Assess the Clinical Effectiveness in AI Prioritising CT Heads

Study background:

Emergency Departments (ED) across the UK are overburdened with increasing patient demand, radiology staff shortages and rising patient wait times. Head injuries are a frequent cause of emergency attendance in the UK with computed tomography(CT) scans usually the first imaging tests to diagnose head injuries and strokes.

A report issued by National Institute of Health and Care Excellence (NICE), confirms that each year 1.4 million people attend emergency departments in England and Wales with head injury. Among the 200,000 patients admitted annually, one-fifth of them suffer from a Traumatic Brain Injury with skull fracture or evidence of brain damage. Head injury is the most common cause of death and disability in people up to the age of 40. Early detection and prompt treatment is vital to save lives and minimise risk of disability, according to the NICE guidelines of Head injury: assessment and early management. Head CT scans are the gold standard for diagnosing these and it is critical that these are performed and reported by Radiologists in line with NICE guidelines.

The potential applications of AI in radiology go well beyond image analysis for diagnostic and prognostic opportunities. It is becoming increasingly clear that AI algorithms have the potential to improve productivity, operational efficiency, and accuracy in diagnostic radiology. AI tools are being developed to aide diagnosis and enhance processes at multiple point in the radiology workflow including:

(a) protocolling the prioritised scan,(b) clinical decision support systems for detection of critical findings, (c) worklist priority adjustment via AI results, and (d) reducing turnaround time through worklist prioritisation and semiautomated structures reporting. The adoption of AI tools is dependent on the demonstration of a tangible effect on patient care and improvement in radiologist workflow.

Thus, in this study, we aim to assess whether real-world implementation of an AI tool which augments (b), (c) and (d) of the imaging life cycle would affect turnaround times.

qER medical device:

qER, a CE Class II approved medical software device, detects, and localizes the presence of six target abnormalities - intracranial haemorrhage, cranial fracture, midline shift, mass effect, atrophy and hypodensities suggestive of infarcts in non-contrast Head-CT scans. A priority status is assigned if any one of the target abnormalities (intracranial haemorrhage, cranial fracture, midline shift or mass effect) is detected by the software, and the user will be able to view a single summary slice listing all the target abnormalities found by qER on the CT scan followed by all slices in scan with the overlay of above abnormalities localization. Alternately, if none of the target abnormalities are detected, the output will indicate that the software has analysed the image and identified no critical findings. qER reports are intended to support certified radiologists and/or licensed medical practitioners for clinical decision making. It is a support tool and, when used with original scans, can assist the clinician to improve efficiency, accuracy, and turnaround time in reading head CTs. It is not to be used to provide medical advice, determine treatment plan, or recommend a course of action to the patient.

Study design:

A multi-centre stepped wedged cluster randomised study will be conducted in 4 NHS hospitals over a 13-month period. Hospitals will be identified and initiated into the qER solution with a 30-day implementation period. The order in which sites will receive the qER intervention will be determined by computer-based randomisation. The stepped wedge design allows delivery of the intervention at an organisational level with evaluation of outcome measures at a patient level. Structuring the implementation through a staged activation in a random order provides important methodological advantages for both qualitative and quantitative elements of the study. The design allows control of adoption bias and adjust for time-based changes in the background patient characteristics at a patient level.

All patients under this pathway would receive an AI reading, and no additional or different tests will be performed as a result of the AI findings. The turnaround time will be the interval between the time the scan was taken to the time when the final scan report becomes available and will be measured in minutes. When qER assistance is used for reporting Head-CT scans and if there is a difference between the output of the qER and the radiologist, the latter will be considered as final for further patient management.

Primary objective:

The primary objective is to assess if qER based reporting and triage significantly reduce turnaround time (TAT) of critical NCCT head reporting for patients attending the emergency department.

Secondary objective(s):

• To assess utility of qER to support emergency department pathways for patients requiring NCCT head and radiology reporting workflow.
• To assess the safety of qER at identifying patients with critical findings on NCCT heads.
• To evaluate the technical performance of qER.
• To conduct a Heath Economic, cost utility analysis of qER.