Registration of Untracked 2D Laparoscopic Ultrasound to CT Images of the Liver Using Multi-Labelled Content-Based Image Retrieval

Laparoscopic Ultrasound (LUS) is recommended as a standard-of-care when performing laparoscopic liver resections it images sub-surface structures such tumours and major vessels. Given that LUS probes are difficult to handle some iso-echoic, registration of pre-operative CT has been proposed an image-guidance method. This problem particularly challenging due the small field view LUS, usually depends on both manual initialisation tracking compose volume, hindering clinical translation. In this paper, we extend approach using Content-Based Image Retrieval (CBIR), removing requirement for or initialisation. Pre-operatively, set possible planes simulated from descriptor generated each image. Then, Bayesian framework employed estimate most likely sequence simulations matches series images. We our CBIR formulation use multiple labelled objects constrain by separating vessels into portal vein hepatic branches. The value new labeled demonstrated in retrospective data 5 patients. Results show that, including untracked time, single image can be registered with accuracies ranging 5.7 16.4 mm success rate 78%. Initialisation could potentially enable translation these fusion techniques.