Automated Detection of Subarachnoid Hemorrhage on CT Head Examinations Utilizing a ResNet Convolutional Neural Network Methodology

Congress:

ECR 2019

Poster Number:

C-3184

Type:

Scientific Exhibit

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

DOI-Link:

https://dx.doi.org/10.26044/ecr2019/C-3184

Fig. 1: Table distribution of Case and Control images between Training and Test Sets

Fig. 2: Subarachnoid hemorrhage example: High-Frequency Salience maps are key

Fig. 3: Grad-CAM Visualization Is Inappropriate for SAH Localization

Fig. 4: False-positive Salience Map activation over normal structures

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...

Methods and materials

Study Population: A retrospective case-control review was performed of all adult CT head examinations on PACS from January 2006 to December 2017 at two academic tertiary care hospitals. In order for a case to be included, single-slice images had to contain a single focus of SAH, to aid in focusing training on a single finding, given the weak localization labelling provided. Images containing multi-compartmental bleeds were excluded as these were hypothesized to possibly compromise training of the network. Control images were derived from independent CT...

Results

Objective Evaluation of Network Performance: 1005 control and 193 case slice images met inclusion/exclusion criteria (Fig. 1). After 5 runs of 200 epochs of training, training accuracy averaged 0.893 with average validation accuracy equaling 0.853. Multiple cross-fold training and validation runs produced validation accuracy values ranging from 0.828 - 0.877. Average training time per epoch was 36 seconds on a single Titan X Pascal (Nvidia, Santa Clara, California). Average time to classification inference per CT head volume was under 3 seconds. Average time to the...

Conclusion

Study Limitations: Both training and validation were limited by the relative low quantity of positive training images as well as the large number of discarded training cases due to multi-compartmental bleed. In general, as with previous studies, we found the relative value of reported ROC accuracy figures to be overly optimistic with respect to accuracy with variable performance when attempting subjective visual validation. Because of the low dataset size, alternate split designs and validation strategies such as cross-validation had to be employed to obtain a...

Personal information

References

[1] Macdonald RL, Schweizer TA. Spontaneous subarachnoid haemorrhage (2017). Lancet.11;389(10069):655-666. [2] Nieuwkamp DJ, Setz LE, Algra A, Linn FH, de Rooij NK, Rinkel GJ (2009). Changes in case fatality of aneurysmal subarachnoid haemorrhage over time, according to age, sex, and region: a meta-analysis. Lancet Neurol. 8(7):635-42. [3] Lawton MT, Vates GE (2017). Subarachnoid Hemorrhage. N Engl J Med. 377(3):257-266. [4] Singer RJ, Ogilvy CS, Rordorf G. Aneurysmal subarachnoid hemorrhage: Epidemiology, risk factors, and pathogenesis. In: Biller J, ed. UpToDate 2018. [5] Frontera J, Claassen J,...