Keywords:
Artificial Intelligence, CNS, CT, Computer Applications-Detection, diagnosis, Computer Applications-General, Trauma, Aneurysms
Authors:
W. Ding, J. W. Luo, J. J. R. Chong; Montreal, QC/CA
DOI:
10.26044/ecr2019/C-3184
Aims and objectives
Spontaneous subarachnoid hemorrhage (SAH) can be a significant complication of trauma or aneurysm rupture that affects approximately 50,000 people in North America each year [1],
among which there is an expected 30% mortality [2] and 50% development of long-term disabilities [3].
Because SAH affects a relatively younger population,
on average individuals aged 40-60 years old [4],
it results in a disproportionate effect with significant loss of productive life years [3].
The clinical management of SAH can be challenging: The initial diagnosis requires timely delivery and accurate imaging assessment,
as the severity of hemorrhage on the initial CT scan is the most important prognostic factor for outcome [5].
The sensitivity of CT decreases rapidly with time,
from nearly 100% within 6 hours to 50% at 5 days [1].
If a patient is delayed in presentation to the hospital,
or if the hemorrhage at onset is subtle,
correct initial CT assessment may not be delivered on-time,
especially in hospitals with large volumes of head CT scans after-hours when the workforce is reduced and scans are assessed by non-neuroradiology experts or trainees.
Thus,
an automated analytical tool able to assess and stratify SAH based on initial routine CT imaging would provide important assistance to the timely detection of SAH and would help prioritize radiology workflow thus reducing the time to treatment and the risk of developing subsequent complications.
Recent advances in Machine Learning and Convolutional Neural Networks (CNN) may aid in achieving this goal.
In a recent study by Lee et al.
[6],
hemorrhagic strokes were detectable with the ResNet CNN architecture [7],
using less than 1000 head CT scan images for training and validation.
Since SAH CT scans have similar graphic properties to hemorrhagic stroke,
we expect that SAH stratification can also be performed using the ResNet-50 architecture and propose to develop and train a network based upon 1000 clinical cases from our local institution.
Our hypothesis is the ResNet-50 will able to perform the detection,
localization,
and quantification of SAH based on non-enhanced brain CT images,
and successfully distinguish SAH from other forms of intracranial hemorrhage.