Assessments for the Results of Surgical Treatments for Brachial Plexus Injuries -- from the Viewpoints of Biomechanics and Neural Network

  • 杜 元坤

Student thesis: Doctoral Thesis

Abstract

Background: Brachial plexus injury (BPI) is a devastating injury mostly caused by high energy trauma There are 2 major types of BPI: upper arm type (C5-6 injury) and total arm type (C5-6-7-8-T1 injury) The priorities of surgical treatments are to restore the shoulder / elbow function and hand function as well There are three topics in this study: (1) Surgical technique and outcome evaluation comparing the total contralateral C7 root transfer to hemi-CC7 transfer on total root avulsion BPI (2) To quantify the muscle strength for post-surgical BPI patients' evaluation comparing the outcomes of single- double- and triple- nerve transfer and (3) To establish a model to predict the prognosis of BPI patients after reconstruction by artificial neural networks (ANN) Materials and Methods: (1) Forty patients who received neurotization for BPI were enrolled in this prospective study Group 1 (n=20) received hemi-CC7 transfer for hand function while group 2 (n=20) received total-CC7 transfer Additional neurotization included spinal accessory phrenic and intercostal nerve transfer for shoulder and elbow function The results were evaluated with an average of 6 years follow-up (2) Forty-three patients who received neurotization for BPI were enrolled Patients were divided into 3 subgroups: single nerve transfer (n=14) double nerve transfer group (n=22) and triple nerve transfer group (n=7) Quantitative strength of shoulder flexion abduction external rotation and elbow flexion were measured by the handheld dynamometer One-way ANOVA and simple linear regression tests were used to examine the differences among three groups and the relationships between progress of motor recovery and follow-up period (3) A sample of 40 patients who have upper arm type and treated with nerve transform were collected The commercial neural network software NeuroSolutions was used for building the model to predict the prognosis of BPI patients Fifteen parameters from clinical record were set as the input variables while the MRC scores were set as the output variables Three common neural network models including linear regression radial basis function and multilayer perceptron were built and used to compare the accuracy of classification results for shoulder abduction and external rotation Results: (1) Group 1 had fewer donor site complications (15%) than group 2 (45%); group 2 had significantly better hand M3 and M4 motor function (65%) than group 1 (35%; p =0 02) There was no difference in sensory recovery Significantly better shoulder function was obtained by simultaneous neurotization on both suprascapular and axillary nerves (2) There were no significant differences in MRC scores among three groups Results of quantitative strength (%non-involved) showed the patients in single group had the trend of larger strength of external rotator than double and triple groups (p=0 04) Patients in single group had larger strength of elbow flexion than those in double (p
Date of Award2014 Aug 27
Original languageEnglish
SupervisorChih-Han Chang (Supervisor)

Cite this

'