CyclopsStrokeQuantifier - Ischaemic Stroke Detection System Using Dynamic CT Aldo von Wangenheim The Cyclops Project, CS Dept., Universidade Federal de Santa Catarina - UFSC Florianopolis - SC - BRAZIL - http://www.cyclops.lisha.ufsc.br
[email protected] Rafael Charnovscki The Cyclops Project, CS Dept., Universidade do Extremo Sul Catarinense - UNESC Av. Universitaria, 1105 - Bairro Universitario - C.P. 3167 - CEP: 88806-000
[email protected] Ronie C.F. Cardoso Telemedicine Lab, University Hospital, UFSC
[email protected] Luiz Felipe de Souza Nobre Internal Medicine Department, Universidade Federal de Santa Catarina (UFSC)
[email protected] Daniel Chaves Radiological Hospital DMI, Sao Jose - SC, BRAZIL
[email protected] Eros Comunello The Cyclops Project, Knowledge-Based Systems Group, Universität Kaiserslautern, Germany
[email protected] Abstract Ischaemic stroke is a very common disease that affects blood vessels in the brain causing cerebral tissue damage. Despite several studies on this subject, the early detection of acute ischaemic stroke is difficult as well its treatment. In this work we present some features and initial results of our research, which includes a new image acquisition protocol and a software tool in order to help physicians to improve the accuracy of the therapy. Keywords: acute ischaemic stroke, ischaemic stroke quantization, penumbra zone, dynamic CT.
1. Problem description The term ischaemic stroke refers to sudden endangering of the cerebral functions due to a variety of histopathological alterations involving one or more, intra or extra-cranial, blood vessels Although we dont't have precise statistics for Brazil, it is known that the mortality rate amongst patients with stroke is 0.5 to 1 within 1,000 individuals. Moreover, stroke is the third main cause of death related to clinical pathologies and the second more frequent cause of neurological morbidity. In the USA stroke is responsible for 150,000 deaths and invalidity of 200,000 people annually. Yet there are estimates about 2 billion people suffering from stroke's after-effects and expenses of 30 billion each year because of this disease[10].
Despite those numbers, physicians in general have little interest in the disease and, as a consequence,little knowledge about it. Most times doctors are not skilled to provide the correct attention[14]. About 80% of the stroke cases are caused by insufficient blood flow (ischaemic stroke)[2], and of these, 75% are due thrombosis and blockage[4]. When the blood flow, in a specific cerebral region, falls bellow of 20ml/100g/min, the cerebral electric activity fails and the symptoms appear. If the blood flow keeps falling to 10ml/100g/min, an ischaemic central area arises resulting in the death of the neurons. Around this central there is a region where the blood flow is reduced, however the perfusion is kept by the collateral circulation. These neurons that don't function but are alive form the penumbra zone. The penumbra has no metabolic activity but this area may be recovered with prompt medical intervention. If the therapy is not started as soon as possible, the cerebral infarct in the central area extends to the margins within a time not exactly known for human patients. The axial computerized tomography (CT) is an important examination and the first one to be requested, since it allow us to differentiate an infarct from an haemorrhage[7]. Once the ischaemic stroke is identified, the therapy can be performed with the utilization of neuroprotector agents or thrombolytics as an attempt to restore the blood flow in the obstructed region. This type of therapy has the purpose to restore the flow as soon as possible in order to limit the neuronal losses in the penumbra area. Currently, these are some of the limitations for adopting this therapy with thrombolytics: • Difficulty for radiologists to identify ischaemic stroke within six hours after onset using only visual examination of standard CT images[16]; • The visualization of signs of ischaemia within this window of time in a CT is itself an independent factor of badly prognostic[17]; • Difficulty for identification and measuring the penumbra zone, which would be the potentially recoverable region. There are several studies to develop early ischaemic stroke detection methods[7]. Although these methods which use magnetic resonance imaging (MRI) are more sensitive, they are difficult to perform in large scale due constraints like examination time and high costs. The specification of a methodology to perform CT examinations supported by a software tool that helps physicians to detect and quantify the exact extension of an ischaemic stroke and to identify penumbra zones, will provide benefits to the therapy, since several works point towards better results when the thrombolytic therapy is started within three hours after the first symptoms.
2. Objectives This work is the first stage of a whole project for development of a radiological protocol and a decision support software that quantifies the penumbra zone and helps physicians to decide whether the thrombolytics therapy is suitable or not. Specific objectives are as follows: • Calculation and graphical representation of the blood flow in the brain based on dynamic CT. For this task we have been using the knowledge previously acquired with the development of the Mammalyzer system [9]; • Using computer analysis for early detection of ischaemic stroke; • Identifying low blood flow regions (penumbra zone) that could be potentially recoverable by applying the therapy;
• Measuring the size of the penumbra zone and comparing it to a future dynamic TC of the same patient; • Validating the software prototype with help of physicians; • Integrating this software with Cerebral Atlas System[15] in order to make correlations between the ischaemic area and the neurologic functions related to that area.
