O29	THE PHYSICS OF EMBOLUS DETECTION BY DOPPLER ULTRASOUND David H Evans Division of Medical Physics, Leicester Warwick Medical School, Leicester, UK Transcranial Doppler Ultrasound is now widely used for the study of cerebral embolism. Like most Doppler ultrasound based techniques however the correct interpretation of the results requires a good understanding of the basic physics of the method. The basic concept of detecting emboli is quite straightforward – if the back-scattered ultrasound power from the moving embolus is significantly higher than the back-scattered power from the surrounding moving blood, then the presence of the embolus can be detected. Arising from this simple principle there are two important questions: 1) Under what circumstances will the additional power from the embolus be sufficient to be detected? 2) If an embolus is detected can we tell something about its size and composition? The detectability of an embolus is determined by a multitude of factors such as the size and composition of the embolus, the ultrasound frequency, the size of the Doppler sample volume, the embolus trajectory and its interaction with the ultrasound beam, filter settings, the point during the cardiac cycle when the embolus traverses the sample volume, the characteristics of the Doppler device, and the compromise that is reached concerning desired sensitivity and specificity. In general even relatively small gas bubbles will be detected, but some larger solid emboli may be missed, and this is perhaps one of the ironies of embolus detection, that it is the potentially more dangerous solid emboli that are most likely to be missed. With regard to size and composition, several techniques have been suggested as being useful for characterising composition, and whilst in general considerable progress has been made in this direction there are still considerable challenges in distinguishing between large particulate emboli and small gaseous emboli, partly because of the same issues that affect detectability. Unfortunately the sizing of emboli remains a distant goal. These various issues, and the potential for significant improvements, are much easier to understand when explained in terms of the relevant physics.
O30	LESSONS FROM RIGHT-TO-LEFT-SHUNT Gian Paolo Anzola MD, Servizio di Neurologia, Ospedale S. Orsola FBF, Brescia – ITALY In recent years , following the surge of interest for patent foramen ovale (PFO) in cryptogenic stroke, the presence of a right-to-left shunt (RLS) through the atrial chambers has been extensively studied in a number of conditions not or marginally related to cerebrovascular disease. Historically the first studies addressed the presence of RLS in scuba divers as a possible abnormality related to decompression sickness of unknown aetiology. Despite initial dispute, there is now robust evidence to claim that patency of foramen ovale increases the risk of developing decompression sickness by 2.5 times (Bove 1999). Patients with PFO-related decompression sickness tend to have early occurrence of symptoms after surfacing and a clinical presentation that indicates brain or upper cervical spinal cord involvement (Germonpré et al., 1998, Wilmshurst et al., 2000). The pathophysiology of PFO-related decompression sickness is likely to derive from paradoxical embolism of nitrogen gas bubbles to the brain or the upper spinal cord (Ries et al., 1999). It has recently been reported that divers with haemodynamically significant RLS may have an increased risk of developing clinically asymptomatic multiple brain lesions (Knaut et al., 1997). However, prospective studies are required to assess whether PFO is to be considered a contraindication to scuba diving. PFO has been found in patients suffering from migraine with aura with approximately the same frequency as that encountered in cryptogenic stroke patients (Del Sette et al., 1998; Anzola et al., 1999). This finding has prompted speculations on the possible role of RLS in increasing the stroke risk in migraineurs and in the pathophysiology of the aura. A recent report showing that migraine with aura attacks are dramatically improved after transcatheter closure of PFO suggests that migraine with aura may indeed be triggered by humoral factors that reach the brain by escaping the pulmonary filter (Wilmshurst et al., 2000). Valsalva-like activities often precede the occurrence of attacks of transient global amnesia (TGA) and abnormalities consistent with hypoperfusion of deep limbic structures have been reported during a typical TGA episode. This had raised the hypothesis that TGA may be triggered by paradoxical embolism of platelets aggregates in the posterior circulation, but the search for an increased frequency of PFO in TGA patients has yielded conflicting results. In the first study, PFO was found in 55% of TGA patients, in 26% of stroke patients and in 26% of controls (Kloetzsch et al., 1966). However, we have been unable to replicate this finding in a cohort of 46 consecutive patients in whom the prevalence of PFO (19.5%) was no higher than that found in 33 age-matched control subjects ( 19,6 %) and nonsignificantly lower than in 31 TIA patients (33.3%) (Anzola et al., 2000). Methodological differences between the two studies can partly explain the discrepancy, but it is likely that inclusion of patients with associated confounding conditions ( migraine, chronic obstructive pulmonary disease ) may have increased the occurrence of PFO in TGA patients in the German study. A further study with very strict inclusion criteria which take into account all possible confounding variables could hopefully clarify this issue. Conditions that determine an increase in pulmonary pressure may facilitate the opening of the virtual interatrial valve and thus promoting shunting of blood to the left heart chambers which in turn might contribute to further desaturation of arterial blood. It is therefore not surprising that RLS has been found in 70% of patients with chronic obstructive pulmonary disease and increased pulmonary pressure (Soliman et al., 1999) and in the same proportion of patients with obstructive sleep apnoea, a condition that ultimately may result in pulmonary hypertension (Shanoudy e al., 1998), although the latter has been disputed (Angeli et al. 1999). In conclusion, from the evidence gathered so far the picture is emerging of an important role of PFO in a number of non-stroke conditions, either as causative factor or as associated condition predisposing to complications. The availability of simple diagnostic techniques such as transcranial Doppler to assess right-to-left shunt will undoubtedly contribute a great deal of knowledge on the relevance in medicine of this hitherto neglected condition.
