Deep Learning for Quantitative Assessment of Acute Ischemic Stroke on Non-Contrast CT

• Applicant: Dr. Sophie Ostmeier
• Subject Area: Clinical Neurology; Neurosurgery and Neuroradiology
• Term: since 2023
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 517316550

A non-contrast Computed Tomography (NCCT) is the first and sometimes only imaging modality to guide the treatment decision of acute stroke in the US. It is fast, cheap, widely available and requires no contrast agent. The goal of the project is the quantification of acute ischemic stroke on NCCT with deep learning to provide information about the stroke volume and location. For the stroke patients in hospitals without advanced stroke imaging (ca. 60% of stroke patients), this will improve the triage of patients for highly effective endovasculare therapy (thrombectomy). Aim 1: To determine a clinical meaningful metric to evaluate the performance of a predesigned deep learning algorithm for acute ischemic stroke. Aim 2: To optimize the deep learning algorithm for the segmentation of acute ischemic stroke on NCCT Aim 3: To modify the deep learning algorithm to incorporates labels from three separate neuroradiologists to reduce uncertainty in the label Aim 4: To investigate correlation of the algorithm predictions for ischemic injury to the final ischemic core volume

• DFG Programme: WBP Fellowship
• International Connection: USA
• Host: Professor Jeremy Josef Heit, Ph.D.