Stream/Area:
-
Subject:
-
Paper ID:
IJSREM69140
Volume:
10 |
11 |
Nov 2026
Pages:
1-6
Publication Frequency:
Monthly (12 issues per year)
License:International Journal of Scientific Research in Engineering and Management follows the Creative Commons Attribution 4.0 (CC BY 4.0) License, allowing free use and distribution with proper attribution.

Department of Information Science and Engineering
Jawaharlal Nehru New College of Engineering, Shivamogga.
Department of Information Science and Engineering
Jawaharlal Nehru New College of Engineering, Shivamogga.
Department of Information Science and Engineering
Jawaharlal Nehru New College of Engineering, Shivamogga.
Department of Information Science and Engineering
Jawaharlal Nehru New College of Engineering, Shivamogga.
Department of Information Science and Engineering
Jawaharlal Nehru New College of Engineering, Shivamogga.

Department of Information Science and Engineering
Jawaharlal Nehru New College of Engineering, Shivamogga.
This paper proposes NMD-CDSS, an explainable multimodal clinical decision support system for analysing neuromuscular disorders. The system uses three different Deep Learning based modal specific pipelines to classify the data. The MRI pipeline uses the ResNet-18 model to classify the image slices into either healthy or muscular dystrophy. The EMG pipeline uses the CNN-BiLSTM model to classify the EMG signals from a needle into healthy, myopathy or neuropathy. The clinical pipeline uses a fine-tuned BioBERT model to classify the clinical text into one of the five diagnostic categories. Individual predictions are visualised using GradCAM, temporal gradient and token attribution. The web interface developed using Flask provides individual predictions, confidence scores and an overall summary report. The results obtained for the various datasets used were 88.50% (±7.21% on patient data) accuracy on the MRI data (87.06% (±5.20%) accuracy on slice data), 96.08% validation accuracy on EMG data (macro-F1 of 0.9532) and 94.74% test accuracy on clinical data (macro-F1 of 0.9463). An attention based fusion module is used for experimentation purposes only as the datasets are not available in a patient paired format. None of these results can be used as evidence of effectiveness in a clinical setting. Real data and clinical validation will be required before being used for diagnostic purposes.
KEYWORDS
This paper proposes NMD-CDSS, a modular explainable multimodal research prototype. The system uses the ResNet-18 model for MRI, CNN-BiLSTM for EMG and BioBERT for clinical text. The web interface was created using Flask. The explanations were provided using Grad-CAM, temporal and token attributions. Each model provided promising results on their own data. (88.50% accuracy on patient data, 96.08% on window data, 94.74% on test data). However, the evaluation varied. The EMG and text results had issues with leakage and the size of the data. The fusion was just for experimentation and has not been validated as there was no paired patient data. Larger individual datasets and clinical validation is required for any practical use.
[1] M. Mancuso, A. Colitta, M. Lavorato, P. Van den Bergh, J. Kirschner, and C. Kornblum, “The most bothersome symptoms in neuromuscular diseases: the ERN EURO-NMD survey,” Orphanet J. Rare Dis., vol. 20, no. 221, pp. 1–10, 2025, doi: 10.1186/s13023-025-03742-z. [2] S. Jose, S. T. George, M. S. P. Subathra, V. S. Handiru, P. K. Jeevanandam, U. Amato, and E. S. Suviseshamuthu, “Robust classification of intramuscular EMG signals to aid the diagnosis of neuromuscular disorders,” IEEE Open J. Eng. Med. Biol., vol. 1, pp. 235–242, 2020, doi: 10.1109/OJEMB.2020.3017130. [3] R. K. Mani and B. Nagaraj, “Diagnosis and classification of neuromuscular disorders using Bi-LSTM optimized with grey wolf optimizer for EMG signals,” Sci. Rep., vol. 15, Art. no. 19274, 2025, doi: 10.1038/s41598-025-03766-2. [4] A. Gorenshtein, Y. Weisblat, M. Khateb, G. Kenan, I. Tsirkin, G. Fayn, S. Geller, and S. Shelly, “AI-based EMG reporting: A randomized controlled trial,” J. Neurol., vol. 272, p. 586, 2025, doi: 10.1007/s00415-025-13261-3. [5] M. Hosny, A. M. Elshenhab, and A. Maged, “Explainable AI-based method for brain abnormality diagnostics using MRI,” Biomed. Signal Process. Control, vol. 100, p. 107184, 2025, doi: 10.1016/j.bspc.2024.107184. [6] Y. Sang and J. Xu, “Evaluation of motor neuron injury in ALS by different parameters of diffusion tensor imaging,” IEEE Access, vol. 8, pp. 72381–72392, 2020, doi: 10.1109/ACCESS.2020.2985064. [7] S. H. Elwan, M. S. Hashem, M. M. Ghanem, S. M. Ahmed, H. S. Mohammed, and B. B. Elsayed, “Role of neuromuscular ultrasound in the evaluation of amyotrophic lateral sclerosis,” Egypt. J. Neurol. Psychiatr. Neurosurg., vol. 61, no. 20, pp. 1–9, 2025, doi: 10.1186/s41983-025-00951-y. [8] F. Hellmann, I. Ristow, L. Well, S. Lohse, M. Anokhin, M. Kuhlen, A. André, and A. Harder, “Artificial intelligence-based tools for precision diagnosis and treatment of neurofibromatosis type 1 associated peripheral and central glial tumors,” Orphanet J. Rare Dis., vol. 20, p. 551, 2025, doi: 10.1186/s13023-025-04093-5. [9] M. Talha, M. Kyrarini, and E. Ali, “Physiological features for classification of different types of peripheral neuropathy using multimodal wearable sensors,” IEEE Access, vol. 14, pp. 8660–8671, 2026, doi: 10.1109/ACCESS.2026.3652722. 10.S. Bista, A. Coffey, M. Mitchell, A. Fasano, S. Dukic, T. Buxo, E. Giglia, M. Heverin, et al., “Abnormal EEG spectral power and coherence measures during pre-motor stage in amyotrophic lateral sclerosis,” IEEE Trans. Neural Syst. Rehabil. Eng., vol. 33, pp. 232–242, 2025, doi: 10.1109/TNSRE.2024.3523109.
Publication Date
11/17/2026
Peer Review Type
Double-Blind Peer Review
Review Rounds: 2
Documents
Article Views
0
Total views since publication
PDF Downloads
0
Total downloads since publication
Geographic Reach
5
Digital Preservation: This article is permanently archived on Zenodo and complies with all data preservation standards across major indexing platforms.
Plagiarism Screening
This manuscript has been checked for plagiarism using TurnitinPlagiarismChecker software.
Conflict of Interest
The authors declare no conflict of interest related to this research work.
Funding Information
No funding information available for this research.
Related research in the same subject area
Loading similar articles...