Seyma Nazli Avci1, Gizem Isiktas1, Eren Berber2,3. 1. Department of Endocrine Surgery, Cleveland Clinic, Cleveland, Ohio, USA. 2. Department of Endocrine Surgery, Cleveland Clinic, Cleveland, Ohio, USA. berbere@ccf.org. 3. Department of General Surgery, Cleveland Clinic, Cleveland, Ohio, USA. berbere@ccf.org.
Abstract
BACKGROUND AND PURPOSE: Parathyroid glands may be detected by their autofluorescence on near-infrared imaging. Nevertheless, recognition of parathyroid-specific autofluorescence requires a learning curve, with other unrelated bright signals causing confusion. The aim of this study was to find out whether machine learning could be used to facilitate identification of parathyroid-specific autofluorescence signals on intraoperative near-infrared images in patients undergoing thyroidectomy and parathyroidectomy procedures. METHODS: In an institutional review board-approved study, intraoperative near-infrared images of patients who underwent thyroidectomy and/or parathyroidectomy procedures within a year were used to develop an artificial intelligence model. Parathyroid-specific autofluorescence signals were marked with rectangles on intraoperative near-infrared still images and used for training a deep learning model. A randomly chosen 80% of the data were used for training, 10% for testing, and 10% for validation. Precision and recall of the model were calculated. RESULTS: A total of 466 intraoperative near-infrared images of 197 patients who underwent thyroidectomy and/or parathyroidectomy procedures were analyzed. Procedures included total thyroidectomy in 54 patients, thyroid lobectomy in 24 patients, parathyroidectomy in 108 patients, and combined thyroidectomy and parathyroidectomy procedures in 11 patients. The overall recall and precision of the model were 90.5 and 95.7%, respectively. CONCLUSIONS: To our knowledge, this is the first study that describes the use of artificial intelligence tools to assist in recognition of parathyroid-specific autofluorescence signals on near-infrared imaging. The model developed may have utility in facilitating training and decreasing the learning curve associated with the use of this technology.
BACKGROUND AND PURPOSE: Parathyroid glands may be detected by their autofluorescence on near-infrared imaging. Nevertheless, recognition of parathyroid-specific autofluorescence requires a learning curve, with other unrelated bright signals causing confusion. The aim of this study was to find out whether machine learning could be used to facilitate identification of parathyroid-specific autofluorescence signals on intraoperative near-infrared images in patients undergoing thyroidectomy and parathyroidectomy procedures. METHODS: In an institutional review board-approved study, intraoperative near-infrared images of patients who underwent thyroidectomy and/or parathyroidectomy procedures within a year were used to develop an artificial intelligence model. Parathyroid-specific autofluorescence signals were marked with rectangles on intraoperative near-infrared still images and used for training a deep learning model. A randomly chosen 80% of the data were used for training, 10% for testing, and 10% for validation. Precision and recall of the model were calculated. RESULTS: A total of 466 intraoperative near-infrared images of 197 patients who underwent thyroidectomy and/or parathyroidectomy procedures were analyzed. Procedures included total thyroidectomy in 54 patients, thyroid lobectomy in 24 patients, parathyroidectomy in 108 patients, and combined thyroidectomy and parathyroidectomy procedures in 11 patients. The overall recall and precision of the model were 90.5 and 95.7%, respectively. CONCLUSIONS: To our knowledge, this is the first study that describes the use of artificial intelligence tools to assist in recognition of parathyroid-specific autofluorescence signals on near-infrared imaging. The model developed may have utility in facilitating training and decreasing the learning curve associated with the use of this technology.
Authors: Fares Benmiloud; Stanislas Rebaudet; Arthur Varoquaux; Guillaume Penaranda; Marie Bannier; Anne Denizot Journal: Surgery Date: 2017-11-06 Impact factor: 3.982
Authors: Emin Kose; Anatoliy V Rudin; Bora Kahramangil; Edwina Moore; Husnu Aydin; Mustafa Donmez; Vikram Krishnamurthy; Allan Siperstein; Eren Berber Journal: Surgery Date: 2019-09-13 Impact factor: 3.982
Authors: Andre Esteva; Alexandre Robicquet; Bharath Ramsundar; Volodymyr Kuleshov; Mark DePristo; Katherine Chou; Claire Cui; Greg Corrado; Sebastian Thrun; Jeff Dean Journal: Nat Med Date: 2019-01-07 Impact factor: 53.440
Authors: Jacek Baj; Robert Sitarz; Marek Łokaj; Alicja Forma; Marcin Czeczelewski; Amr Maani; Gabriella Garruti Journal: Molecules Date: 2020-04-09 Impact factor: 4.411