Reliability of automated software to assign e-ASPECTS to CT scans for acute ischemic changes (S8.006)
2017
Objective: To evaluate the reliability of automated software to evaluate CT scans for acute ischemic changes (AIS) when compared to human readers with varying levels of experience. Background: Ischemic stroke is a leading cause of morbidity and mortality. Currently, computed tomography (CT) is used for assessment of patients presenting with symptoms of acute ischemic stroke. ASPECT scoring is a topographic 10-point scoring system to rate early ischemic changes and is correlated to outcome after stroke. ASPECT rating requires significant expertise and is typically only performed in tertiary stroke centres. New software designed to automatically assign e-ASPECT scores to CT scans has been developed (Brainomix Limited). Design/Methods: 165 CT scans with corresponding diffusion weighted (DWI) MRI within 60 minutes were selected from the ethics approved KSS database. These scans were evaluated by physicians with various levels of experience (from resident to stroke fellow) in assigning ASPECTS for AIS. Subjects were not aware of the affected side. The same scans were also processed using the Brainomix software to determine the e-ASPECTS, without specifying the affected side. MR DWI ASPECTS was obtained by a 2-physician consensus reading; performance was evaluated by comparing CT ASPECTS to MRI ASPECTS. Results: e-ASPECTS reliably predicted the affected side (Kappa 0.57). Trichotomized ASPECTS revealed that e-ASPECTS software performed similarly to physicians with greater experience (e-ASPECTS Kappa 0.65). Region wide analysis demonstrated that e-ASPECTS performed as well as the physician readers (e-ASPECTS Kappa 0.54 (0.49, 0.58) Conclusions: e-ASPECTS was able to assign ASPECTS scores to CT scans with a reliability similar to the physician readers. There was moderate agreement for both physician readers and e-ASPECTS compared to MRI. Disclosure: Dr. Hoyte has nothing to disclose. Dr. Al Sultan has nothing to disclose. Dr. Finkelstein has nothing to disclose. Dr. Boyko has nothing to disclose. Dr. Fok has nothing to disclose. Dr. Pordeli has nothing to disclose. Dr. Horn has nothing to disclose. Dr. Neweduk has nothing to disclose. Dr. Yu has nothing to disclose. Dr. Appireddy has nothing to disclose. Dr. Tham has nothing to disclose. Dr. Menon has nothing to disclose.
Keywords:
- Correction
- Source
- Cite
- Save
- Machine Reading By IdeaReader
0
References
0
Citations
NaN
KQI