Artificial Neural Network: The Alternative Method to Obtain the Dimension of Ankle Bone Parameters

• Main Authors: R., Daud, Mas Ayu, Hassan, Salwani, Mohd Salleh, Siti Haryani, Tomadi, Mohammed Rafiq, Abdul Kadir, Raghavendran, Hanumantharao Balaji, Tunku, Kamarul
• Format: Article
• Language: English
• Published: Medwell Journals 2017
• Subjects: TJ Mechanical engineering and machinery
• Online Access: http://umpir.ump.edu.my/id/eprint/20082/1/fkm-2017-masayu-Artificial%20Neural%20Network%20The%20Alternative%20Method.pdf; http://umpir.ump.edu.my/id/eprint/20082/; http://docsdrive.com/pdfs/medwelljournals/jeasci/2017/2782-2787.pdf
• Institution: Universiti Malaysia Pahang

Current ankle morphometric measurement tools involve the use of radiographic techniques which maybe rmacceptable to many ethical committees due to the radiation exposure to subjects. In the present study, we propose an alternative method of ankle morphometric measurement using neural network computational model based solely on existing data measurements and demographic information. The reliability and prediction power of this technique were examined and compared with the morphometric measurements of normal subjects using Computed Tomography (CT) scan measurements and Multiple Linear Regression (1.1LR) method of prediction. The Artificial Nemal Network (ANN) used in the present study was based on two-layer feed forward network. The network system included a hidden layer sigmoid transfer fllllction and a linear transfer fllllction in the output layer. For network training, standard levenberg-marquardt algorithm was used. The input used consisted of a set of demographic data (age, height and weight) while the output obtained from the analyses consisted of ankle morphometric measurements (Trochlea Tali Length (TTL) Talar Anterior Width (TaA W) Sagittal Radius of talar (SRTa) Tibia Length (TiL) Tibia Width (TiW) Widtli!LengthRatio of Talar (WLR Ta) and Widtli!Length Ratio of Tibia(WLRTi)). The applicability and accuracy of these alternative methods were evaluated by comparing the predicted values from our computational analysis with the normal CT values of 15 randomly selected volrmteers. Furthermore, our prediction values were also compared with the values predicted using the 1.1LR method. The ANN method showed a greater capacity of prediction and was folllld to estimate the ankle joint morphometric measurements with a low percentage of error and high correlative values with the measurements obtained through the use of CT scan. In addition, the ANN method was also noted to be better in predicting ankle measurements than the 1.1LR method as demonstrated by the lower average of standard deviations: SANN~ 1.35, SMLR ~ 2.20 for females and SANN~ 1.81, SMLR ~ 4.07 for males. The ANN method is potentially better alternative to predict ankle morphometric measurements than CT scan and 1.1LR methods.