Computer-aided detection of parathyroid glands using a dual-RGB/NIR imaging and deep learning technique

Parathyroid glands (PGs), small endocrine glands in the neck, control calcium levels in the body and are crucial to maintaining homeostasis. Accidental removal of or direct damage to healthy parathyroid glands during thyroid surgery may occur due to its small size and similar appearance to surrounding anatomical structures, potentially leading to postoperative hypocalcemia. Thus precise and quick detection of normal parathyroid glands in real-time during surgery can improve the surgical outcome. In this study, we introduce a deep learning system (YOLOv5) based on dual RGB/NIR imaging for Computer-aided detection (CADe) of PG with high accuracy. This model can effectively detect parathyroid glands in real-time as it also includes the confidence level, which can help surgeons make decisions. We tested a computer-aided detection (CADe) using the co-registered RGB/NIR camera and ex-vivo thyroid tissue specimen. The average precisions of models were significantly higher when trained by the dual-RGB/NIR (0.99) data than NIR (0.94) and RGB (0.96) data alone at a high confidence threshold (0.7). The proposed CADe may increase the parathyroid detection rates clinically.

[1]  Q. Duh,et al.  Current state of intraoperative use of near infrared fluorescence for parathyroid identification and preservation. , 2020, Surgery.

[2]  M. Wallace,et al.  Efficacy of Real-Time Computer-Aided Detection of Colorectal Neoplasia in a Randomized Trial. , 2020, Gastroenterology.

[3]  Jieping Ye,et al.  Object Detection in 20 Years: A Survey , 2019, Proceedings of the IEEE.

[4]  Frederic De Leeuw,et al.  Parathyroid gland management using optical technologies during thyroidectomy or parathyroidectomy: A systematic review. , 2018, Oral oncology.

[5]  S. Natsugoe,et al.  Intraoperative Identification of the Parathyroid Gland with a Fluorescence Detection System , 2017, World Journal of Surgery.

[6]  Ronald M. Summers,et al.  Deep Convolutional Neural Networks for Computer-Aided Detection: CNN Architectures, Dataset Characteristics and Transfer Learning , 2016, IEEE Transactions on Medical Imaging.

[7]  Ali Farhadi,et al.  You Only Look Once: Unified, Real-Time Object Detection , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[8]  Pietro Perona,et al.  Microsoft COCO: Common Objects in Context , 2014, ECCV.

[9]  Matthew D. Keller,et al.  Near-infrared autofluorescence for the detection of parathyroid glands. , 2011, Journal of biomedical optics.

[10]  Jude W. Shavlik,et al.  Transfer Learning via Advice Taking , 2010, Advances in Machine Learning I.

[11]  M. Fried,et al.  Assessment of the morbidity and complications of total thyroidectomy. , 2002, Archives of otolaryngology--head & neck surgery.

[12]  J. Gavilán,et al.  Complications following thyroid surgery , 1991, Archives of otolaryngology--head & neck surgery.

[13]  J. Phay,et al.  Intraoperative Parathyroid Autofluorescence Detection in Patients with Primary Hyperparathyroidism , 2019 .

[14]  E. Berber,et al.  Autofluorescence imaging of parathyroid glands: An assessment of potential indications. , 2019, Surgery.