Artificial neural network to predict patient body circumferences and ligament thicknesses
Authors: Vaughan, N., Dubey, V.N., Wee, M.Y.K. and Isaacs, R.
Journal: Proceedings of the ASME Design Engineering Technical Conference
Volume: 2 A
DOI: 10.1115/DETC2013-13088
Abstract:An artificial neural network has been implemented and trained with clinical data from 23088 patients. The aim was to predict a patient's body circumferences and ligament thickness from patient data. A fully connected feed-forward neural network is used, containing no loops and one hidden layer and the learning mechanism is back-propagation of error. Neural network inputs were mass, height, age and gender. There are eight hidden neurons and one output. The network can generate estimates for waist, arm, calf and thigh circumferences and thickness of skin, fat, Supraspinous and interspinous ligaments, ligamentum flavum and epidural space. Data was divided into a training set of 11000 patients and an unseen test data set of 12088 patients. Twenty five training cycles were completed. After each training cycle neuron outputs advanced closer to the clinically measured data. Waist circumference was predicted within 3.92cm (3.10% error), thigh circumference 2.00cm, (2.81% error), arm circumference 1.21cm (2.48% error), calf circumference 1.41cm, (3.40% error), triceps skinfold 3.43mm, (7.80% error), subscapular skinfold 3.54mm, (8.46% error) and BMI was estimated within 0.46 (0.69% error). The neural network has been extended to predict ligament thicknesses using data from MRI. These predictions will then be used to configure a simulator to offer a patient-specific training experience. Copyright © 2013 by ASME.
Source: Scopus
ARTIFICIAL NEURAL NETWORK TO PREDICT PATIENT BODY CIRCUMFERENCES AND LIGAMENT THICKNESSES
Authors: Vaughan, N., Dubey, V.N., Wee, M.Y.K. and Isaacs, R.
Journal: PROCEEDINGS OF THE ASME INTERNATIONAL DESIGN ENGINEERING TECHNICAL CONFERENCES AND COMPUTERS AND INFORMATION IN ENGINEERING CONFERENCE, 2013, VOL 2A
ISBN: 978-0-7918-5585-0
Source: Web of Science (Lite)
Artificial neural network to predict patient body circumferences and ligament thicknesses
Authors: Vaughan, N., Dubey, V.N., Wee, M.Y.K. and Isaacs, R.
Conference: ASME 2013 International Design Engineering Technical Conferences & Computers and Information in Engineering Conference
Dates: 4-7 August 2013
Source: Manual
Preferred by: Venky Dubey
Artificial Neural Network to Predict Patient Body Circumferences and Ligament Thicknesses
Authors: Vaughan, N., Dubey, V.N., Wee, M.Y.K. and Isaacs, R.
Conference: ASME 2013 Computers and Information in Engineering Conference, IDETC/CIE 2013
Dates: 4-8 August 2013
Journal: asmedigitalcollection.asme.org/
Publisher: ASME
Place of Publication: asmedigitalcollection.asme.org/
Abstract:An artificial neural network has been implemented and trained with clinical data from 23088 patients. The aim was to predict a patient’s body circumferences and ligament thickness from patient data. A fully connected feed-forward neural network is used, containing no loops and one hidden layer and the learning mechanism is back-propagation of error. Neural network inputs were mass, height, age and gender. There are eight hidden neurons and one output. The network can generate estimates for waist, arm, calf and thigh circumferences and thickness of skin, fat, Supraspinous and interspinous ligaments, ligamentum flavum and epidural space. Data was divided into a training set of 11000 patients and an unseen test data set of 12088 patients. Twenty five training cycles were completed. After each training cycle neuron outputs advanced closer to the clinically measured data. Waist circumference was predicted within 3.92cm (3.10% error), thigh circumference 2.00cm, (2.81% error), arm circumference 1.21cm (2.48% error), calf circumference 1.41cm, (3.40% error), triceps skinfold 3.43mm, (7.80% error), subscapular skinfold 3.54mm, (8.46% error) and BMI was estimated within 0.46 (0.69% error). The neural network has been extended to predict ligament thicknesses using data from MRI. These predictions will then be used to configure a simulator to offer a patient-specific training experience.
Source: Manual