Advancing edema detection: Harnessing the power of machine learning and near infrared spectroscopy for cerebral and cerebellar edema assessment

Abstract

Edema, characterized by brain swelling, is a common response observed in various brain injuries. Timely detection of edema is crucial to mitigate the associated risks and improve patient care. This study evaluates the efficacy of CEREBO®, a non-invasive machine learning-powered near-infrared spectroscopy (mNIRS) based device, in detecting edema. The study was conducted on 234 participants with suspected head injuries who underwent simultaneous CEREBO® scans and CT head scans. The results of the study showed that CEREBO® effectively identified edematous lobes, achieving a sensitivity of 95.7%, specificity of 97%, and accuracy of 96.9% for cases with intracranial hemorrhage (ICH). Additionally, for cases without ICH, the device exhibited a sensitivity of 100%, specificity of 97.2%, and accuracy of 97.2%. Two cases were reported where CEREBO® failed to detect edematous ICH. The study highlights the potential of CEREBO® as a valuable tool for early detection of pre-symptomatic edema and ICH, enabling timely interventions and improved patient care. The findings support the reliability of near-infrared spectroscopy as a diagnostic modality for edema.

Keywords: CEREBO®; Cerebellar Edema; Cerebral edema; Near infrared spectroscopy; Non-invasive detection.