增强现实
医学
乳房再造术
腹壁下动脉穿支皮瓣
计算机视觉
投影(关系代数)
人工智能
虚拟现实
计算机图形学(图像)
计算机科学
算法
乳腺癌
癌症
内科学
作者
David J Cholok,Marc Fischer,Christoph W Leuze,Michael Januszyk,Bruce L. Daniel,Arash Momeni
标识
DOI:10.1097/prs.0000000000010592
摘要
Autologous breast reconstruction yields improved long-term aesthetic results but requires increased resources of practitioners and hospital systems. Innovations in radiographic imaging have been used increasingly to improve the efficiency and success of free-flap harvest. Augmented reality (AR) affords the opportunity to superimpose relevant imaging on a surgeon's native field of view, potentially facilitating dissection of anatomically variable structures. To validate the spatial fidelity of AR projections of deep inferior epigastric perforator flap (DIEP) relevant anatomy, comparisons of 3D models and their virtual renderings were performed by four independent observers. Measured discrepancies between the real and holographic models were evaluated.3D-printed models of DIEP relevant anatomy were fabricated from CTA data from 19 de-identified patients. The corresponding CTA data was similarly formatted for the Microsoft Hololens to generate corresponding projections. Anatomic points were initially measured on 3D models, after which, the corresponding points were measured on the Hololens projections from two separate vantages. Statistical analyses, including Generalized Linear Modeling, were performed to characterize spatial fidelity regarding translation, rotation, and scale of holographic projections.Amongst all participants, the median translational displacement at corresponding points was 9.0 mm, 12.1 mm, and 13.5 mm between the real 3D model and V1, 3D model and V2, and between V1 and V2, respectively.Corresponding points, including topography of perforating vessels for the purposes of breast reconstruction can be identified within millimeters, but there remain multiple independent contributors of error, most notably the participant and location at which the projection is perceived.
科研通智能强力驱动
Strongly Powered by AbleSci AI