INTRODUCTION:

Lower back pain is one of the most popular diseases that cause extreme discomfort for most adults. More than 80% of the world population is affected due to this disease; especially after a certain age and among them higher owing is inclined towards women due to hormonal and reproductive factors.¹ Usually, the ache experienced is in the lower part of the body, as it supports most of the weight of the body. One of the main reasons for lower back pain is the dislocation of a disc between the vertebra.² Usually, lower back pain can be felt when we lift a heavy object, sit in a single wrong posture position for a long time, or if we have any injury or accident.³ Acute lower back pain is experienced due to sudden shock or injury to the ligament muscle that supports the back. The pain can be caused due to tear, rupture, strain or spasms in the muscles and ligaments.⁴

The experience of LBP may include a tingling or burning sensation, a dull achy feeling, or sharp pain. The pain may be mild, or it can be so severe that makes the person unable to move.

Few indirect reasons for LBP include kidney stone, pregnancy complications, infection in the spine (e.g. diskitis, abscess), ovarian cancer, and ovarian cysts. LBP can also bag some potential pain that can worsen the situation i.e., pain in the leg, hip, or the bottom of the foot.⁵ Patients may also experience weakness in your legs and feet. LBP can also be caused by an assortment of issues with any parts of the complex, interconnected system of spinal muscles, nerves, bones, plates or ligaments in the lumbar spine.⁶LBP can be categorised into normal and abnormal categories. The potential reasons include: (a) The broadened nerve roots in the lower back that go to the legs may be damaged. (b) The little nerves may be damaged. (c) The expansive matched lower back muscles (erector spine) may be stressed. (d)The bones, tendons or joints may be harmed. (e) An intervertebral plate may be dislocated.

An aggravation or issue with any of these structures can cause lower back pain or potential pain that transmits or is alluded to different parts of the body. Numerous lower back issues likewise cause back muscle fits,⁷ that can cause extreme pain and inability. While lower back pain is to a great degree normal, the side effects and seriousness of LBP change enormously. A straightforward lower back muscle strain may be torment enough to require an emergency visit, while a declining circle may cause just mild, irregular uneasiness. One of the major issues is that the group of people encountering or not encountering the disease is wrongly categorized.⁸ Once LBP is diagnosed, treatment can be done immediately to prevent the worsening potential impact.⁹

In our proposed approach, Random Forest is used as the classification algorithm which results in comparatively better accuracy.¹⁰ Based on 13 attributes which are selected by performing the statistical method, Principal Component Analysis,¹¹ the patient can be classified under normal or abnormal category once their diagnosis is confirmed.¹² Figure 1 depicts the categorization. Parameter tuning is an emerging technique that optimises the performance of the classification algorithm in which it is used, and maintains the stability of the code.¹³ The accuracy rate for Random forest is 71.25% but after applying the proposed technique we can witness that it has been increased to 85.80%.

LITERATURE SURVEY:

Parameters decision for SVM is extraordinarily complex and exceptionally hard to unwind by normal methodology; so parameter tuning approach of SVM was proposed based on Quantum evolutionary algorithm that has better search capacity.¹⁴ An idea was proposed for two-level classification of data; where data mining methods were explored to identify the suitable method for efficient classification on the diabetic dataset.¹⁵ Ensemble classifier increases not only the performance of the classification, but also the confidence of the results. The motivation beyond using ensemble classifiers is that the results are less dependent on peculiarities of a single training set and because the ensemble system outperforms the best base classifier in the ensemble.¹⁶ An enhancement of parameter tuning technique was proposed which in turn proposes factual techniques to discover the parameter setting of an artificial intelligence technique, harmony Search (HS) calculation. The examination demonstrates that HMCR value is not worthy.¹⁷ Another work demonstrates the effect of acupuncture on treating chronic lower back pain. Acupuncture is a key component of Chinese medicine (TCM). This article also evaluates the comparison between acupuncture and the other treatments available.¹⁸

