Techniques, algorithms and database for the processing of multiscanning-EMG signals

• Autores: Mahima Kallingal Muraleedharan
• Directores de la Tesis: Armando Malanda Trigueros (dir. tes.), Javier Navallas Irujo (codir. tes.)
• Lectura: En la Universidad Pública de Navarra ( España ) en 2026
• Idioma: inglés
• Número de páginas: 106
• Programa de doctorado: Programa de Doctorado en Biotecnología por la Universidad Pública de Navarra
• Enlaces
  • Tesis en acceso abierto en: Academica-e
• Resumen

  • This thesis focuses on developing a framework for the recording, processing, and analysing motor unit potential (MUP) scans obtained from multiscanning-EMG technique. Starting from the raw signal recording, it describes the experimental setup, recording protocol, and signal processing steps that transform muscle activity into structured MUP-scans suitable for quantitative analysis of motor unit (MU) parameters. MU parameters such as MUP duration, MU territory extent and MU fractions were automatically estimated. To support validation, a dedicated MATLAB tool, the DTF-marking tool was developed to allow neurophysiologist to interactively mark motor unit duration, territory extent, and fractions. These annotations served as the ground truth used for evaluating the automatic algorithms. For the MUP duration, a correlation-based algorithm was implemented and validated with the ground truth markings. This approach outperformed Nandedkar method that was adapted to work with Multiscanning data, particularly in the start marker detection. The algorithm excluded silent areas and cannula regions, thereby providing more physiologically meaningful results. For the MU territory extent, two complementary methods were developed: a double threshold method (M1) and an image processing method (M2) that excluded the cannula region to reduce artefacts. M1 showed close agreement with the ground truth for both the start and the end markings. M2 performed comparably for the start markers but showed a significant difference for the end markers. MU fraction detection was based on the image processing algorithm, which achieved a sensitivity of 78.6 %. Its performance was evaluated using two analyses: first, using the entire dataset of MUP-scans (A1), and second, with MUP-scans consisting of an equal number of fractions in automatic detection and ground truth analysis (A2). In addition, depth and signal quality were analysed for the detected fractions. The results showed that the fractions located in deeper and central regions were more numerous and had a higher signal-to-noise ratio. The superficial fractions were longer but exhibited greater variability. Despite challenges such as the exclusion of far-field signal and occasional false positives, this framework provides the first reproducible approach for automatic quantification of the MU structure using multiscanning-EMG. The approach creates new opportunities for applying the multiscanning technique in both research and clinical practice, where it can complement conventional EMG. By establishing a reliable approach to analyzing motor unit architecture, this work contributes to deepen physiological understanding and paves the way for future clinical applications. In particular, it has the potential to advance the diagnosis, monitoring, and treatment of neuromuscular disorders.