Javier A Luzon1,2, Rahul P Kumar3, Bojan V Stimec4, Ole Jakob Elle3,5, Arne O Bakka6,7, Bjørn Edwin6,3,8, Dejan Ignjatovic6,7. 1. Institute of Clinical Medicine, Faculty of Medicine, University of Oslo, Oslo, Norway. jaluzon@gmail.com. 2. Division of Surgery, Department of Digestive Surgery, Akershus University Hospital, Lørenskog, Norway. jaluzon@gmail.com. 3. The Intervention Centre, Oslo University Hospital, Oslo, Norway. 4. Faculty of Medicine, Teaching Unit, Anatomy Sector, University of Geneva, Geneva, Switzerland. 5. Department of Informatics, Faculty of Mathematics and Natural Sciences, University of Oslo, Oslo, Norway. 6. Institute of Clinical Medicine, Faculty of Medicine, University of Oslo, Oslo, Norway. 7. Division of Surgery, Department of Digestive Surgery, Akershus University Hospital, Lørenskog, Norway. 8. Department of Hepatopancreatobiliary Surgery, Oslo University Hospital-Rikshospitalet, Oslo, Norway.
Abstract
BACKGROUND: 3D vascular anatomy roadmaps are currently being implemented for surgical planning and navigation. Quality of the reconstruction is critical. The aim of this article is to compare anatomical completeness of models produced by manual and semi-automatic segmentation. METHODS: CT-datasets from patients included in an ongoing trial, underwent 3D vascular reconstruction applying two different segmentation methods. This produced manually-segmented models (MSMs) and semi-automatically segmented models (SAMs) which underwent a paired comparison. Datasets were delivered for reconstruction in 4 batches of 6, of which only batch 4 contained patients with abnormal anatomy. Model completeness was assessed quantitatively using alignment and distance error indexes and qualitatively with systematic inspection. MSMs were the gold standard. Assessed vessels were those of interest to the surgeon performing D3-right colectomy. RESULTS: 24 CT-datasets (13 females, age 44-77) were used in a paired comparative analysis of 48 3D-models. Quantitatively, SAMs showed structural improvement from Batch 1 to 3. Batch 4, with abnormal vessels, showed the highest error-index values. Qualitatively, 91.7% of SAMs did not contain all mesenteric branches relevant to the surgeon. In SAMs, 1 (12.5%) right colic artery-RCA scored as a complete vessel. 3 (37.5%) RCAs scored as incomplete and 4 (50%) RCAs were absent. 6 (25%) of 24 middle colic arteries-MCA scored as complete vessels. 11 (45.8%) scored as incomplete while 7 (29.2%) MCAs were absent. 13 (54.2%) of 24 ileocolic arteries-ICA were complete vessels. 11 (45.8%) scored as incomplete. None (0%) were absent. Additionally, it was observed that 10 (41.7%) of SAMs contained all their jejunal arteries, when compared to MSMs. Calibers of "complete" vessels were significantly higher than in "missing" vessels (MCA p < 0.001, RCA p = 0.016, ICA p < 0.001, JAs p < 0.001). CONCLUSION: Despite acceptable results from quantitative analysis, qualitative comparison indicates that semi-automatically generated 3D-models of the central mesenteric vasculature could cause considerable confusion at surgery.
BACKGROUND: 3D vascular anatomy roadmaps are currently being implemented for surgical planning and navigation. Quality of the reconstruction is critical. The aim of this article is to compare anatomical completeness of models produced by manual and semi-automatic segmentation. METHODS: CT-datasets from patients included in an ongoing trial, underwent 3D vascular reconstruction applying two different segmentation methods. This produced manually-segmented models (MSMs) and semi-automatically segmented models (SAMs) which underwent a paired comparison. Datasets were delivered for reconstruction in 4 batches of 6, of which only batch 4 contained patients with abnormal anatomy. Model completeness was assessed quantitatively using alignment and distance error indexes and qualitatively with systematic inspection. MSMs were the gold standard. Assessed vessels were those of interest to the surgeon performing D3-right colectomy. RESULTS: 24 CT-datasets (13 females, age 44-77) were used in a paired comparative analysis of 48 3D-models. Quantitatively, SAMs showed structural improvement from Batch 1 to 3. Batch 4, with abnormal vessels, showed the highest error-index values. Qualitatively, 91.7% of SAMs did not contain all mesenteric branches relevant to the surgeon. In SAMs, 1 (12.5%) right colic artery-RCA scored as a complete vessel. 3 (37.5%) RCAs scored as incomplete and 4 (50%) RCAs were absent. 6 (25%) of 24 middle colic arteries-MCA scored as complete vessels. 11 (45.8%) scored as incomplete while 7 (29.2%) MCAs were absent. 13 (54.2%) of 24 ileocolic arteries-ICA were complete vessels. 11 (45.8%) scored as incomplete. None (0%) were absent. Additionally, it was observed that 10 (41.7%) of SAMs contained all their jejunal arteries, when compared to MSMs. Calibers of "complete" vessels were significantly higher than in "missing" vessels (MCA p < 0.001, RCA p = 0.016, ICA p < 0.001, JAs p < 0.001). CONCLUSION: Despite acceptable results from quantitative analysis, qualitative comparison indicates that semi-automatically generated 3D-models of the central mesenteric vasculature could cause considerable confusion at surgery.
Entities:
Keywords:
3D modeling; Colorectal surgery; Image-guided surgery; Patient-specific computational modeling; Personalized medicine
Authors: Javier A Luzon; Bojan V Stimec; Arne O Bakka; Bjørn Edwin; Dejan Ignjatovic Journal: Int J Comput Assist Radiol Surg Date: 2020-09-28 Impact factor: 2.924