Techniques such as grid search, random search, Estimation of Distribution Algorithms (EDAs), bio-enlivened metaheuristics can be used for solving various issues enecountered. However, one of the major concern is with respect to the time complexity.¹⁹ A survey has been carried out that details the most common image processing techniques and comparison among them. The survey consists of various classification techniques such as Artificial Neural Network (ANN), Decision Tree (DT), Fuzzy Classification and more.²⁰ A research work that highlights the approach of spine disorder detection using physical spine data as a parameter and to achieve this most significant parameter is selected and then by using unsupervised learning algorithms such as PCA and RF, the aim is achieved.²¹ An approach has been proposed that mainly focuses on a novel metric that captures the stability of random forest predictions, which is the key for scenarios that require successive predictions. The theorems will have influence all over the metrics, but the ER-R will be the area to think about. No matter how metrics key stability bits upon the factor but the RFP calculation struck the major part.²²

An approach has been proposed that focuses on reducing the time required for the approval of new medicine in the pharmaceutical industry by the Food and Drug Administration (FDA) with the help of binary classification technique.²³ Another proposed technology conquers any hindrance between hyperparameter streamlining and group learning by performing Bayesian streamlining of a group with respect to its hyperparameters. This technique comprises of steps such as: building a settled size group, upgrading the setup of one classifier of the group at every cycle of the hyperparameter streamlining calculation, taking into thought the communication with alternate models while assessing potential exhibitions.²⁴ This study demonstrates the effect of the prolonged usage of laptop that results in neck, shoulder and low back pain. The main idea was to study the risk factor to which the laptop users are exposed to.²⁵ A study that has highlighted the effect of lower back pain in nursing staff in terms of expressions of behaviour, emotional disposition and other social activities which highlights the restrictions it has bought in the capacity of interpersonal commitment.²⁶ Survey was conducted to show the effectiveness of rehabilitation exercises therapy in numerous crucial phases which in turn provides vital information for treating chronic low back pain.²⁷

The IPC technique was adopted by researchers due to its globally optimized feature. Better efficiency was obtained while experimenting on a huge image database for road signal obstacle detection using adaptive tuning.²⁸ A modified Bayesian information criterion (MBIC) was proposed for selecting an optimal tuning parameter for the adaptive LASSO (least absolute shrinkage and selection operator). The adaptive LASSO obtained by minimizing the MBIC correctly distinguishes the true auto-regression coefficients from zero asymptotically.²⁹ A comparative study was conducted to determine the effectiveness of Mckenzie and William exercise in mechanical low back pain by using a VAS scale and suggesting effective exercise as per the patient’s situation.³⁰ Comprehensive analysis on Lower back pain was performed and discussion was carried out on disordered physiological processes, diagnosis and the respective treatment that can be considered.³¹

Map Matching Algorithms (MMA) was created to investigate spatial ambiguities that emerge during the time spent doing out GPS estimations. There is an absence of efficient parameter tuning approaches for streamlining the MMA execution.³² Meenakshi et. al. has analyzed the merits and demerits of machine learning algorithms. The main goal is to identify a better machine learning approach that can lead to accurate learning with less false positive rate.³³ Object tracking process technique was introduced to adapt the scene condition variations. More precisely, this approach learns how to tune the parameters to cope with the tracking context variations.³⁴ Analysis on the effect of active stretching and ankle mobilization on low back pain with the patient having pronated foot were conducted. It was concluded that this convectional therapy helps in decreasing pain and increasing the quality of life with pronated foot subject.³⁵

PROPOSED ARCHITECTURE:

Several classification techniques (KNN, Logistic Regression, CART, Naïve Bayes, etc.) were used in the medical domain to get effective results. In this research paper, parameter tuning is used which will act as a catalyst in terms of code stability and will improve the accuracy rate. Being a medical centric problem statement, any negligence can result in a fatal outcome; so along with the accuracy, response time is also considered.

Here, 3 parameters are considered for improving the accuracy rate along with Random Forest as the compatibility of both are high, simple and intuitive. Data will be read from the spine dataset (310 Observations, 14 Attributes, 13 Numeric Predictors, 1 Binary Class Attribute, No Demographics).³⁶ Table 1 depicts the dataset view.

