Computational Approaches for Acute Traumatic Brain Injury Image Recognition

Abstract

In recent years, there have been major advances in deep learning algorithms for image recognition in traumatic brain injury (TBI). Interest in this area has increased due to the potential for greater objectivity, reduced interpretation times and, ultimately, higher accuracy. Triage algorithms that can re-order radiological reading queues have been developed, using classification to prioritize exams with suspected critical findings. Localization models move a step further to capture more granular information such as the location and, in some cases, size and subtype, of intracranial hematomas that could aid in neurosurgical management decisions. In addition to the potential to improve the clinical management of TBI patients, the use of algorithms for the interpretation of medical images may play a transformative role in enabling the integration of medical images into precision medicine. Acute TBI is one practical example that can illustrate the application of deep learning to medical imaging. This review provides an overview of computational approaches that have been proposed for the detection and characterization of acute TBI imaging abnormalities, including intracranial hemorrhage, skull fractures, intracranial mass effect, and stroke.

Key Findings on Computed Tomography of the Head that Predict Death or the Need for Neurosurgical Intervention From Traumatic Brain Injury

“…15 Machine learning models for CTH interpretation may currently be useful for triage purposes, where abnormal findings can be brought to the attention of radiologists for definitive diagnosis. 16 Future innovations may significantly expand the capabilities of automated CTH interpretation, but the simplicity of our current schema allows for rapid manual interpretation of CTH findings by clinicians.…”

“…Various algorithms have been developed that are successful in identifying certain abnormalities, including hemorrhage, hematoma volume, midline shift, and localization of abnormal findings. 15,16 Although these algorithms are powerful, it is difficult to design algorithms that can identify the wide range of abnormal findings that can be seen on CTH after TBI. 15 Machine learning models for CTH interpretation may currently be useful for triage purposes, where abnormal findings can be brought to the attention of radiologists for definitive diagnosis.…”

Artificial Intelligence and Medicine

“…With the advancement of medical imaging technology and the enhancement of clinical diagnosis precision, DL-based clinical diagnosis approaches have been intensively developed. A DL-based brain CT image processing system was introduced to achieve automatic identification of acute neurological events such as stroke (24,25). Zhu et al suggested an automatic diagnosis approach for ischemic stroke based on DL, with a sensitivity of 76.9%, a specificity of 84.0%, and an accuracy of 80.5%, which can offer doctors with acute ischemic stroke (26).…”

Head Injury and Intracranial Hemorrhages: A CT Imaging Study of Trauma Patients

“…In the majority of studies, 0.5-1% of patients with mild TBI who CTH have traumatic intracranial findings, and 5-10% of these patients also require neurosurgery (11). However, mild TBI is a hazy condition that the Glasgow Coma Scale can variously classify (GCS) 13-15, GCS 14-15, or GCS 15, and it can happen with or without loss of consciousness or retrograde amnesia, which has complicated research on it in the context of diagnostic evaluation (12).…”

“…Although these are significant factors, they shouldn't be overemphasized because a single CTH has significantly less radiation exposure than most other computed tomography scans (13). Additionally, unrecognized intracranial hematoma continues to have significant clinical and legal ramifications for emergency medicine practitioners (12). This last factor is likely responsible for the growing use of CTH because no decision tool has ever demonstrated 100% sensitivity across a population tested (14).…”