A probabilistic SVM based decision system for pain diagnosis
Introduction
Low back pain (LBP) affects a large proportion of the population in the world. Sixty to ninety percent of the adult population is at risk of developing low back pain in their lives (Andersson, 1999, Skovron et al., 1994). In 1997, LBP costs US 171 billion in the industrial setting of the US (Leigh, Markowitz, Fahs, Shin, & Landrigan, 1997), indirect costs related to days lost from work are substantial, with approximately 2% of the US work force compensated for back injuries each year (Andersson, 1999). Researchers and clinicians have developed several treatment regimes for LBP, which includes nonsteroidal anti-inflammatory drugs (NSAIDs), skeletal muscle relaxants, opioid analgesics, benzodiazepines, systemic corticosteroids, antidepressant medications, and antiepileptic drugs (Cherkin et al., 1998, Bernstein et al., 2004, Luo et al., 2004a, Luo et al., 2004b, Di Iorio et al., 2000). However, the efficacy of one treatment regime may vary significantly in populations. In addition, conflicting results were reported in previous researches, for example, sufficient evidence have found to support the usage of NSAIDs and skeletal muscle relaxants for LBP, and limited evidence for other treatments (Deyo, 1996). But another research recommended acetaminophen for mild or moderate pain other than NSAIDs (Chou & Huffman, 2007). Since selecting treatment regime especially pharmacologic therapy is usually associated with the trade-off between benefits and harms, it is critical to diagnosis LBP precisely and evaluates the efficacy of treatment for individual patients.
On the other hand the pathology underlying LBP remains largely unknown. Since most of LBP are happened together with mechanical function alteration, many researchers explore the relationship between LBP and various mechanical properties. Particularly, static or dynamic paraspinal surface electromyography (sEMG) has been observed increasing in chronic LBP patients (Arena et al., 1989, Arena et al., 1990), while some long-term pathology are related with increased stiffness (Comerford & Mottram, 2001). Hu, Wong, Lu, and Kawchuk (2009) found that the correlation of sEMG and muscle stiffness signal may have relation with the treatment efficacy. However, only the general expert knowledge is not sufficient to form clinical diagnosis tools for individual patients since there are various uncertainties in individual samples such as measurement noise, sensor location variance and diversity of patients’ specificities.
In this paper, a decision system is proposed for the diagnosis of LBP. The decision system is the integration of a qualitative knowledge model and a quantitative model. The expert knowledge is integrated into the decision system to formulate the qualitative knowledge model. To reduce the uncertainties in sample patients, probabilistic SVM (PSVM) (Yang & Li, submitted for publication) is employed to formulate the quantitative model. PSVM is an extended modeling method developed from support vector machine (SVM) and has the advantages of SVM, such as the ability to work at high dimensionality, good generalization performance and global optimal solution. Moreover, PSVM could handle uncertainties in data samples, while SVM is sensitive to uncertainties (Matić Guyon and Vapnik, 1996, Zhang, 1999). Similar to SVM, PSVM works by mapping samples in input space to high dimensional feature space, in which those samples could be separated linearly. In PSVM, the uncertain parts are characterized by a probabilistic distribution function (PDF) of the separating margin in feature space. A probabilistic separating margin is created in PSVM instead of deterministic margin in SVM. Through the creation of probabilistic margin, more information is integrated in PSVM, thus PSVM works better than SVM in uncertain circumstances.
Section snippets
Problem description
In clinical practice, surface EMG (sEMG) (Hu, Shiu, Mak, & Luk, 2010), lumbar stiffness (Kawchuk, Liddle, Fauvel, & Johnston, 2006) were applied to evaluate low back pain treatment as shown in Fig. 1a and b. Thereafter, a correlation of sEMG and stiffness was proposed. However, it was questioned about the linear correlation between sEMG and stiffness. As well, it is still unknown whether the changes of sEMG and stiffness have been associated with pain relief.
In this study, paraspinal stiffness
Methodology design
Since medical researchers have done much works on the relations among sEMG, muscle stiffness signals and the diagnosis of LBP, in this paper, the existing general expert knowledge is integrated into the design of the decision system. To model the individual patients’ uncertainties a data based model is constructed. The decision system consists of two parts: the qualitative knowledge model and quantitative model. The knowledge model is a deterministic model. It is constructed by translating
Results and discussion
Twenty-one patients with LBP aged 24–59 (mean 42.4, SD = 10), were recruited. There are 5 female patients and 16 male patients in subject pool. All enrolled participants provided written consent. This protocol was approved by the Institutional Review Board of The University of Hong Kong.