Table 1: Dataset view

pelvic_ incidence	pelvic_ tilt	lumbar_ lordosis_ angle	sacral_ slope	pelvic_ radius	degree_ spondylolis thesis	pelvic_ slope	Direct_ tilt	thoracic_ slope	cervical_ tilt	sacrum_ angle	Scoliosis _ slope
63.02	22.55	39.60	40.47	98.67	-0.25	0.74	12.56	14.53	15.30	-28.65	43.51
39.05	10.06	25.01	28.99	114.40	4.56	0.41	12.88	17.53	16.78	-25.53	16.11
68.83	22.21	50.09	46.61	105.98	-3.53	0.47	26.83	17.486	16.65	-29.03	19.22
69.29	24.65	44.31	44.64	101.86	11.21	0.36	23.56	12.70	11.42	-30.47	18.83
49.71	9.65	28.31	40.06	108.16	7.91	0.54	35.49	15.95	8.87	-16.37	24.91
40.25	13.92	25.12	26.32	130.32	2.23	0.78	29.32	12.00	10.40	-1.51	9.65
53.43	15.86	37.16	37.56	120.56	5.98	0.19	13.85	10.71	11.37	-20.51	25.94
45.36	10.75	29.03	34.61	117.27	-10.67	0.13	28.81	7.76	7.60	-25.11	26.35
43.79	13.53	42.69	30.25	125.00	13.28	0.19	22.70	11.42	10.59	-20.02	40.02
36.68	5.01	41.94	31.67	84.241	0.66	0.36	26.20	8.73	14.91	-1.70	21.43

MODULES:

Data Collection: Pre-collected data is downloaded from Kaggle which is in csv format. This dataset contains 12 columns and 310 rows.

Data Preparation/Pre-processing: After fetching the data, pre-processing is being performed to ensure the robustness of the dataset and to avoid any NA or dirty data. Cleaning the dataset may require removing duplicate, correct errors, deal with missing values, normalization, datatype conversion, etc. This technique may vary from dataset to dataset.

Train and Test the dataset: Once the correctness of the data is ensured, then the whole dataset is divided into training and testing dataset. The main goal is to make the prediction or answer the question as accurately as possible. Mostly the suggested ratio for a partition of the dataset is 70:30 but it can vary. Each iteration in the process is considered as a training step.

Choose Model: Selection of the model(s) for different task varies as per the demand of the problem statement. As we are planning to classify Lower back pain, we will be choosing any classification technique. Now classification is done to label the data which is categorized according to the attribute value.

Bagging: It resamples the original training dataset with replacement and is performed to avoid any type of fitting, let it be underfitting or overfitting. The primary objective of it is to improve the stability and accuracy of the machine learning algorithm used and it also calculates the correct and incorrect classification.

Cross-Validation: In this technique, the model is trained by the subset of the data set then the other subset is used to test the data to check whether the model got the correct pattern from the training data or not. The main job performed by cross-validation is it rechecks the result accuracy. The number of folds needs to be set.

Parameter Tuning: This is referred to as hyperparameter tuning; the model is tuned to improve the performance. The main purpose is to provide code stability and high accuracy.

Experimental Evaluation:

Table 2: Evaluation result of the proposed architecture

Mapping in parallel	mode = socket	cpus = 8	elements =5
[Tune] Result:	mtry=9	nodesize=29	acc.test.mean= 1
[Tune] Result:	mtry=2	nodesize=23	acc.test.mean=0.858

Table 2 depicts the increase in the accuracy rate after applying parameter tuning to the Random Forest classification.⁴⁰ From the WEKA survey, it was seen that the random forest accuracy rate is 71.25% but after applying the proposed technique we can see that it has been increased to 85.80%.

CONCLUSION:

For classification of Lower Back Pain, 13 attributes were considered which described the physical characteristics of the patient such as pelvic incidence, pelvic tilt, lumbar lordosis angle, sacral slope, pelvic radius, degree spondylolisthesis, pelvic slope, Direct tilt, thoracic slope, cervical tilt, sacrum angle, and scoliosis slope. The effective contribution is observed in terms of the effectiveness in the process of classification. Random Forest is used as a classification algorithm to classify the patient by labelling them whether they have LBP or not. Parameter tuning technique results in enhancement of the performance of the classification algorithm in which it is used. We witnessed the increment in the accuracy rate after applying parameter tuning.

REFERENCES:

1. Gasibat, Q., Simbak, N., Aziz, A. A., and Musa, R. M. The Effect of Strength Training Programme in the Enhancement of Trunk and Hip Muscles Activations among Healthy Females Subject. Research Journal of Pharmacy and Technology, (2017), 10(9), 2845-2850.

2. Staartjes, V. E., Quddusi, A., Klukowska, A. M., and Schröder, M. L. Initial classification of low back and leg pain based on objective functional testing: a pilot study of machine learning applied to diagnostics. European Spine Journal, (2020), 29(7), 1702-1708.

