Fully automated detection and segmentation of intracranial aneurysms in subarachnoid hemorrhage on CTA using deep learning

Rahil Shahzad,Jan Borggrefe,David Maintz,Marc Schlamann,Christoph Kabbasch,Boris Krischek,Lukas Goertz,Lenhard Pennig,Michael Perkuhn,Frank Thiele

doi:10.1038/s41598-020-78384-1

Abstract

In aneurysmal subarachnoid hemorrhage (aSAH), accurate diagnosis of aneurysm is essential for subsequent treatment to prevent rebleeding. However, aneurysm detection proves to be challenging and time-consuming. The purpose of this study was to develop and evaluate a deep learning model (DLM) to automatically detect and segment aneurysms in patients with aSAH on computed tomography angiography. In this retrospective single-center study, three different DLMs were trained on 68 patients with 79 aneurysms treated for aSAH (2016–2017) using five-fold-cross-validation. Their outputs were combined to a single DLM via ensemble-learning. The DLM was evaluated on an independent test set consisting of 185 patients with 215 aneurysms (2010–2015). Independent manual segmentations of aneurysms in a 3D voxel-wise manner by two readers (neurosurgeon, radiologist) provided the reference standard. For aneurysms > 30 mm3 (mean diameter of ~ 4 mm) on the test set, the DLM provided a detection sensitivity of 87% with false positives (FPs)/scan of 0.42. Automatic segmentations achieved a median dice similarity coefficient (DSC) of 0.80 compared to the reference standard. Aneurysm location (anterior vs. posterior circulation; P = .07) and bleeding severity (Fisher grade ≤ 3 vs. 4; P = .33) did not impede detection sensitivity or segmentation performance. For aneurysms > 100 mm3 (mean diameter of ~ 6 mm), a sensitivity of 96% with DSC of 0.87 and FPs/scan of 0.14 were obtained. In the present study, we demonstrate that the proposed DLM detects and segments aneurysms > 30 mm3 in patients with aSAH with high sensitivity independent of cerebral circulation and bleeding severity while producing FP findings of less than one per scan. Hence, the DLM can potentially assist treating physicians in aSAH by providing automated detection and segmentations of aneurysms.

Highlights

In aneurysmal subarachnoid hemorrhage, accurate diagnosis of aneurysm is essential for subsequent treatment to prevent rebleeding
CT-angiography (CTA) is performed immediately upon radiological proof of SAH with sensitivity rates for detection of aneurysms ranging between 85–98% compared to digital subtraction angiography (DSA), which is considered as the gold standard for aneurysm imaging[5,6]
The development of a deep learning model (DLM) to automatically detect and segment intracranial aneurysms would be of valuable assistance to the radiologist

Summary

Introduction

In aneurysmal subarachnoid hemorrhage (aSAH), accurate diagnosis of aneurysm is essential for subsequent treatment to prevent rebleeding. The purpose of this study was to develop and evaluate a deep learning model (DLM) to automatically detect and segment aneurysms in patients with aSAH on computed tomography angiography. In this retrospective single-center study, three different DLMs were trained on 68 patients with 79 aneurysms treated for aSAH (2016–2017) using five-fold-cross-validation. We demonstrate that the proposed DLM detects and segments aneurysms > 30 mm[3] in patients with aSAH with high sensitivity independent of cerebral circulation and bleeding severity while producing FP findings of less than one per scan. Abbreviations ACA Anterior cerebral artery aSAH Aneurysmal subarachnoid hemorrhage CNN Convolutional neural network CTA CT-angiography DSA Digital subtraction angiography DSC Dice similarity coefficient DLM Deep learning model. This is of particular interest due to the growing workload and consequent fatigue of radiologists, which correlates with increased risk to miss relevant findings[16]

Objectives

Methods

Discussion

Conclusion