Deep Learning Analysis of In Vivo Hyperspectral Images for Automated Intraoperative Nerve Detection

被引:24
作者
Barberio, Manuel [1 ,2 ,3 ]
Collins, Toby [2 ]
Bencteux, Valentin [1 ]
Nkusi, Richard [4 ]
Felli, Eric [1 ]
Viola, Massimo Giuseppe [3 ]
Marescaux, Jacques [2 ]
Hostettler, Alexandre [2 ]
Diana, Michele [2 ,5 ]
机构
[1] IHU Strasbourg, Dept Res, Inst Image Guided Surg, F-67091 Strasbourg, France
[2] IRCAD, Dept Res, Res Inst Digest Canc, F-67091 Strasbourg, France
[3] Osped Card G Pan, Dept Surg, I-73039 Tricase, Italy
[4] IRCAD Africa, Dept Res, Res Inst Digest Canc, 2 KN 30 ST, Kigali, Rwanda
[5] Photon Instrumentat Hlth, ICUBE Lab, F-67412 Strasbourg, France
关键词
hyperspectral imaging; artificial intelligence; tissue recognition; intraoperative navigation tool; optical imaging; deep learning; precision surgery; convolutional neural network; CLASSIFICATION; VISUALIZATION; SURGERY;
D O I
10.3390/diagnostics11081508
中图分类号
R5 [内科学];
学科分类号
1002 ; 100201 ;
摘要
Nerves are critical structures that may be difficult to recognize during surgery. Inadvertent nerve injuries can have catastrophic consequences for the patient and lead to life-long pain and a reduced quality of life. Hyperspectral imaging (HSI) is a non-invasive technique combining photography with spectroscopy, allowing non-invasive intraoperative biological tissue property quantification. We show, for the first time, that HSI combined with deep learning allows nerves and other tissue types to be automatically recognized in in vivo hyperspectral images. An animal model was used, and eight anesthetized pigs underwent neck midline incisions, exposing several structures (nerve, artery, vein, muscle, fat, skin). State-of-the-art machine learning models were trained to recognize these tissue types in HSI data. The best model was a convolutional neural network (CNN), achieving an overall average sensitivity of 0.91 and a specificity of 1.0, validated with leave-one-patient-out cross-validation. For the nerve, the CNN achieved an average sensitivity of 0.76 and a specificity of 0.99. In conclusion, HSI combined with a CNN model is suitable for in vivo nerve recognition.
引用
收藏
页数:20
相关论文
共 59 条
[1]   Hyperspectral imaging and quantitative analysis for prostate cancer detection [J].
Akbari, Hamed ;
Halig, Luma V. ;
Schuster, David M. ;
Osunkoya, Adeboye ;
Master, Viraj ;
Nieh, Peter T. ;
Chen, Georgia Z. ;
Fei, Baowei .
JOURNAL OF BIOMEDICAL OPTICS, 2012, 17 (07)
[2]   Detection and Analysis of the Intestinal Ischemia Using Visible and Invisible Hyperspectral Imaging [J].
Akbari, Hamed ;
Kosugi, Yukio ;
Kojima, Kazuyuki ;
Tanaka, Naofumi .
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, 2010, 57 (08) :2011-2017
[3]  
[Anonymous], 2016, P BIG DAT SPAC BIDS1, DOI DOI 10.2788/854791
[4]   Hyperspectral imaging for tissue classification, a way toward smart laparoscopic colorectal surgery [J].
Baltussen, Elisabeth J. M. ;
Kok, Esther N. D. ;
de Koning, Susan G. Brouwer ;
Sanders, Joyce ;
Aalbers, Arend G. J. ;
Kok, Niels F. M. ;
Beets, Geerard L. ;
Flohil, Claudie C. ;
Bruin, Sjoerd C. ;
Kuhlmann, Koert F. D. ;
Sterenborg, Henricus J. C. M. ;
Ruers, Theo J. M. .
JOURNAL OF BIOMEDICAL OPTICS, 2019, 24 (01)
[5]  
Barberio Manuel, 2018, Current Directions in Biomedical Engineering, V4, P399, DOI 10.1515/cdbme-2018-0095
[6]   Quantitative serosal and mucosal optical imaging perfusion assessment in gastric conduits for esophageal surgery: an experimental study in enhanced reality [J].
Barberio, Manuel ;
Felli, Eric ;
Pizzicannella, Margherita ;
Agnus, Vincent ;
Al-Taher, Mahdi ;
Seyller, Emilie ;
Moulla, Yusef ;
Jansen-Winkeln, Boris ;
Gockel, Ines ;
Marescaux, Jacques ;
Diana, Michele .
SURGICAL ENDOSCOPY AND OTHER INTERVENTIONAL TECHNIQUES, 2021, 35 (10) :5827-5835
[7]   HYPerspectral Enhanced Reality (HYPER): a physiology-based surgical guidance tool [J].
Barberio, Manuel ;
Longo, Fabio ;
Fiorillo, Claudio ;
Seeliger, Barbara ;
Mascagni, Pietro ;
Agnus, Vincent ;
Lindner, Veronique ;
Geny, Bernard ;
Charles, Anne-Laure ;
Gockel, Ines ;
Worreth, Marc ;
Saadi, Alend ;
Marescaux, Jacques ;
Diana, Michele .
SURGICAL ENDOSCOPY AND OTHER INTERVENTIONAL TECHNIQUES, 2020, 34 (04) :1736-1744
[8]   STANDARD NORMAL VARIATE TRANSFORMATION AND DE-TRENDING OF NEAR-INFRARED DIFFUSE REFLECTANCE SPECTRA [J].
BARNES, RJ ;
DHANOA, MS ;
LISTER, SJ .
APPLIED SPECTROSCOPY, 1989, 43 (05) :772-777
[9]  
Bishop C.M., 2006, MACH LEARN, V128
[10]   Hyperspectral Imaging in the Medical Field: Present and Future [J].
Calin, Mihaela Antonina ;
Parasca, Sorin Viorel ;
Savastru, Dan ;
Manea, Dragos .
APPLIED SPECTROSCOPY REVIEWS, 2014, 49 (06) :435-447