An Intelligent Decision-Making Method for Rehabilitation Assistive Device Matching inSpinal Cord Injury Patients Based on K-Modes Clustering and Weighted K-Nearest Neighbor
Li Sujiao1,2#*, Shao Jiang1#, Liu Qi1, Chen Shizheng3, Yu Hongliu1,2#
1(Institute of Rehabilitation Engineering and Technology, University of Shanghai for Science and Technology, Shanghai 200093, China) 2(Shanghai Engineering Research Center of Assistive Devices, Shanghai 200093, China) 3(China Rehabilitation Research Center, Beijing 100068, China)
Abstract：The injury segments, nature, and limb muscle strength characteristics of patients with spinal cord injuries are complex and variable. Inappropriate rehabilitation assistive device fitting has long hindered the overall rehabilitation process and efficacy for such patients. This study proposed a smart assistive device fitting decision-making method (KKA) that integrated K-modes clustering with weighted KNN, aiming to enhance the precision and efficiency of assistive device fitting. The study collected clinical data from 600 spinal cord injury patients across different hospitals. A 13-member expert panel comprising cross-disciplinary, cross-professional, and frontline assistive device fitting expert′s consensus determined nine feature attributes, including injury segment, injury nature, limb muscle strength and different fitting schemes. The KAA method comprised three key steps: (1) using entropy weighting and the analytic hierarchy process to determine feature weights and construct a weighted feature space; (2) optimizing the case library structure using the K-modes clustering algorithm to effectively reduce classification bias; and (3) combining weighted KNN case retrieval with a rule-based inference model to generate personalized adaptation prescriptions. The algorithm′s performance was validated using 600 static medical records and 40 medical records tracked dynamically over a one-month period. The study demonstrated that KAA performed exceptionally well in the adaptation of four types of assistive devices, achieving average accuracy and recall rates of 94.6-97.9% and 89.5-97.2%, respectively, indicating that the KAA algorithm possessed high effectiveness and reliability in the field of rehabilitation assistive product recommendation. Additionally, dynamic clinical tracking assessments indicated that the optimal threshold for SCIM improvement in the KAA algorithm was 24.2%, while the optimal threshold for ICF improvement was 2.3%. The AUC values for both SCIM and ICF improvement levels were greater than 0.7, indicating that the KAA method had high diagnostic value in assessing the effectiveness and applicability of rehabilitation assistive product fitting for patients. In conclusion, the KAA method demonstrated significant advantages in terms of fitting accuracy, efficiency, and rehabilitation outcomes, and is expected to provide reliable technical support for precise rehabilitation in spinal cord injury patients.
[1] Morone G, Pirrera A, Iannone A, et al. Development and use of assistive technologies in spinal cord injury: a narrative review of reviews on the evolution, opportunities, and bottlenecks of their integration in the health domain [J]. Healthcare-Basel, 2023, 11 (11):1646. [2] Bradbury EJ, McMahon SB. Spinal cord repair strategies: why do they work [J]? Nature Reviews Neuroscience, 2006, 8(7): 644-653. [3] Seijas V, Schrepfer L, Posada AM, et al. Evidence-based recommendations for the rehabilitation and management of the ageing population with spinal cord injury: a systematic review of clinical practice guidelines[J]. European Journal of Physical and Rehabilitation Medicine, 2024, 60(3):433-444. [4] Wyndaele M, Wyndaele JJ. Incidence, prevalence and epidemiology of spinal cord injury: what learns a worldwide literature survey?[J] Spinal Cord, 2006, 9(44): 523-529. [5] Scherer MJ. The change in emphasis from people to person: introduction to the special issue on assistive technology[J]. Disability and Rehabilitation, 2002, 24(1-3):1-4. [6] Davies A, De Souza LH, Frank AO. Changes in the quality of life in severely disabled people following provision of powered indoor/outdoor chairs [J]. Disability and Rehabilitation, 2003, 6(25): 286-290. [7] 李晞,吴小高. 