**5. Conclusions**

Computational Intelligence in Electromyography Analysis – 48 A Perspective on Current Applications and Future Challenges

with temporary functional compensation after MIS.

tracts during the progression of cerebrovascular disease.

The one-year follow-up established a statistically significant dynamics in the MEP and CMCT outcomes. During the acute phase of the ischaemic stroke in the symptomatic hemisphere we found prolonged CMCT and reduced MEP amplitudes, similar to the findings by other authors [Segura et al, 1990]. When abnormal MEP amplitude is found, it is most likely the case of mainly functional disturbances of the pyramidal tract conduction, with a pathogenesis due to the acute disorder of cerebral circulation [Braune et al, 1996]. This may probably explain the prolonged latencies of MEP and reduced MEP amplitudes in the asymptomatic hemisphere. The decrease of the CMCT and the significant increase of the MEP amplitude (without reaching the normal values) at month 3, an according to Hadjipetrova et al [1993] – even at the 20-th day after the index event – could be explained

According to some authors, the MEP amplitudes are a more sensitive marker than CMCT in the view of assessing the damage of the corticospinal tracts as a result of the brain ischaemia. After the acute phase, there might be a facillitation at cortical level, which could allow an increase in the MEP amplitude and may eventually explain the increases in the MEP amplitude at month 3 in our patients. In particular, the MEP amplitudes in the asymptomatic hemisphere achieve the normal values at month 3 and month 12. In the same time, it is known that the minor ischaemic strokes are most likely of lacunar type (i.e., "deep", "subcortical"). The proportion of cortical clinical syndromes in our patients is relatively small and we could hypothesize that the abnormal MEP amplitudes might be present also in distant ischaemic lesion as it was shown earlier [Laloux et al,

At month 12, even in patients without neurological deficit and without recurrent cerebrovascular accidents, the CMCT is increased in both hemispheres. MEP amplitudes at month 12 are also reduced in the view of the normal values. These later changes are most likely due to the appearance of new asymptomatic structural changes of the corticospinal

Last but not least, following the revealed correlations, we were also able to create predictive models for the outcomes at month 12. For both the CMCT and MEP amplitude, the regression models were based on the initial measures at month 1 and symptomatics (i.e., pertaining to symptomatic or asymptomatic hemisphere). We confirmed our results further by using such probabilistic approach as artificial neural networks modelling. ANN proved to be very useful in the current analysis as it allowed us to assess the role of potential predictors of CMCT and MEP at month 12 as continuous outcomes, without the possible constraints of parametric models (e.g., normal distribution of the outcome, etc.). To note, ANN had been successfully used in other medical fields [Mecocci, 2002; Grossi, 2011; Azarkhish et al, 2012; etc.] and, in neurology, in particular [Mecocci, 2002; Shanthi et al, 2009; for a recent overview see Atanassova &

**4. Discussion** 

1991].

Dimitrov, 2011].

This *post-hoc* analysis of one-year follow-up clinical trial data, obtained in 40 patients with minor ischaemic stroke, but without neurological deficit or recurrent cerebrovascular incidents, established a statistically significant dynamics in the MEP and CMCT outcomes after transcranial magnetic stimulation. During the acute phase of the ischaemic stroke (at day 7 in month 1) we performed an initial measurement on the motor cerebral cortex bilaterally and on C7 with consecutive conduction of MEPs by surface electrodes from isometrically slightly contracted muscle *abductor policis brevis* and found prolonged CMCT and reduced MEP amplitudes in the symptomatic hemisphere. By following consecutive measurements at the end of month 3 and month 12, we revealed that the CMCT was increased in both hemispheres and MEP amplitudes were reduced, thus both remaining with abnormal values. At the interim measurement, CMCT were shorter but still abnormal in both hemispheres while the MEP amplitudes were lower, mostly in the symptomatic hemisphere. The changes at the end of the follow-up are most likely due to the appearance of new asymptomatic structural changes of the corticospinal tracts during the progression of the cerebrovascular disease.

We observed a parallel dynamics and found correlations between CMCT and MEP at various times, preserving a significant asymmetry among the two hemisepheres. There was a statistically significant correlation between the initial values of CMCT and MEP and the outcome measurements at month 12. Thе parametric regression modelling indicated that CMCT outcomes at month 12 can be predicted by the initial values at month 1 and whether or not these have been observed in the symptomatic or asymptomatic hemisphere. The same is valid for the MEP amplitude outcomes at month 12, although the role of the symptomatics as a predictor is with a marginal statistical significance. The probabilistic ANN modelling confirmed the role of early CMCT (month 1) and hemisphere symptomatics in predicting the outcome at month 12. Given the dynamics of CMCT and MEP changes, we could postulate that cerebrovascular disease progression post-MIS may have most likely determined the subclinical damage of the pyramidal tract and its underlying mechanisms.
