MRI image analysis to assess stroke damage

Today in the lab, I spent the entire time doing image analysis, using the MRI scans collected during an experiment that was testing possible new treatments for stroke. The image below shows an example of those images and the type of analysis I was doing is described in more detail below. Basically, a lot of mathematics was required, although the process was semi-automated with the use of certain software specifically designed for this type of analysis. Still, it is a lengthy process, and it will probably take several days to complete the analysis for the entire data set. Regardless, I thoroughly enjoyed the process so far and it was very exciting getting familiar with the image analysis software.

This is an example of the image analysis I performed today in the lab. These are brain scans acquired 30 minutes and 7 days post-stroke, in which I manually delineated in red and white the areas of brain damage. Note: the three treatment types have been concealed.
This is an example of the image analysis I performed today in the lab. These are brain scans acquired 30 minutes and 7 days post-stroke, in which I manually delineated in red and white the areas of brain damage. Note: the three treatment types have been concealed.

Detailed methods of today's image analysis

(A little bit too heavy on the details, but worth a read if you are interested in the topic)

Initial infarct at 30 minutes after stroke was defined as the hypo-intense area on quantitative apparent diffusion coefficient (ADC) maps that were constructed from diffusion-weighted MRI images. This area on each slice was multiplied by the slice thickness, and all eight slices were summed to calculate the initial infarct volume for each brain. Additionally, final infarct 7 days after stroke was defined by manually delineating the hyperintense area on T2-weighted MRI images. Infarct area from each slice was multiplied by slice thickness and summed to calculate the lesion volume for each brain. Ipsilateral and contralateral hemispheric volumes were similarly defined by manually delineating the hemispheres on T2-weighted MRI images.

Ischaemic stroke results in the breakdown of the blood-brain barrier, leading to vasogenic brain oedema due to water influx into the damaged tissue (Ayata and Ropper, 2002). This causes an extension of the ischaemic brain lesion, which accounts for about 30% of the observed infarct volume (Gerriets et al., 2004). To prevent overestimation of infarct volume from MRI measurements, initial and final infarct volumes [equation (1)] were corrected for swelling of the ipsilateral hemisphere using equation (2), published in Swanson et al (1990); and for compression of the contralateral hemisphere using equation (3), published in Gerriets et al (2004).

Equations 1, 2, 3
Equations 1, 2, 3

Infarct volume as a percentage of the ipsilateral hemisphere (% hemispheric lesion volume, % HLV) was calculated using equation (4) published in Gerriets et al (2004), to allow comparison of different total brain volumes.

Equation 4
Equation 4

A little bit more background on the general experiment

A DWI scan was obtained at 30 minutes post-stroke to allow quantitative ADC map production and assessment of ischaemic tissue. Treatment was administered intracerebroventricularly via an osmotic pump inserted subcutaneously at the back of the neck. The pump was connected to a cannula inserted into the right lateral ventricle via a burr hole drilled in the contralateral skull. Recovery was allowed for seven days, then T2-weighted MRI images were obtained to assess final infarct volume.

Magnetic resonance imaging (MRI) is commonly used in the diagnosis and management of acute ischemic stroke. A good method of detecting acute ischaemia is by using diffusion weighted imaging (DWI), an MRI sequence that uses opposing magnetic field gradients to detect tissue in which water diffusion is restricted due to cytotoxic oedema (Fisher et al., 1992), which is evident in the acute phase of cerebral ischaemia. Cytotoxic oedema is the swelling of neurons, glia, and endothelial cells, due to the release of toxic factors after cerebral injury (Hemphill, Beal and Gress, 2001). Increased DWI signal in the brain is typically observed within a few minutes of stroke and leads to a reduction of the ADC value of the tissue, a quantitative measure of the degree of water restriction (Merino and Warach, 2010).

Another type of MRI, T2-weighted imaging is also used as a reliable measure of ischaemic lesion beyond the acute stages. T2-weighted images show areas of vasogenic oedema, signifying the volume of the infarct. Vasogenic oedema is the influx of solutes and fluids into the brain due to the breakdown of the blood-brain barrier (Hemphill, Beal and Gress, 2001).


  • Ayata, C. and Ropper, A.D. (2002). Ischaemic brain oedema. Journal of Clinical Neuroscience, 9, pp.113–124.
  • Fisher, M., Sotak, C.H., Minematsu, K., Li, L. (1992). New magnetic resonance techniques for evaluating cerebrovascular disease. Annals of Neurology, 32, pp.115-122.
  • Gerriets, T., Stolz, E., Walberer, M., Müller, C., Kluge, A., Bachmann, A., Fisher, M., Kaps, M., Bachmann, G. (2004). Noninvasive Quantification of Brain Edema and the Space-Occupying Effect in Rat Stroke Models Using Magnetic Resonance Imaging. Stroke, 35(2), pp.566-571.
  • Hemphill, J.C., Beal, M.F., Gress, D.R. (2001). Critical care in neurology. In: Braunwald, E., Fauci, A.S., Kasper, D.L., Hauser, S.L., Longo, D.L., Jameson, J.L., eds. Harrison's Principles of Internal Medicine. 15th ed. Mc Graw Hill: New York. pp. 2491-2498.
  • Merino, J.G. and Warach, S. (2010). Imaging of acute stroke. Nature Reviews Neurology, 6, pp.560-571.
  • Swanson, R.A., Morton, M.T., Tsao-Wu, G., Savalos, R.A., Davidson, C., Sharp, F.R. (1990). A semiautomated method for measuring brain infarct volume. Journal of Cerebral Blood Flow and Metabolism, 10, pp.290-293.