Since not all patients are respondent to physical therapy, first goal of the model is to discriminate the patients which are respondent to treatment from those which are not respondent. Patients that have
Conclusion
A probabilistic SVM based decision system is proposed to handle classification and treatment evaluation of low back pain. The decision system consists of two parts: the deterministic qualitative knowledge model and the probabilistic quantitative model. Qualitative knowledge model is constructed from expert knowledge from medical literatures and serve as nominal model for the decision system. PSVM is employed to construct the quantitative model. The dynamics learned from patient samples by PSVM
Acknowledgement
Authors would like to thank Dr. Xinjiang Lu, Dr. Xiaogang Duan, and Mr. Hongtao Liu for their valuable discussions. The project is partially supported by a SRG grant from City University of Hong Kong (7008057), a GRF grant from RGC of Hong Kong (GRF 712408E) and S.K. Yee Medical Foundation (207210).
References (27)
Epidemiological features of chronic low-back pain
Lancet
(1999)- et al.
Electromyographic recordings of 5 types of low back pain subjects and non-pain controls in different positions
Pain
(1989) - et al.
Temporal stability of paraspinal electromyographic recordings in low back pain and non-pain subjects
International Journal of Psychophysiology
(1990) A descriptive study of VAS and VRS for the assessment of post operative pain in orthopeadic patients
Journal of Pain Symptom Management
(1999)Identification of contributing variables using kernel-based discriminant modeling and reconstruction
Expert Systems with Applications
(2007)- et al.
Functional stability re-training: Principles and strategies for managing mechanical dysfunction
Manual Therapy
(2001) - et al.
Credit scoring with a data mining approach based on support vector machines
Expert Systems with Applications
(2007) - et al.
Lumbar muscle electromyographic dynamic topography during flexion-extension
Journal of Electromyography and Kinesiology
(2010) - et al.
Creation of an asymmetrical gradient of back muscle activity and spinal stiffness during asymmetrical hip extension
Clinical Biomechanics
(2009) - et al.
The accuracy of ultrasonic indentation in detecting simulated bone displacement: A comparison of three techniques
Journal of Manipulative and Physiological Therapeutics
(2006)
Particle swarm optimization for parameter determination and feature selection of support vector machines
Expert Systems with Applications
The use of muscle relaxant medications in acute low back pain
Spine
Medication use for low back pain in primary care
Spine
Cited by (27)
Clinical Decision Support Systems for Triage in the Emergency Department using Intelligent Systems: a Review
2020, Artificial Intelligence in MedicineCitation Excerpt :This makes it a non-probabilistic binary linear classifier. Another disadvantage of this model is that it is very sensitive to uncertainties [75], and prone to over-fitting in a high dimensional space [76]. However, it has a good generalization ability, being also robust for high dimensional data [77].
Using a posterior probability support vector machine model to assess soil quality in Taiyuan, China
2019, Soil and Tillage ResearchCitation Excerpt :Applying Bayesian theory to the SVM method, the posterior probability not only represents the different contributions of each sample to the classifier and indicates the uncertainty of the sample category but also avoids estimation of the density function (Wu et al., 2003). The posterior probability SVM has been used regularly to deal with real-world applications (Li et al., 2013; Yang et al., 2011). However, few evaluations of soil quality with fuzzy property have been assessed by posterior probability SVM.
A novel effective diagnosis model based on optimized least squares support machine for gene microarray
2018, Applied Soft Computing JournalDesign and evaluation of a decision support system for pain management based on data imputation and statistical models
2016, Measurement: Journal of the International Measurement ConfederationDempster-Shafer theory-based robust least squares support vector machine for stochastic modelling
2016, NeurocomputingCitation Excerpt :Most discussions about the LS-SVM are based on the assumption that the training data are not corrupted by noise [6,13]. In contrast, there are also many studies into ways to reduce the effect of noise on the LS-SVM model [12,14,15]. A range of loss functions have been employed to improve robustness such as, the weighted SVM [10], the reweighting SVM [16,17] and Nesterov’s SVM (NESVM) [18].
Melt index prediction by least squares support vector machines with an adaptive mutation fruit fly optimization algorithm
2015, Chemometrics and Intelligent Laboratory SystemsCitation Excerpt :Finally, Section 4 concludes this paper. The support vector machine method introduced by Vapnik [34] is a reliable tool for solving pattern recognition and classification problems, and is used to solve many problems successfully in lots of fields [35–39]. The LSSVM is an extension of SVM, which applies the linear least squares criteria to the loss function instead of inequality constraints [40].