Ibrahim Anagheem1

1 St. Petersburg Electrotechnical University “LETI” (St. Petersburg, Russia)
1 anaghemibrahem66@gmail.com

Problem statement. Current methods for diagnosing musculoskeletal disorders, especially when assessing posture in the frontal and sagittal planes, are limited in analyzing dynamic conditions, such as walking or other forms of natural movement. Static methods do not reflect the actual behavior of the spine during movement, which reduces the accuracy of diagnosing pathologies such as scoliosis, pelvic asymmetry, and hyperlordosis. Additional difficulties arise when analyzing video due to camera shake, changes in viewing angles, and frame quality. Optical tracking systems are also subject to perspective distortion, which complicates the quantitative assessment of posture parameters. Objective. Development of an improved algorithm for measuring and classifying spinal deformities during movement. The main goal is to provide a stable coordinate system adaptive to patient movement, automatic and accurate determination of spinal markers, as well as quantitative analysis of curvature and asymmetry of the spinal column. Particular attention is paid to the correct processing of video data, extraction of key frames, compensation of jitter, and the use of neural network methods for classification of the deformation type. Results. A complex algorithm was developed that includes several interrelated stages. At the pre-processing stage, video material is standardized (up to MP4 format, 1080p), jitter compensation is performed using stabilization, and key frames are selected using computer vision methods based on convolutional and recurrent neural networks (CNN and RNN). To ensure the accuracy of the analysis, reduced compression is used, which allows maintaining the high image quality necessary for accurate assessment of textures and anatomical landmarks of the spine. Next, the camera is initialized and the coordinate axes are synchronized; if necessary, the video is matched with biomechanical motion signals. Deformation measurements are performed based on the YOLOv6 algorithm, with subsequent refinement of the results using CNN and LSTM models, which ensures accurate extraction of spine coordinate data under motion conditions. The post-processing stage includes noise filtration, smoothing and refinement of the spinal curvature boundaries. Then, the detected deformities are automatically classified, including the determination of the type and severity of such pathologies as scoliosis, kyphosis and lordosis. The final stage is the visualization of the results in the form of curvature and deviation graphs, as well as the generation of a structured report with a quantitative assessment and recommendations for clinical use. Practical significance. The proposed algorithm opens up new prospects in the diagnosis of posture disorders, providing the ability to accurately measure in conditions of natural patient movement without the need for expensive equipment or strictly controlled conditions. Its use allows for an objective assessment of the effectiveness of various orthopedic devices, such as corsets, insoles and exoskeletons, as well as monitoring the progression of spinal pathologies at early stages. In addition, the algorithm can be adapted to individualize the rehabilitation process depending on the dynamics of changes in spinal curvature, which makes it a valuable tool in clinical practice. Due to its versatility and efficiency, the development is easily integrated into mobile and telemedicine solutions, which significantly expands the scope of its application. In total, this approach represents an important step towards the creation of intelligent diagnostic systems of the new generation, combining high accuracy, adaptability and accessibility.to biological tissues associated with the specifics of MIS interventions and allow to improve the quality and safety of surgical care.

Anagheem Ibrahim. An improved algorithm for the analysis and diagnosis of spinal deformities using dynamic image processing and machine learning. Biomedicine Radioengineering. 2025. V. 28. № 5. P. 85–90. DOI: https:// doi.org/10.18127/j15604136-202505-17 (In Russian)

1. Anagim Ibragim, Smirnova L.M., Yuldashev Z.M. Pribor dlya issledovaniya pozy cheloveka vo frontal'noj ploskosti pri hod'be. Biomedicinskaya radiotekhnika. 2025. T. 28. № 2. S. 56–62. DOI: https:// doi.org/10.18127/j15604136-202502-09 (In Russian).
2. Cheng Y., Liu Q., Zhao C., Zhu, X., Zhang G. Razrabotka i realizaciya mediapleera na baze FFmpeg. In: Software Engineering and Knowledge Engineering: Theory and Practice: V 2. Springer Berlin Heidelberg. 2012. P. 867–874.
3. Dunaev A.V. Metod i ustrojstvo ocenki funkcional'nogo sostoyaniya mikrocirkulyatorno-tkanevyh sistem organizma cheloveka na osnove mnogoparametricheskoj opticheskoj diagnostiki.. Vestnik rossijskih universitetov. Ser.: Radioelektronika. 2020. T. 23. № 4. S. 77–91 (In Russian).
4. Movellan J.R. Uchebnoe posobie po metodike Gabora Fil'try. Dokument s otkrytym iskhodnym kodom. 2002. 40. S. 1–23 (In Russian).
5. Tanure M.K., Pinejro A.P., Olivejra A.S. Ocenka nadezhnosti izmerenij ugla Kobba s ispol'zovaniem ruchnyh i cifrovyh metodov. The Spine. 2010. V. 10(9). S. 769–774 (In Russian).