O31	FUNDAMENTAL FEATURES OF MICROEMBOLIC TRACKS ON POWER M-MODE DOPPLER (PMD) FOR AUTOMATIC EMBOLUS DETECTION MA Moehring Spencer Technologies, Seattle, USA Background: Transcranial power m-mode Doppler (PMD) observes microembolic tracks (ET) and blood flow signatures with depth on the vertical axis, time on the horizontal, and power showing as intensity colored red or blue according to mean velocity. The aim of this investigation is to determine essential characteristics of PMD ET, for automatically detecting microemboli. Method: A 2 MHz PMD system (Spencer Technologies TCD100M) having 33 sample gates placed at 2 mm intervals from 23 to 87mm depth was used for this work with 8kHz PRF, 7cm/s clutter rejection and 6mm sample volume size. Doppler power versus depth was constructed every 4ms, using all 33 sample gates. 300 ET signals were collected from a subject with two mechanical valves, utilizing a variety of beam alignments with basal cerebral vessels. Results: The hallmark of ET signatures on PMD is a continuous track with intensity that rises minmally 3 to 6dB above the background blood flow signal, with length at least twice the Doppler sample volume length. An ET has a sloping appearance, with postive slope when directed toward the probe and negative away. An ET can have direction contrary to the surrounding blood flow if the blood in which it moves is a minority contributor to the mean velocity. An ET can produce mulitple spatially contiguous tracks if it stops and then continues on at a later time, or moves into a branch vessel while in the ultrasound beam. ET thickness in the depth dimension is proportional to sample volume size. Conclusions: A new standard for automatic detection of microemboli using PMD and not using spectral information will include embolus track amplitude of 3 to 6 dB above the background, track length greater than twice the sample volume size, confirmation of slope, and track thickness on the order of the sample volume size. Collection of emboli from carotid lesions is underway to groom this approach. This work supported by NIH Grant 2R44HL57108-02.
O32	EVALUATION OF THROMBOLYSIS WITH TRANSCRANIAL ULTRASONICATION USING A RAT AUTOLOGOUS THROMBOEMBOLISM MODEL T. Saguchi ^1, 2, T. Ishibashi ^1, M. Akiyama ^2, H. Onoue ^2, T. Abe ^2, H. Furuhata ^1 Medical Engineering Laboratory^1, Department of Neurosurgery^2 , Jikei University School of Medicine, Tokyo, Japan Background: We are developing intracranial thrombolytic therapy by maens of transcranial ultrasonication (TUS). The efficacy of thrombolysis with transcranial ultrasonication in a rat autologous thromboembolism model was investigated. Methods: Twenty-four adult male Wistar rats weighing 300-400 g were used. The animal model used was previously reported by Zhang et al. in 1997. Animals were assigned to three groups: 1) non-therapy group, 2) trombolytic therapy without TUS group (intravenous administration of 1.2 mg tissue plasminogen activator (tPA)), and 3) thrombolytic therapy with TUS group. We selected a frequency of 488.2 kHz for TUS. In order to avoid tissue heating, ultrasonication (US) was applied in 2 minutes cycles, with a pause of 30 seconds over a period of 10 minutes. US was not applied for 5 minutes. These 10-minute sessions, which followed the 5-minutes breaks, were repeated four times. The total intermittent application time of US was 32 minutes. Twenty-four hours after embolization, the rats were deeply anesthetized and killed by transcardial perfusion with 50ml of heparinized (5U/ml) saline and 10% bufferd formaline, and then their brains were carefully removed. Using a rat brain matrix, each brain was cut into 2-mm thick coronal blocks, for a total of seven blocks per animal. The specimens were stained with hematoxylin and eosin. The difference in an area of cerebral infarction was examined among the above-mentioned three groups. Results: There was a significant difference between the non-therapy group and the thrombolytic therapy group in the area of cerebral infarction, and a tendency for the area of cerebral infarction to be smaller in the thrombolytic therapy with TUS group than that of the thrombolytic therapy without TUS group. Conclusions: The thrapeutic efficacy of thrombolytic therapy with TUS was confirmed in this experiment using a rat autologous thromboembolism model.