3. Lazennec, J. Y., Brusson, A., and Rousseau, M. A. Lumbar-pelvic-femoral balance on sitting and standing lateral radiographs. Orthopaedics and Traumatology: Surgery and Research, (2013), 99(1), S87-S103.

4. Ung, H., Brown, J. E., Johnson, K. A., Younger, J., Hush, J., and Mackey, S. Multivariate classification of structural MRI data detects chronic low back pain. Cerebral cortex, (2014), 24(4), 1037-1044.

5. Papadopoulos, E. C., and Khan, S. N. Piriformis syndrome and low back pain: a new classification and review of the literature. Orthopedic Clinics, (2004), 35(1), 65-71.

6. Rudwaleit, M., Metter, A., Listing, J., Sieper, J., and Braun, J. Inflammatory back pain in ankylosing spondylitis: a reassessment of the clinical history for application as classification and diagnostic criteria. Arthritis and Rheumatism: Official Journal of the American College of Rheumatology, (2006), 54(2), 569-578.

7. Kajbafvala, M., Rahmani, N., Bandpei, M. A. M., and Salavati, M. Eligibility of the movement-based classification systems in the diagnosis of patients with low back pain: A Systematic Review. Journal of Bodywork and Movement Therapies (2020).

8. Sandag, G. A., Tedry, N. E., and Lolong, S. Classification of lower back pain using K-Nearest Neighbor algorithm. In 2018 6th International Conference on Cyber and IT Service Management (CITSM) (2018) (pp. 1-5). IEEE.

9. Koh, J., Alomari, R. S., Chaudhary, V., and Dhillon, G. Lumbar spinal stenosis CAD from clinical MRM and MRI based on inter-and intra-context features with a two-level classifier. In Medical Imaging 2011: Computer-Aided Diagnosis (2011), (Vol. 7963, p. 796304). International Society for Optics and Photonics.

10. Kelkar, K. M., and Bakal, J. W. Hyper Parameter Tuning of Random Forest Algorithm for Affective Learning System. In 2020 Third International Conference on Smart Systems and Inventive Technology (ICSSIT) (2020), (pp. 1192-1195). IEEE.

11. Polyak, B. T., and Khlebnikov, M. V. Principle component analysis: Robust versions. Automation and Remote Control, (2017), 78(3), 490-506.

12. Fritz, J. M., Cleland, J. A., and Childs, J. D. Subgrouping patients with low back pain: evolution of a classification approach to physical therapy. journal of orthopaedic and sports physical therapy, (2007), 37(6), 290-302.

13. Wang, G., Xu, J., and He, B. A novel method for tuning configuration parameters of spark based on machine learning. In 2016 IEEE 18th International Conference on High Performance Computing and Communications; IEEE 14th International Conference on Smart City; IEEE 2nd International Conference on Data Science and Systems (HPCC/SmartCity/DSS), (2016), (pp. 586-593). IEEE.

14. Luo, Z., Wang, P., Li, Y., Zhang, W., Tang, W., and Xiang, M. Quantum-inspired evolutionary tuning of SVM parameters. Progress in Natural Science, (2008), 18(4), 475-480.

15. Naresh, K., Prabakaran, N., Kannadasan, R., and Boominathan, P. Diabetic Medical Data Classification using Machine Learning Algorithms. Research Journal of Pharmacy and Technology, (2018), 11(1), 97-100.

16. Tarek, S., Abd Elwahab, R., and Shoman, M. Gene expression based cancer classification. Egyptian Informatics Journal, (2017), 18(3), 151-159.

17. Tongchan, T., Pomsing, C., and Tonglim, T. Harmony search algorithm's parameter tuning for traveling salesman problem. In 2017 International Conference on Robotics and Automation Sciences (ICRAS), (2017), (pp. 199-203). IEEE.

18. Swathy, S., and Sethu, V. G. Acupuncture and lower back pain. Research Journal of Pharmacy and Technology, (2015), 8(8), 991.

19. Rojas-Domínguez, A., Padierna, L. C., Valadez, J. M. C., Puga-Soberanes, H. J., and Fraire, H. J. Optimal hyper-parameter tuning of SVM classifiers with application to medical diagnosis. Ieee Access, (2017), 6, 7164-7176.

20. Srilatha, K., and Ulagamuthalvi, V. A Comparative Study on Tumour Classification. Research Journal of Pharmacy and Technology, (2019), 12(1), 407-411.

