Using Artificial Neural Network to Determine Favorable Wheelchair Tilt and Recline Usage in People with Spinal Cord Injury: Training ANN with Genetic Algorithm to Improve Generalization

Fu, Jicheng; Genson, Jerrad; Jan, Yih‐Kuen; Jones, Mark P.

doi:10.1109/ictai.2011.13

Cited by 2 publications

(9 citation statements)

References 14 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This study is different from our previous approach [11] in that ANN will predict numerical values instead of discrete classifications. This is the so called regression learning, which is very challenging.…”

Section: Introductionmentioning

confidence: 72%

“…Moreover, the learned function g can be used to cross-validate the outputs of another function f [11] that was designed to classify the skin perfusion changes into three discrete categories, i.e., positive (1), neutral (0), and negative (-1). Given a data item 〈a, s, d, l, c, t, r〉, we will be confident with the predictions in the following three cases:…”

Section: Discussionmentioning

confidence: 99%

“…Our previous study [11] showed that there were no generic rules that were suitable for all the wheelchair users on how to effectively use wheelchair power seat functions. The skin blood flow of people with SCI responded very differently to wheelchair tilt and recline settings.…”

Section: Introductionmentioning

confidence: 99%

“…As a result, the clinical guidance on wheelchair tilt and recline usage should be customized for individual persons. We attempted to address this issue by using the artificial neural network (ANN) to classify favorable tilt and recline settings that could result in skin perfusion increase for individual persons [11]. The limitation of that approach, however, is that it can only make discrete classifications, namely, positive, neutral, or negative.…”

Section: Introductionmentioning

confidence: 99%

“…Experimental results demonstrate that the proposed approach is very promising: The ANN trained by PSO generalizes significantly better than the one trained by BP. By combining our previous classification approach [11] and the proposed PSO-based regression approach, our intelligent system can cross-validate the outputs of the two approaches to ensure even better prediction quality.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

Towards an intelligent system for clinical guidance on wheelchair tilt and recline usage

Wiechmann

Jan

et al. 2012

2012 Annual International Conference of the IEEE Engineering in Medicine and Biology Society

View full text Add to dashboard Cite

We propose to construct an intelligent system for clinical guidance on how to effectively use power wheelchair tilt and recline functions. The motivations fall into the following two aspects. (1) People with spinal cord injury (SCI) are vulnerable to pressure ulcers. SCI can lead to structural and functional changes below the injury level that may predispose individuals to tissue breakdown. As a result, pressure ulcers can significantly affect the quality of life, including pain, infection, altered body image, and even mortality. (2) Clinically, wheelchair power seat function, i.e., tilt and recline, is recommended for relieving sitting-induced pressures. The goal is to increase skin blood flow for the ischemic soft tissues to avoid irreversible damage. Due to variations in the level and completeness of SCI, the effectiveness of using wheelchair tilt and recline to reduce pressure ulcer risks has considerable room for improvement. Our previous study indicated that the blood flow of people with SCI may respond very differently to wheelchair tilt and recline settings. In this study, we propose to use the artificial neural network (ANN) to predict how wheelchair power seat functions affect blood flow response to seating pressure. This is regression learning because the predicted outputs are numerical values. Besides the challenging nature of regression learning, ANN may suffer from the overfitting problem which, when occurring, leads to poor predictive quality (i.e., cannot generalize). We propose using the particle swarm optimization (PSO) algorithm to train ANN to mitigate the impact of overfitting so that ANN can make correct predictions on both existing and new data. Experimental results show that the proposed approach is promising to improve ANN's predictive quality for new data.

show abstract

Section: Introductionmentioning

confidence: 72%

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Towards an intelligent system for clinical guidance on wheelchair tilt and recline usage

Wiechmann

Jan

et al. 2012

2012 Annual International Conference of the IEEE Engineering in Medicine and Biology Society

View full text Add to dashboard Cite

show abstract

Assistive locomotion strategies for active lower limb devices

Figueiredo

Moreno

Santos

2017

2017 IEEE 5th Portuguese Meeting on Bioengineering (ENBENG)

View full text Add to dashboard Cite

First, I want to thank to my parents and siblings, in particular to my mother by unconditional support over these last few years. Besides to provide me the best conditions for academic training, she was also an emotional pillar, indicating to me always the best way forward. Thank you for all your strength and encouragement. I have great pride in you.Thank you to my mentors, Cristina Santos and Juan Moreno.Professor Cristina, I am eternally grateful to you by the your support and encouragement, either professionally, either in personal situations. Thank you by the knowledge that you conveyed me, and by the opportunities that you provided me, mainly the possibility to accomplish Erasmus placement in Consejo Superior de Investigaciones Científicas (CSIC). Thanks by the constant messages of courage, which helped me to overcome some challenges that emerged over this year.Juan, thank you for the excellent reception in the CSIC and by letting me join to research team of BioMot project. I am also very grateful for all your support and by the knowledge that you provided me, which were fundamental to initialize my investigation in the context of rehabilitation. Thank you by your excellent collaboration.I also thank the collaboration and the fellowship of all my colleagues of the CSIC.

show abstract

Using Artificial Neural Network to Determine Favorable Wheelchair Tilt and Recline Usage in People with Spinal Cord Injury: Training ANN with Genetic Algorithm to Improve Generalization

Cited by 2 publications

References 14 publications

Towards an intelligent system for clinical guidance on wheelchair tilt and recline usage

Towards an intelligent system for clinical guidance on wheelchair tilt and recline usage

Assistive locomotion strategies for active lower limb devices

Contact Info

Product

Resources

About