O33	NEW ULTRASONIC TRANSDUCER FOR CHARACTERIZATION OF MICROEMBOLI P.Palanchon^1,2, J. Klein^2 and N. de Jong^1 Erasmus MC, Rotterdam The Netherlands Dijkzigt Academish Ziekenhuis,Rotterdam, The Netherlands Background: High harmonic components contained in the frequency spectrum of the echo scattered by a microemboli has been recently proposed as a way to differentiate solid emboli from gaseous emboli. This technique, based on RF signals, requires the use of large frequency band transducer to detect the nonlinear high frequency components scattered by gas bubbles. Method: In this study, we introduce a new transducer design with separate transmitting and receiving frequency bands. This transducer is composed of a PZT ceramic as the transmitting part emitting at a frequency f0 of 500kHz. A PVDF layer is glued on the top of the ceramic and used for the reception of frequencies ranging from 50kHz up to 20MHz. This new transducer was used to characterize gaseous and particulate emboli. The measurements were performed using gas bubbles diameters ranging from 15mm up to 100mm and solid particles between 50mm and 500mm. Results: Depending on their size, gaseous emboli are leading to the formation of a subharmonic (f0/2), ultraharmonic (3f0/2) or higher harmonics (2f0, 3f0) components in the frequency spectrum while solid particles and other bubble sizes behave only linearly. Conclusions: Overall, this study shows the feasibility of such a transducer design to characterize selective bubble size from other particles.In the future, such a transducer could be used for clinical diagnosis.
O34	MICROEMBOLIC SIGNAL PREDICTS RECURRENT STROKE IN ACUTE STROKE PATIENTS WITH MIDDLE CEREBRAL ARTERY STENOSIS Ka Sing Wong, Shan Gao, Tjark Hansberg, Yu Leung Chan, Wynnie W.M. Lam, Dirk W. Droste, Richard Kay, E. Bernd Ringelstein. The Chinese University of Hong Kong, Hong Kong SAR and University of Münster, Münster, Germany ABSTRACT Background Cerebral embolism is a common cause of stroke. Microembolic signals (MES) detected by transcranial Doppler ultrasound represent ongoing embolism but the lack of reliable data about its clinical relevance hinders its widespread use in clinical practice. Methods We prospectively monitored 114 consecutive acute ischemic stroke patients for MES over 30minutes, including 85 patients with symptomatic and 29 patients with asymptomatic middle cerebral artery (MCA) stenosis. The signals on digital audiotape were analysed by an independent observer who was blind of all other data. All patients were followed for the occurrence of recurrent stroke in the MCA territory. Results MES were found in 24 (28.2%) of 85 symptomatic patients, and in 1 (3.4%) of 29 asymptomatic patients (Pearson Chi-square P=0.005). The mean number of MES was 18 (range, 1-102). MES were more common in patients with severe stenosis (10/21, 48%) than in those with mild-moderate stenosis (4/26, 15%) (Pearson Chi-square P=0.02). There was no significant difference in stroke severity on admission between MES-positive and MES-negative patients (Pearson Chi-square P=0.79). During follow-up for a mean of 13.6 (range 1-32) months, the presence of MES was the only predictor of a further ischemic stroke, (Cox regression, adjusted OR 8.45; 95% CI 1.69-42.22; P=0.01) even after controlling for age, sex, diabetes, hypertension, previous stroke, smoking, acute treatment and degree of MCA stenosis. Conclusions MES predicts further cerebral ischemia in acute stroke patients with MCA stenosis. This procedure should become a routine investigation and might identify a group of patients who are most likely to benefit from antithrombotic treatment.
O35	Real-Time Identification AND archival of MicroemboliC EVENTS Using A KNOWLEDGE BASED System L. Fan*, D.H. Evans* and A.R. Naylor** Departments of *Medical Physics & **Surgery, University of Leicester, UK Background: Detection of microemboli in the cerebral circulation using Transcranial Doppler ultrasound provides valuable clinical information, but current subjective methods of embolic signal detection and analysis have significant drawbacks. The purpose of this project was to implement a reliable and accurate real-time system for detection and archival of embolic signals. Methods: An automatic knowledge based system was implemented using a PC and a digital signal processing board containing an ADSP-21065L processor (Analog Devices Inc.). Time, frequency and neighbourhood information is extracted in the FFT-based front-end, and information sent to a blackboard system before a decision is made to identify microemboli and reject artefacts in real-time. For each event detected, a one-second segment of raw signal, together with details of absolute time, signal-to-clutter-ratio, and frequency are saved for further verification. The system displays a sonogram corresponding to 3.2s of signal, and any signal segment within this period can be selected and displayed in the time domain without interrupting real-time detection and archival. A preliminary test has been conducted in real-time with 74 separate Doppler signal segments from 29 carotid endarterectomy patients and 4 normal volunteers. The data, with a total duration of 185 minutes and containing at least 575 artefacts and 253 seconds of diathermy interference, was analysed automatically using the system and also manually off-line by three human experts. Results: Using a majority voting method with a 7dB detection threshold (1151 candidate events and 495 ‘true’ embolic events), the automated system reached a sensitivity of 96.6% and a specificity of 96.0%, compared with values for the three human experts of: 97.0% and 97.4%, 94.8% and 99.4%, 98.9 and 99.1%. Conclusions: These preliminary results show that this new real-time knowledge based system has a sensitivity and specificity close to that of human experts, when they perform their analyses under ideal laboratory conditions.