Abdullah Al et al. evaluated the performance of colonial walk, a lightweight machine learning method, in localizing aortic root landmarks on CT angiographies (CCTA) for transcatheter aortic valve implantation (TAVI) planning. Colonial walk uses a two-phased approach, estimating the aortic valve region and subsequently refining the localization of each landmark (three hinges, three commissures, and two coronary ostia) with a regression tree-based multi-walker search. Using 71 CCTA scans (40 non-TAVI and 31 pre-operative TAVI with heavy calcification), two experiments were performed: cross validation on non-TAVI scans with TAVI scans and cross validation on all scans. Colonial walk’s average landmark error was 2.04 ± 1.11 mm overall, with higher accuracy than the previous random tree walk (RTW) method (p < 0.05). Annulus and ostium diameter errors were below 2mm for TAVI and non-TAVI scans. Processing time on a single-core GPU was 12 ms per scan. Colonial walk outperforms RTW and is faster than manual analysis of CCTAs in clinical settings.