Challenge Motivation


Intracerebral aneurysms are found in 3% of the general population, and some groups have a higher risk. If an aneurysm ruptures it causes bleeding in the brain (subarachnoid haemorrhage). [1]  Early detection of intracranial aneurysms, as well as accurate measurement and assessment of shape, is important in clinical routine. This enables careful monitoring of the growth and rupture risk of aneurysms to allow informed treatment decisions to be made [2]. Currently, contrast-enhanced computed tomography angiography scans (CTA) and non-contrast 3D time-of-flight magnetic resonance angiography (TOF-MRA) are the most common imaging techniques for this purpose. 

However, intracranial aneurysm detection and measurement can sometimes be difficult – especially for small aneurysms [1]. It has been cited that about 10% of the aneurysms, mostly small ones, are still missed [3]. For small aneurysms (<5mm) it has been reported that detection by radiologists from MRAs can have a sensitivity as low as 35% [4]. 

Increased knowledge on risk factors for aneurysm presence, such a positive family history for the disease, has led to more individuals being preventively screened with MRA [5]. With more patients being screened, it is becoming important to reduce the clinical workflow duration, whilst still allowing the accurate detection and diagnosis of an aneurysm. Automatic methods of detection of aneurysms from TOF-MRAs would allow the speed of clinical workflow to be increased, without compromising accuracy. 

Furthermore, automated volumetric segmentation would enable more reliable measurements and characteristics of aneurysms to be derived and considered for rupture risk prediction. For example, it is known that shape characteristics such as non-spherical and lobular shape are associated with elevated rupture risk [6, 7, 8]. Based on these shape features, and the associated rupture risk, a more informed treatment decision can be made. Shape of an unruptured intracranial aneurysm can also have an effect on the treatment outcome of a patient. Shape features of the aneurysms, derived from volumetric segmentations, could further aid treatment complication prediction models [9].

Technical Point of View

Various different (semi-) automatic methods for the detection and segmentation of intracranial aneurysms exist [10, 11].  Many detection methods are developed for CTA or Digital Subtraction Angiography (DSA) 2D images [12, 13]. However, in the clinic, MRI is best suited for regular follow-up as it requires neither intravenous contrast agent nor radiation. In addition, some treated (e.g. coiled) aneurysms can create large artefacts on CTA, so it is often necessary to assess for recanalization on MRA without artefacts. As TOF-MRA is increasingly used in clinical routine, characterisation and rupture risk assessment of aneurysms for MRA are becoming more important [14]. Hence, there is a need for accurate detection and segmentation methods from TOF-MRA. Aneurysms can be small, have very different shapes and occur at many different locations. In addition, fusiform widening of branching vessels can mimic small aneurysms. This leads to an exciting technical challenge to automatically detect and segment aneurysms, and includes generating creative and novel methods for medical image segmentation.


The purpose of this challenge is to automatically detect and segment intracranial aneurysms from TOF-MRA images. Automatic detection can aid a radiologist in diagnosis of intracranial aneurysms and will likely speed up the clinical workflow. Volumetric segmentation allows analysis of the size and shape of the aneurysms which may provide new biomarkers for use in rupture risk prediction models. Eventually, this may result in more informed decisions being made with regard to treatment of intracranial aneurysms.

Full challenge design document can be found here:


  1. A. Keedy, “An overview of intracranial aneurysms,” McGill Journal of Medicine, vol. 9, no. 2. pp. 141–146, 2006.
  2. J. M. Wardlaw and P. M. White, “The detection and management of unruptured intracranial aneurysms,” Brain, vol. 123, no. 2, pp. 205–221, 2000.
  3. P. M. White, J. M. Wardlaw, and V. Easton, “Can noninvasive imaging accurately depict intracranial aneurysms? A systematic review,” Radiology, vol. 217, no. 2, pp. 361–370, 2000.
  4. P. M. White, E. M. Teasdale, J. M. Wardlaw, and V. Easton, “Intracranial aneurysms: CT angiography and MR angiography for detection – Prospective blinded comparison in a large patient cohort,” Radiology, vol. 219, no. 3, pp. 739–749, 2001.
  5. M. J. H. Wermer, I. C. van der Schaaf, A. Algra, and G. J. E. Rinkel, “Risk of rupture of unruptured intracranial aneurysms in relation to patient and aneurysm characteristics: an updated meta-analysis.,” Stroke, vol. 38, no. 4, pp. 1404–10, 2007.
  6. D. Backes et al., “ELAPSS score for prediction of risk of growth of unruptured intracranial aneurysms,” Neurology, vol. 88, no. 17, pp. 1600–1606, 2017.
  7. A. E. Lindgren et al., “Irregular Shape of Intracranial Aneurysm Indicates Rupture Risk Irrespective of Size in a Population-Based Cohort,” Stroke, vol. 47, no. 5, pp. 1219–1226, 2016.
  8. M. L. Raghavan, B. Ma, and R. E. Harbaugh, “Quantified aneurysm shape and rupture risk,” J. Neurosurg., vol. 102, no. 2, pp. 355–362, 2009.
  9. W. Ji et al., “Risk score for neurological complications after endovascular treatment of unruptured intracranial aneurysms,” Stroke, vol. 47, no. 4, pp. 971–978, 2016.
  10. C. M. Hentschke, O. Beuing, R. Nickl, and K. D. Tönnies, “Automatic cerebral aneurysm detection in multimodal angiographic images,” IEEE Nucl. Sci. Symp. Conf. Rec., no. October, pp. 3116–3120, 2012.
  11. H. Arimura et al., “Computerized detection of intracranial aneurysms for three-dimensional MR angiography: Feature extraction of small protrusions based on a shape-based difference image technique,” Med. Phys., vol. 33, no. 2, pp. 394–401, 2006.
  12. H. Duan, Y. Huang, L. Liu, H. Dai, L. Chen, and L. Zhou, “Automatic detection on intracranial aneurysm from digital subtraction angiography with cascade convolutional neural networks,” Biomed. Eng. Online, vol. 18, no. 1, p. 110, 2019.
  13. N. Sulayman, M. Al-Mawaldi, and Q. Kanafani, “Semi-automatic detection and segmentation algorithm of saccular aneurysms in 2D cerebral DSA images,” Egypt. J. Radiol. Nucl. Med., 2016.
  14. A. Lane, P. Vivian, and A. Coulthard, “Magnetic resonance angiography or digital subtraction catheter angiography for follow-up of coiled aneurysms: Do we need both?,” J. Med. Imaging Radiat. Oncol., vol. 59, no. 2, pp. 163–169, 2015.