3. Methods To develop the first software prototype a group of 4 patients have been used. Three patients were suspected to have ischaemic stroke and one patient was healthy. They have been submitted previously to an axial CT with slices of 1 cm thick in order to exclude the cases with haemorrhagic stroke. The brain area chosen to acquire the images is the region of the medium cerebral artery, since it is important for the blood flow supply. Immediately after the patients have taken the contrast agent, dynamic CT images with 5 mm thick have been acquired (the first slice is considered as a native slice). The results obtained with the software analysis were compared to the patients' anamnese and to the blood flow quantification from the healthy patient. Firstly, we developed a DICOM-compatible software module to load the CTs from a radiological image database. This software has an interface which allow us to visualize and to start all the calculations related to perfusion maps and signal variation through a certain time. For the calculation of the signal variation and construction of the perfusion maps it is initially calculated the percentual signal raise curve relative to the same point on the native slice (slice number 1) for each image region corresponding to the cerebral parenchima. The image regions showing cerebral parenchima are automatically chosen by the software, depending on both its radiological density and characteristics of its neighbourhood. The contrast agent absorption curves (percentual signal raise curves) are stored to generate perfusion maps on each slice and they can be visualized by simply moving (dragging) the mouse on any slice. The perfusion maps for each slice are then further calculated through the comparison of the relative contrast agent absorption variations, previously calculated for each point of that slice, with the mean contrast agent absorption in a volume of reference (VOR) manually selected by the user with the mouse. Thus, the user determines himself/herself any region in the slice to represent the cerebral area which is chosen to be a reference parameter (VOR) to the brain as a whole. After the manual selection of a VOR region, the system calculates for all slices the curve that represents the relative mean contrast agent absorption passing through the VOR. This curve, called the curve of reference (COR), will finally be used as a reference for the calculation of the perfusion maps for each slice. To create the perfusion maps, it is calculated the percentual difference (positive or negative) between the COR and the percentual signal raise curves on each pixel on the slice. Thus, the perfusion map shows the relative difference between the signal raise in the VOR (represented by the COR) and the signal raise of the selected slice. This procedure permits the construction of a comparative map between a contrast agent perfusion taken as a reference for that patient and all pixels of the parenchima. This methodology also allow us to: • choose a healthy cerebral area of the patient as a reference, in order to take account of his/her own characteristics as age, health status and other information; • create several perfusion maps based on different brain regions to compare the contrast agent absorption to the typical absorption of a cortex area or an area with some patology.
4. Results The results obtained so far are promising since the tests for the 4 patients are consistent. The software hasn't found any suspicious area of stroke in the healthy patient examination. However, the system has found areas with low or no perfusion in the patients suspected to have ischaemic stroke. For these patients the software marked low perfusion areas, which are related to the symptoms presented at the anamnese. At this moment there is a first version of the software being validated at the medical partners of our project and a large-scale study has been started. This largescale study is being started through the acquisition of a reference dataset of dynamic CTs acquired from a control group that is being selected among students of the UFSC. Figure 1 is a screenshot of the software prototype's GUI (Graphical User Interface) showing some images which had been acquired from a patient suspected to have ischaemic stroke.
Figure 1: System's user interface showing a) tomography slice number 15, b) signal variation curve related to the VOR (green line) compared to the other volume chosen (yellow line), c) perfusion map on slice number 16 showing perfusion in the brain, and d) VOR limits (green). Notice in the perfusion map (c) that the areas turning to purple show higher perfusion areas than the VOR marked as green (d). Yellow, orange and red areas indicate lower perfusion.
5. Conclusions A treatment for the ischaemic stroke has been difficult due to the lack of suitable techniques that permit precocious identification. The accuracy of the treatment depends on the penumbra zone identification, however there hasn't been an adequate protocol to locate and measure it. Nowadays the patients with potentially recoverable cerebral areas aren't treated, so that they may have a permanent cerebral lesion.
Throughout this first stage of the work we have considered that the understanding about the behavior of the penumbra zone is one of the most important goals. Besides providing a diagnostic tool with utmost importance, this work aims to add important and substantial information in the clinical routine associated to the manipulation of patients with acute ischaemic stroke. Furthermore, it will be possible to personalize the treatment if we can differentiate reversible ischaemic regions from those not reversible[8] so that we may recommend or not the therapy.
