gms | German Medical Science

Objective: We recently reported on intra-operative blood perfusion measurement based on indocyanine green (ICG) fluorescence. Before the method can be established in practice its validity has to be shown.

Methods: The algorithm, originally developed for ICG video perfusion (Viper), was run on raw CT-perfusion sequences of 10 patients (4 SAH, 4 infarctions, 1 trauma, 1 abscess). The sampling point for the arterial input function (AIF) was selected manually. The algorithm computed perfusion measures using a constrained algebraic approach (least-squares minimization, subject to the residue function being positive). The resulting images for the time to maximum of the residue function (Tmax) and for the mean transit time (MTT) were compared to those computed by standard CT-perfusion software, STROKETOOL-CT.

For comparison, the images produced by both algorithms were divided into eight brain regions: anterior, lateral, medial, and posterior, for left and right side, respectively. For each patient, region, and algorithm, we manually assessed whether it showed normal or disturbed blood flow. The theoretically expected number of differences in assessment between the two algorithms was modelled as a binomial distribution and statistically compared to the observed differences. Difference probabilities of 0.05 (“significant”) and 0.1 (“trend”) were used. On the 0.1 level, the expected number of differences was 13 and 7 on the 0.05 level.

Results: Tmax images did not differ between the two algorithms: We counted nine different assessments, all observed in two patients with infarction. For the remaining patients we had perfect matching. Concerning MTT images, 30 differences were counted, distributed over all diagnoses. Illustrative examples are show in Figures 1 and 2. MTT images showed similarities in shape, but ICG images are smoother, with moderate values, compared to the patchy CT-images with more extreme values.

Conclusion: For Tmax, which is the easiest to compute, the Viper algorithm for ICG perfusion produced essentially the same results as a standard CT-perfusion algorithm. Computation of MTT is significantly more complex, involving additional steps in which numerical errors due to noise are accumulated. The differences in MTT images are most likely due to different ways of dealing with these inaccuracies and need further investigation.

Figure 1 [Fig. 1]

Figure 2 [Fig. 2]