Early evaluation and prompt treatment are key elements for the correct management of acute ischemic stroke (AIS). The early identification of an accurate prognosis including advanced neuroimaging currently lacks a posterior circulation clot burden score (pc-CBS), and a multimodal prediction tool of posterior circulation stroke treated with endovascular treatment (EVT).
The project will be structured in a two-phase analysis using as derivation population the ASTRAL acute stroke databank of the Lausanne University Hospital: (1) development of a pc-CBS correlating with functional long-term outcome; and (2) development of a pre-EVT and post-EVT outcome prediction nomogram for patients treated up to 24 hours for posterior circulation AIS. Development of the pc-CBS and of the prognostic tools will compare the performance of classic logistic regression with two machine learning algorithms (xgboost and neural network). Both phases will undergo external validation with pooled data from other stroke centers.
We expect to find a good performance of the pc-CBS and of the nomograms thanks to the combination of multimodal neuroimaging variables with clinical and procedural variables; machine learning may further improve the score’s performance, and external validation will contribute to its validity.
The project has the impact to develop clinical tools able to support difficult acute clinical decisions including EVT. If externally validated, these tools may also be useful for patient selection in clinical trials.