6. References [1] BARBER P.A.; DARBY, D.G.; DESMOND, P.M.; et al. Prediction of stroke outcome with echoplanar perfusion- and diffusion-weighted magnetic resonance imaging. Neurology, 51:418-426, 1998. [2] BENNETT, J.C.; PLUM, F.; et al. Cecil Txtbook of Medicine, 20th edition. Bennett & Plum, Philadelphia, pp 2271 e 2272, 1996. [3] CENIC, C.; NABAVI, D.G.; CRAEN, R.A.; GELB, A.W.; LEE, T. Dynamic CT Measurement of Cerebral Bloob Flow: A Validation Study. Am. J. Neuroradiol., 20:63-73, 1999. [4] FIESCHI, C.; ARGENTINO, C.; LENZI, G.L.; SACCHETTI, M.L.; TONI, D.; BOZZAO L. Clinical and instrumental evaluation of patients with ischemic stroke within the first 6 hours. J. Neurol. Sci., 91:311-321, 1989. [5] HACKE, W.; KASTE, M.; FIESCHI, C.; TONI, D.; LESAFFRE, E.; VON KUMMER, R.; BOYSEN, G.; BLUHMKI E.; HOXTER, G.; MAHAGNE, M.H.; et al. Intravenous thrombolysis with recombinant tissue plasminogen activator for acute hemispheric stroke. The European Cooperative Acute Stroke Study ECASS. J. Am. Med. Assoc., 274(13):1017-25, 1995. [6] HACKE, W.; KASTE, M.; FIESCHI, C.; TONI, D.; LESAFFRE, E.; VON KUMMER, R.; et al. Randomised double-blind placebo-controlled thrombolytic therapy with intravenous alteplase in acute ischemic stroke (ECASSII). Lancet, 352:1245-1251, 1998. [7] HOGGARD, N.; WILKINSON, I.D.; GRIFFITHS, P.D. The imaging of ischemic stroke. Clinical Radiology, 56:171-183, 2001. [8] KLOTZ, E.; KÖNIG, M. Perfusion measurements of the brain: using dynamic CT for the quantitative assessment of cerebral ischemia in acute stroke. European Journal of Radiology, 30:170-184, 1999. [9] KRECHEL, D.; MAXIMINI, R.; WILLE, P. R.; STEIL, M.; Von WANGENHEIM, A. Approaches for MRMammography Matching Algorithms. In: MRM 2000 - SECOND INTERNATIONAL CONGRESS ON MR-MAMMOGRAPHY, 2000, Jena. MRM 2000 - Second International Congress on MR-Mammography. Jena: Klinikum der Friedrich-Schiller-Universität Jena, 2000. v.1. p.215-219. [10] MATCHAR, D.B.; DUNCAN, P.W. Cost of stoke. Stroke Clinical Updates, 5:9-12, 1994. [11] MAYER, T.E.; HAMANN, J.B.; ROSENGARTEN, B.; KLOTZ, E.; WIESMANN, M.; MISSLER, U.; SCHULTE-ALTEDORNEBURG, G.; BRUECKMANN, H.J. Dynamic CT Perfusion Imaging of Acute Stroke. Am. J. Neuroradiol., 21:1441-1449, 2000. [12] NATIONAL INSTITUTE OF NEUROLOGICAL DISORDERS AND STROKE, rt-PA Stroke Study Group. Tissue plasminogen activator for acute ischemic stroke. New Engl. J. Med., 333:1581-1587, 1995. [13] REICHENBACH, J.R.; RÖTHER, J.; JONETZ-MENTZEL, L.; et al. Acute Stroke Evaluated by Time-toPeak Mapping during Initial and Early Follow-up Perfusion CT Studies. Am. J. Neuroradiol., 20:1842-1850, 1999. [14] TOOLE, J.F.; Cerebrovascular Disorders; fourth edition; New York; Raven Press, 129-150, 1990.
[15] WAGNER, H. M. Atlas Cerebral Digital: Desenvolvimento de uma Ferramenta Computacional para Mapeamento Funcional e Anatomico de Áreas Cerebrais, Baseado no Atlas de Talairach. 2001. 77f.. Master Thesis in Computer Science - Departamento de Informatica e Estatistica, Universidade Federal de Santa Catarina, Florianopolis. [16] WARDLAW, J.M.; DORMAN, P.J.; LEWIS, S.C.; SANDERCOCK, P.A.G. Can stroke physicians and neuroradiologists identify signs of early cerebral infarction on CT? Journal of Neurology, Neurosurgery & Psychiatry, 67(5):651-653, 1999. [17] WARDLAW, J.M.; LEWIS, S.C.; DENNIS, M.S.; COUNSELL, C.; McDOWAL, M. Is visible infarction on computed tomography associated with an adverse prognosis in acute ischemic stroke? Stroke , 29:1315-9, 1998. [18] WINTERMARK, M.; MAEDER, F.; VERDUN, F.R.; et al. Using 80 kVp versus 120 kVp in Perfusion CT Measurement of Regional Cereral Blood Flow. Am. J. Neuroradiol., 21:1881-1884, 2000.
