文献速递鼻科

面部识别软件显著提升计算机辅助导航精度

Accuracy of computer-assisted navigation: significant augmentation by facial recognition software.
PubMed 文献 · Int Forum Allergy Rhinol · 2017-07-14

面部识别在中鼻甲头、中鼻甲腋和蝶窦前壁的平均误差分别为0.47 mm、0.33 mm和0.39 mm,且在这三个位点均显著优于表面配准;配准时间未受不利影响。

要点

  • 面部识别配准在中鼻甲头、中鼻甲腋和蝶窦前壁的精度显著优于传统表面配准。
  • 面部识别配准的平均误差在中鼻甲头为0.47 mm、中鼻甲腋为0.33 mm、蝶窦前壁为0.39 mm。
  • 整合面部识别软件未对配准时间产生不利影响。

核心结论

  • 在这个前瞻性研究中,自动面部识别软件与传统表面配准相比显著提高了图像导航配准的精度。

主要发现

  • 面部识别在中鼻甲头(MTH)平均误差0.47 mm,在中鼻甲腋(MTA)0.33 mm,在蝶窦前壁(SS)0.39 mm,在鼻尖(NT)0.36 mm。
  • 面部识别在中鼻甲头比表面配准平均精确0.43 mm(p = 0.002)。
  • 面部识别在中鼻甲腋比表面配准平均精确0.44 mm(p < 0.001)。
  • 面部识别在蝶窦前壁比表面配准平均精确0.40 mm(p < 0.001)。
  • 整合面部识别软件未对配准时间产生不利影响。

关键数值与推荐

  • 面部识别配准精度:中鼻甲头平均0.47 mm,中鼻甲腋0.33 mm,蝶窦前壁0.39 mm,鼻尖0.36 mm。
  • 面部识别较表面配准的精度提升:中鼻甲头平均0.43 mm(p = 0.002),中鼻甲腋平均0.44 mm(p < 0.001),蝶窦前壁平均0.40 mm(p < 0.001)。

临床怎么用

  • 在鼻窦和颅底手术中,可考虑使用面部识别软件进行图像导航配准以提高精度,且不延长配准时间。

局限性

  • 原文未述局限性

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

BACKGROUND: Over the past 20 years, image guidance navigation has been used with increasing frequency as an adjunct during sinus and skull base surgery. These devices commonly utilize surface registration, where varying pressure of the registration probe and loss of contact with the face during the skin tracing process can lead to registration inaccuracies, and the number of registration points incorporated is necessarily limited. The aim of this study was to evaluate the use of novel facial recognition software for image guidance registration. METHODS: Consecutive adults undergoing endoscopic sinus surgery (ESS) were prospectively studied. Patients underwent image guidance registration via both conventional surface registration and facial recognition software. The accuracy of both registration processes were measured at the head of the middle turbinate (MTH), middle turbinate axilla (MTA), anterior wall of sphenoid sinus (SS), and nasal tip (NT). RESULTS: Forty-five patients were included in this investigation. Facial recognition was accurate to within a mean of 0.47 mm at the MTH, 0.33 mm at the MTA, 0.39 mm at the SS, and 0.36 mm at the NT. Facial recognition was more accurate than surface registration at the MTH by an average of 0.43 mm (p = 0.002), at the MTA by an average of 0.44 mm (p < 0.001), and at the SS by an average of 0.40 mm (p < 0.001). The integration of facial recognition software did not adversely affect registration time. CONCLUSION: In this prospective study, automated facial recognition software significantly improved the accuracy of image guidance registration when compared to conventional surface registration.

来源:Int Forum Allergy Rhinol,查看原文。版权归原作者所有。

引用

  1. Glicksman JT, Reger C, Parasher AK, et al. Accuracy of computer-assisted navigation: significant augmentation by facial recognition software. Int Forum Allergy Rhinol. 2017. PMID: 28707807. doi:10.1002/alr.21982

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。