21. Abdullah, A. A., Yaakob, A., and Ibrahim, Z. Prediction of spinal abnormalities using machine learning techniques. In 2018 International conference on computational approach in smart systems design and applications (ICASSDA) (2018), (pp. 1-6). IEEE.

22. Liu, C. B., Chamberlain, B. P., Little, D. A., and Cardoso, Â. Generalising random forest parameter optimisation to include stability and cost. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases (2017), (pp. 102-113). Springer, Cham.

23. Indraja, B., and Annapurani, K. Classification of medicines using naive bayes classifier. Research Journal of Pharmacy and Technology, (2018), 11(5), 1940-1944.

24. Lévesque, J. C., Gagné, C., and Sabourin, R. Bayesian hyperparameter optimization for ensemble learning. arXiv preprint arXiv:1605.06394, (2016).

25. Affrin, J. H. Effect of Prolonged usage of Laptop Related to Neck-Shoulder and Low Back Pain. Research Journal of Pharmacy and Technology, (2018), 11(3), 1217-1219.

26. Roupa, Z., Vassilopoulos, A., Sotiropoulou, P., Makrinika, E., Noula, E., Faros, E., and Marvaki, C. The problem of lower back pain in nursing staff and its effect on human activity. Health science journal, (2008), 2(4).

27. Gasibat, Q., Mesrati, M. H., Musa, R. M., and Zidan, A. A. Effective Recovery and Control of Chronic Low Back Pain by using Rehabilitation Exercises Therapy. Research Journal of Pharmacy and Technology, (2019), 12(9), 4313-4323.

28. Lucas, Y., Domingues, A., Driouchi, D., and Treuillet, S. Design of experiments for performance evaluation and parameter tuning of a road image processing chain. EURASIP Journal on Advances in Signal Processing, (2006), 1-10.

29. Kwon, S., Lee, S., and Na, O. Tuning parameter selection for the adaptive lasso in the autoregressive model. Journal of the Korean Statistical Society, (2017), 46(2), 285-297.

30. Jeganathan, A., Kanhere, A., and Monisha, R. A comparative study to determine the effectiveness of the mckenzie exercise and williams exercise in mechanical low back pain. Research Journal of Pharmacy and Technology, (2018), 11(6), 2440-2443.

31. Urits, I., Burshtein, A., Sharma, M., Testa, L., Gold, P. A., Orhurhu, V., ... and Kaye, A. D. Low back pain, a comprehensive review: pathophysiology, diagnosis, and treatment. Current pain and headache reports, (2019), 23(3), 1-10.

32. Blazquez, C. A., Ries, J., and Miranda, P. A. Towards a parameter tuning approach for a map-matching algorithm. In 2017 IEEE International Conference on Vehicular Electronics and Safety (ICVES), (2017), (pp. 85-90). IEEE.

33. Meenakshi, K., Safa, M., Karthick, T., and Sivaranjani, N. A novel study of machine learning algorithms for classifying health care data. Research Journal of Pharmacy and Technology, (2017), 10(5), 1429.

34. Chau, D. P., Thonnat, M., Bremond, F., and Corvee, E. Online parameter tuning for object tracking algorithms. Image and Vision Computing, (2014), 32(4), 287-302.

35. Kim, D. J., and Kim, J. H. Effects of Mckenzie Exercise on Back Pain and Physical Fitness. Group, (2019), 53(1), 53-338.

36. Sammy [Online]. Available: https://www.kaggle.com/sammy123/lower-back-pain-symptoms-dataset, (2016).

37. Song, L., Minku, L. L., and Yao, X. The impact of parameter tuning on software effort estimation using learning machines. In Proceedings of the 9th international conference on predictive models in software engineering, (2013) (pp. 1-10).

38. Lenka, S., Mishra, S., and Victor, N. IoT based Neonatal Monitoring in the ICU. Research Journal of Pharmacy and Technology, (2019), 12(6), 2885-2888.

39. Popescu, A., and Lee, H. Neck pain and lower back pain. Medical Clinics, (2020), 104(2), 279-292.

40. Venables, W. N., and Smith, D. M. The R development core team. An Introduction to R, (2003), Version, 1(0).

Received on 13.08.2020 Modified on 17.02.2021

Research J. Pharm.and Tech 2022; 15(4):1573-1578.

DOI: 10.52711/0974-360X.2022.00262