Study on the Prognostic Prediction Model of Patients With Acute Intracerebral Hemorrhage by Artificial Intelligence

Capital's Funds for Health Improvement and Research

Spontaneous intracerebral hemorrhage(SICH) is the most lethal and disabling stroke. Timely and accurate assessment of patient prognosis could facilitate clinical decision making and stratified management of patients and is important for improving patient clinical prognosis. However, current studies on the prediction of prognosis of patients with SICH are limited and only include a single variable, with less precise results and inconvenient clinical application, which may lead to delays in effective patient treatment. Our group's previous studies on SICH showed that hematoma heterogeneity and the degree of contrast extravasation within the hematoma are closely related to the clinical outcome of patients, but they are difficult to describe quantitatively based on imaging signs. Based on this, we propose to use radiomics to quantitatively extract hematoma features from NCCT and CTA images, combine them with patients' clinical information and laboratory tests, study their relationship with the prognosis of cerebral hemorrhage, and use artificial intelligence to establish a rapid and accurate prognostic prediction model for patients with SICH, which is of great significance to guide clinical individualized treatment.

Inclusion Criteria:

• 1. aged 18-80 years; 2. patients diagnosed with acute cerebral hemorrhage by CT examination; 3. complete non-contrast CT and CTA images; 4. the time interval from onset to the first baseline CT and CTA examination is less than 6 hours; 5. follow-up data within 3 months; 6. agree and sign a written document.

Exclusion Criteria:

• 1. Patients with secondary aneurysm hemorrhage; 2. Patients with secondary hemorrhage of cerebrovascular malformation; 3. Patients with dissecting aneurysm hemorrhage; 4. Patients with cerebral infarction hemorrhage transformation; 5. Patients lost to follow-up within 3 months; 6. CT or CTA images have a heavy artefact.