Visual understanding of liver vessels anatomy between living donor-recipient (LDR) pair can assist surgeons to optimize transplant planning by avoiding non-targeted arteries, which may cause severe complications. We presented a new method for LDR liver pair matching. In our approach, liver vessel systems were segmented from CTA volumes using incremental concept of cascade feature mapping as an optimal alternative for adapting new training data. The segmented vessels were then mapped to tree representation using ternary tree data structure based on (Michel’s X types) hepatic artery and portal vein variants. Matching process efficiently employed ternary tree in-order traversing to find appropriate match between given recipient against its possible donor variants. In liver transplantation problems to maintain diversity of vessels structure, our constructed model is adaptive for new data. Experimental analysis shows that newly added input helps to maximize the chance of finding an appropriate match, which adds up to our accuracy level in terms of increasing the success rate.
Our workflow comprises three major steps. Firstly, liver vessels segmentation is carried out using CTA volumes as an input to the CIL model. Secondly, segmented vessels are mapped to a tree representation via a well-known hepatic artery and portal vein variants. Finally, tree matching step compares the recipient tree with possible available donor tree variants to find the best match. There is a clear clinical motivation of this study, which is to provide surgeons with a comprehensive pre-operative evaluation of the liver vascular system to understand the variations of LDR vascular anatomy. Such evaluations can reduce the risks to donors, e.g., toxic exposure of liver parenchyma, and maximize the benefits to recipient, e.g., to avoid non-optimal pairing (recipient rejection). Clinically, such solutions will provide surgeons comprehensive guidance for pre-operative evaluation of liver vascular system to understand the variations of LDR vascular anatomy.