Ab1524 a novel accelerometry-based method for early detection of peripheral neuropathy associated with systemic autoimmune rheumatic diseases

Background Systemic autoimmune rheumatic diseases (SARDs) such as systemic lupus erythematosus (SLE), rheumatoid arthritis (RA), primary Sjögren’s syndrome (SS), and systemic sclerosis (SSc) occasionally affect the peripheral nervous system. Nerve conduction studies (NCS) can almost certainly confirm the diagnosis of SARD-associated neuropathy. The NCS method is usually considered the gold standard for neuropathy assessment, although this method is based on physicians’ knowledge and experience. However, it is painful for patients, long-lasting, prone to errors, and can’t be used for routine follow-up. There is an urgent need for potential alternative diagnostic screening methods for implementation in everyday practice. Wearable sensor devices, such as accelerometers, are instruments that can be utilized to acquire data during different activities. Their size and wirelessness, lower cost, portability, and use in home-based and real-life situations are a few of the advantages. Objectives To evaluate a telemedicine wearable device with a machine learning algorithm that can be used as a screening and tracking tool for SARD-related neuropathy. Methods A monocentric, diagnostic study was conducted at the Institute of Rheumatology in 2020. The participants were healthy volunteers and SARD patients who had suspected neuropathy. The participant started with the NCS examination; electrodes were placed on the limbs, and amplitude, latency, and conduction velocity of n. medianus, n. ulnaris, n. peroneus, n. tibialis, and n. supraspinatus (motor and sensory fibres) were measured. The novel method consists of four wearable sensors placed over the middle of the hands and feet. The subject performed six exercises with open and closed eyes. Raw data was sent through the Bluetooth connection from the sensors to the tablet and then via WiFi connection to the central server for further analysis. A wearable device uses a specific mathematical algorithm that transforms signals from the accelerometer and gyroscope into specific values. The outcome is defined as a binary variable: whether or not neuropathy exists. Results The study included 23 participants (9 ♂ and 14 ♀), 11 with SARDs (45.8%). Of the total number of SARD participants, 8 (72.7%) had neuropathy confirmed with the NCS examination. The features (such as acceleration or power) obtained with signal processing were examined, and only those that can be used to discriminate SARD-related neuropathy are presented (Table 1). The model for binary classification was developed and presented in Table 2. As shown, the sensitivity and specificity are satisfactory, but the confidence intervals are still wide. Positive predictive value is significantly lower compared to negative predictive value. Conclusion Wearable sensors represent accurate and promising technology for the diagnosis of neuropathies related to SARDs. Further studies are needed to evaluate the true accuracy of the technology. Table 1. The features used to discriminate neuropathy Neuropathy (NCS) p value No (15) Yes (8) EXC 1 heel-toe walk SDP LL Heels 10hz 0.01±0.01 0.01±0.003 0.02 SDP Acc Norm LL Heels 0.1±0.05 0.06±0.03 0.01 EXC 2 tandem walk SigPow Acc Norm RL Min 1.5±0.2 1.33±0.13 0.07 EXC 3 heel-knee test CorrSeg Acc Norm LL ClE 0.03±0.03 0.01±0.02 0.03 DiffSig Acc Norm LL_LL OpE 0.06±0.1 0.02±0.015 0.2 DiffSig Acc Norm RL_RL OpE 0.05±0.1 0.03±0.06 0.5 CorrSeg Acc Norm LL ClE Std 0.7±0.23 0.44±0.16 0.03 EXC 4 Romberg test Var Acc Norm LL CloseEye 0.1±0.15 0.28±0.3 0.04 EXC 5 postural tremor StatPos AbsDiff RA Max 0.02±0.01 0.01±0.01 0.07 Var Acc Norm LA BefAft 1.2±0.6 0.71±0.2 0.05 EXC 6 finger-nose DiffSig Acc Norm RA_RA OpE 1.0±0.02 0.97±0.1 0.01 Exp Acc Norm RA TotPow 0.003±0.01 -0.02±0.03 0.01 Table 2. Validation of new proposed model for neuropathy screening Model NCS - (n=15) NCS + (N=8) Sn Sp PPV NPV 1 FT EXC 1 + 4 FT EXC 3 WS + 3 7 0.875 (0.466-0.993) 0.800 (0.513-0.946) 0.700 (0.513–0.946) 0.923 (0.621–0.996) WS - 12 1 Abb. FT – feature, WS – wearable sensors, NCS – nerve condution studies Acknowledgements The authors acknowledge the company “DIVS Neuroinformatics” for providing equipment to us, for signal processing and data analysis. We also acknowledge the collaboration of patients and other rheumatologists at the Institute of Rheumatology for participating in this study. Disclosure of Interests Zoran Veličković: None declared, Slavica Pavlov Dolijanovic Speakers bureau: Pfizer, Novartis, Eli Lilly, Abbvie, Nina Tomonjic: None declared, Saša Janjić: None declared, Biljana Stojic: None declared, Goran Radunovic Speakers bureau: Pfizer, Novartis, Eli Lilly, Abbvie, Grant/research support from: Novartis, Abbvie.