Matthias Seise, Abstract Co-Author: Nothing to Disclose

Marina Kolesnik, Presenter: Nothing to Disclose

Mika Pollari MSc, Abstract Co-Author: Nothing to Disclose

Sonja Koestenbauer, Abstract Co-Author: Nothing to Disclose

Ursula Mayrhauser, Abstract Co-Author: Nothing to Disclose

Philipp Stiegler, Abstract Co-Author: Nothing to Disclose

Horst R. Portugaller MD, Abstract Co-Author: Nothing to Disclose

We presented a processing pipeline for accurate fusion of histological and CT data and, ultimately, for assessing the true correspondence between histological lesion and lesion visible in CT after RFA. Tested in pig experiments the pipeline satisfied the clinical precision target of 5 mm which is the typical safety margin around the lesion induced by the RFA intervention. Thus the pipeline is suitable for validation of the RFA treatment.

Histological validation of a RFA-induced lesion provides ground-truth about the true lesion size, thus benchmark verification of success or failure of the treatment. In the liver, lesions can be highly irregular due to large blood vessels acting as heat sinks, which make assessment based on simple descriptors such as volume, elongation or single slices highly uncertain. The IMPPACT project (www.imppact.eu, funded by the European Commission under FP7 ICT Programme) has developed a processing pipeline for establishing accurate correspondence between the histological lesion and lesion visible in CT as used for RFA guiding.

For evaluation, all steps of the computational pipeline were performed twice by independent observers. The precision of histology lesion reconstruction and registration into CT was evaluated as distance between lesion contours. On average it was well below 2mm with a maximum of 5mm. For comparison, the precision of manually segmented CT lesions was 0.40 mm (SD: 0.42 mm, MAX: 4.01 mm). The distance between CT and histology lesion was 0.85 mm (SD: 1.02 mm, MAX: 5.46 mm).

Data and histology preparation was designed specifically for accurate reconstruction of the histological volume and registration into the CT data. Both were solely based on segmented vessels to avoid introducing correlation between lesion features. High-resolution CT (MicroCT) of the histology block before slicing was used to correctly reconstruct the histology lesion with surrounding vessels from large scale histology slides. Furthermore, in situ perfusion fixation of the liver ensured minimal deformation of the liver tissue before removal enabling accurate registration of reconstructed histology and CT.

Seise, M, Kolesnik, M, Pollari, M, Koestenbauer, S, Mayrhauser, U, Stiegler, P, Portugaller, H, Framework for Histological Validation of Radiofrequency Ablation (RFA) in Porcine Liver.  Radiological Society of North America 2011 Scientific Assembly and Annual Meeting, November 26 - December 2, 2011 ,Chicago IL. http://archive.rsna.org/2011/11008890.html