. From an individual patient's point of view, an improved diagnosis through CAD may mean the difference between a difficult life with paralysis and full recovery following a CSI.

3. Healthcare managers.
The lifelong healthcare costs resulting from CSI place a huge burden on the NHS, spinal cord injury centres, institutions such as hospitals and nursing homes. The effectiveness of CAD, coupled with other factors like cost, will inform healthcare policy (NICE in the UK, insurance providers in the USA) providing it as an option to CSI patients.

4. Regulators.
The scope of work in this project is to determine the feasibility of cervical spine CAD to improve CSI patient outcomes. The project can contribute to the evidence of whether positive outcomes can come about as a result of CAD. This clinical evidence may be part of regulatory submissions to relevant bodies in order to receive approval for CAD as a medical device for non-investigational use.

5. Policy-makers.
Due to loss of wages and inability to work, many patients must rely on state support for care. The economic cost of CSI is estimated to be £1 billion per year in the UK alone (Borius et al., 2010). Improved detection of CSI has the potential to reduce these costs.

6. UK industry.
The first CAD systems were targeted towards mammography; and now breast CAD systems are routinely used for clinical screening of breast cancer. CAD has now widened its application areas and some segments of the CAD market are seeing widespread acceptance, and CAD systems are being researched for diverse applications in the body including prostate, liver, and bones. The bone-CAD market is at a nascent stage and is expected for high growth (Frost & Sullivan, 2011). The general CAD market is valued at US$600 million worldwide and is predicted to grow at 15.6% annually. Ultimately we see C-spine CAD as part of routine clinical practice and CAD to contribute to preserving UK competence in the medical technology sector.