我国残疾人辅助器具服务工作的现状及展望 [J]. 残疾人研究, 2016 (3): 43-47. [8] Abbas SR, Seol H, Abbas Z, et al. Exploring the role of artificial intelligence in smart healthcare: a capability and function-oriented review[J]. Healthcare (Basel), 2025, 13(14):1642. [9] Agarwal S, Chin WY, Vasudevan L, et al. Digital tracking, provider decision support systems, and targeted client communication via mobile devices to improve primary health care[J]. Cochrane Database of Systematic Reviews, 2025, 4(4):CD012925. [10] Shortliffe EH, Axline SG, Buchanan BG, et al. An artificial intelligence program to advise physicians regarding antimicrobial therapy[J]. Computers and Biomedical Research, an International Journal, 1973, 6(6):544-560. [11] Montaña D, Campos-Roca Y, Pérez CJ. A diadochokinesis-based expert system considering articulatory features of plosive consonants for early detection of Parkinson′s disease[J]. Computer Methods and Programs in Biomedicine, 2018, 154:89-97. [12] Ali L, Niamat A, Khan L JA, et al. An optimized stacked support vector machines based expert system for the effective prediction of heart failure[J]. IEEE Access, 2019, 7: 54007-54014. [13] Rivera RF, Pagaduan RA, Caliwag J.A, et al. A mobile expert system using fuzzy logic for diagnosing kidney diseases[C]// Proceedings of the 2nd International Conference on Information Science and Systems. Tokyo: ICISS, 2019: 161-165. [14] López-Canay J, Casal-Guisande M, Pinheira A, et al. Predicting COPD readmission: an intelligent clinical decision support system[J]. Diagnostics (Basel). 2025, 15(3):318. [15] De Ramón Fernández A, Ruiz Fernández D, Prieto Sánchez MT. A decision support system for predicting the treatment of ectopic pregnancies[J]. International Journal of Medical Informatics, 2019, 129:198-204. [16] Lin CT, Chang KC, Lin CL, et al. An intelligent telecardiology system using a wearable and wireless ECG to detect atrial fibrillation[J]. IEEE Transactions on Information Technology in Biomedicine, 2010, 3(14): 726-733. [17] Lourenço JW, de Jesus PAC, Schaefer JL, et al. Challenges and strategies for the development and diffusion of assistive technologies[J]. Disability and Rehabilitation: Assistive Technology, 2025, 23:1-14. [18] Banyai AD, Brişan C. Robotics in physical rehabilitation: systematic review[J]. Healthcare (Basel). 2024, 29;12(17):1720. [19] 范秀琴,喻洪流,杨宇辉,等. 基于案例推理-规则推理混合推理的脊髓损伤智能辅具适配系统[J].中国康复医学杂志,2022,37(8):1084-1088. [20] Duankhan P, Sunat K,Soomlek C. An adaptive smell agent optimization with binomial crossover and linnik flight for engineering optimization problems[C]// 28th International Computer Science and Engineering Conference (ICSEC). Khon Kaen: IEEE, 2024: 1-6. [21] Roberts TT, Leonard GR, Cepela DJ. Classifications in brief: American spinal injury association (ASIA) impairment scale[J]. Clinical Orthopaedics and Related Research, 2017,5(475):1499-1504. [22] Cheng Q. Structure entropy weight method to confirm the weight of evaluating index[J]. Systems Engineering-Theory & Practice, 2010, 7(30):1225-1228. [23] 梁吉业,白亮,曹付元.基于新的距离度量的K-Modes聚类算法[J].计算机研究与发展,2010,47(10):1749-1755. [24] Yang ZG, Li HQ, Zhu LP, et al. A case based method to predict optimal k value for k-NN algorithm[J]. International Journal of Fuzzy Systems, 2017,1(33):55-65. [25] 桑应宾.基于K近邻的分类算法研究[D]. 重庆:重庆大学,2009. [26] 官赛萍,靳小龙,贾岩涛,等.面向知识图谱的知识推理研究进展[J].软件学报,2018,29(10):2966-2994. [27] Boddu AS, Jan A. A systematic review of machine learning algorithms for breast cancer detection[J]. Tissue Cell, 2025,95:102929. [28] 王越,张春青,王悦,等.国内外医用康复器械分类与管理现状[J].中国医疗设备,2022,37(3):163-166. [29] 贾俊平, 何晓群, 金勇进. 统计学 [M]. 第9版. 北京: 中国人民大学出版社, 2025. [30] Hu F, He K, Wang C, et al. STRFLNet:spatio-temporal representation fusion learning network forEEG-based emotion recognition[J]. IEEE Transactions on Affective Computing, 28 Oct, 2025 [Epub ahead of print]. [31] Banville H, Chehab O, Hyvarinen A, et al. Uncovering the structure of clinical EEG signals with self-supervised learning[J]. Journal of Neural Engineering, 2021, 18(4): 046020.
