Real-time intraoperative depth estimation in transsphenoidal surgery using deep learning: A feasibility study

IF 1.8 4区医学 Q3 CLINICAL NEUROLOGY

Journal of Clinical Neuroscience Pub Date : 2026-05-01 Epub Date: 2026-02-14 DOI:10.1016/j.jocn.2026.111910

Olivier Zanier , Aron Alakmeh , Raffaele Da Mutten , Alessandro Carretta , Matteo Zoli , Diego Mazzatenta , Carlo Serra , Luca Regli , Victor E. Staartjes

{"title":"Real-time intraoperative depth estimation in transsphenoidal surgery using deep learning: A feasibility study","authors":"Olivier Zanier , Aron Alakmeh , Raffaele Da Mutten , Alessandro Carretta , Matteo Zoli , Diego Mazzatenta , Carlo Serra , Luca Regli , Victor E. Staartjes","doi":"10.1016/j.jocn.2026.111910","DOIUrl":null,"url":null,"abstract":"<div><h3>Purpose</h3><div>Endoscopic endonasal and transcranial approaches are used for the resection of various pathological lesions in neurosurgery, especially pituitary adenomas, craniopharyngiomas, chordomas, or meningiomas. The video feed provided by endoscopes is generally two-dimensional, which can hinder depth perception. Thus, generating three-dimensional imaging without the need for special endoscopes using deep learning might be beneficial for enhanced intraoperative orientation.</div></div><div><h3>Methods</h3><div>DINOv2 is a pre-trained deep-learning model published by Meta in 2023. One of its capabilities is to estimate the depth in two-dimensional images. In this study, we explore the application of DINOv2 to the video feed of eight transsphenoidal endonasal surgeries. The results were evaluated for quality by both a senior neurosurgeon and a resident neurosurgeon. Furthermore, depth estimations from a randomly selected subset of 488 images taken from the videos were semi-quantitatively compared against manual segmentations for the estimation of deep, intermediate, and superficial areas.</div></div><div><h3>Results</h3><div>Using DINOv2, numeric depth maps were generated, and colormaps were created for depth visualization. Although these colormaps were not perfect, they aligned well with the subjective assessment of depth in the video feed by a senior neurosurgeon as well as a resident neurosurgeon. Semi-quantitative validation of the model’s estimations yielded a mean overall DICE Similarity Index of 0.48. These semi-quantitative results should be interpreted with caution, as the cutoffs used for model depth predictions and manual segmentation are not standardized.</div></div><div><h3>Conclusions</h3><div>Through the application of DINOv2, we were able to estimate depth in endoscopic imaging from transsphenoidal endonasal surgeries by generating numeric maps and depth colormaps. This illustrates the potential of deep learning-based depth estimations, which in the future could contribute to improving intraoperative orientation. It also highlights the opportunities in using artificial intelligence to augment endoscopic video feeds.</div></div>","PeriodicalId":15487,"journal":{"name":"Journal of Clinical Neuroscience","volume":"147 ","pages":"Article 111910"},"PeriodicalIF":1.8000,"publicationDate":"2026-05-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Clinical Neuroscience","FirstCategoryId":"3","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0967586826000615","RegionNum":4,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"2026/2/14 0:00:00","PubModel":"Epub","JCR":"Q3","JCRName":"CLINICAL NEUROLOGY","Score":null,"Total":0}

引用次数: 0

Abstract

Purpose

Endoscopic endonasal and transcranial approaches are used for the resection of various pathological lesions in neurosurgery, especially pituitary adenomas, craniopharyngiomas, chordomas, or meningiomas. The video feed provided by endoscopes is generally two-dimensional, which can hinder depth perception. Thus, generating three-dimensional imaging without the need for special endoscopes using deep learning might be beneficial for enhanced intraoperative orientation.

Methods

DINOv2 is a pre-trained deep-learning model published by Meta in 2023. One of its capabilities is to estimate the depth in two-dimensional images. In this study, we explore the application of DINOv2 to the video feed of eight transsphenoidal endonasal surgeries. The results were evaluated for quality by both a senior neurosurgeon and a resident neurosurgeon. Furthermore, depth estimations from a randomly selected subset of 488 images taken from the videos were semi-quantitatively compared against manual segmentations for the estimation of deep, intermediate, and superficial areas.

Results

Using DINOv2, numeric depth maps were generated, and colormaps were created for depth visualization. Although these colormaps were not perfect, they aligned well with the subjective assessment of depth in the video feed by a senior neurosurgeon as well as a resident neurosurgeon. Semi-quantitative validation of the model’s estimations yielded a mean overall DICE Similarity Index of 0.48. These semi-quantitative results should be interpreted with caution, as the cutoffs used for model depth predictions and manual segmentation are not standardized.

Conclusions

Through the application of DINOv2, we were able to estimate depth in endoscopic imaging from transsphenoidal endonasal surgeries by generating numeric maps and depth colormaps. This illustrates the potential of deep learning-based depth estimations, which in the future could contribute to improving intraoperative orientation. It also highlights the opportunities in using artificial intelligence to augment endoscopic video feeds.

查看原文本刊更多论文

应用深度学习进行蝶窦手术术中实时深度估计的可行性研究

目的内镜下经鼻和经颅入路用于神经外科各种病理病变的切除，尤其是垂体腺瘤、颅咽管瘤、脊索瘤和脑膜瘤。内窥镜提供的视频馈送通常是二维的，这可能会阻碍深度感知。因此，无需特殊内窥镜而使用深度学习生成三维成像可能有助于增强术中定位。方法dinov2是Meta在2023年发表的预训练深度学习模型。它的功能之一是估计二维图像的深度。在本研究中，我们探讨了DINOv2在8例经蝶窦内鼻手术视频馈送中的应用。结果由高级神经外科医生和住院神经外科医生对质量进行评估。此外，从视频中随机选择的488张图像子集的深度估计与人工分割的深度，中间和表面区域的估计进行了半定量比较。结果使用DINOv2生成数值深度图，并创建颜色图进行深度可视化。尽管这些颜色图并不完美，但它们与资深神经外科医生和住院神经外科医生对视频中深度的主观评估非常吻合。模型估计的半定量验证产生了0.48的平均总体DICE相似指数。这些半定量的结果应该谨慎解释，因为用于模型深度预测和人工分割的截止值是不标准化的。结论通过应用DINOv2，我们能够通过生成数值图和深度颜色图来估计经蝶窦内鼻手术的内镜成像深度。这说明了基于深度学习的深度估计的潜力，这在未来可能有助于改善术中定位。它还强调了使用人工智能来增强内窥镜视频馈送的机会。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Journal of Clinical Neuroscience 医学-临床神经学

CiteScore

4.50

自引率

0.00%

发文量

402

审稿时长

40 days

期刊介绍： This International journal, Journal of Clinical Neuroscience, publishes articles on clinical neurosurgery and neurology and the related neurosciences such as neuro-pathology, neuro-radiology, neuro-ophthalmology and neuro-physiology. The journal has a broad International perspective, and emphasises the advances occurring in Asia, the Pacific Rim region, Europe and North America. The Journal acts as a focus for publication of major clinical and laboratory research, as well as publishing solicited manuscripts on specific subjects from experts, case reports and other information of interest to clinicians working in the clinical neurosciences.