Authors:

• Thirumalesu Kudithi, School of Technology, The Apollo University, Chittoor, India.
• J Balajee, Department of Computer Science and Engineering, Mother Theresa Institute of Engineering and Technology, Palamaner, Chittoor, Andhra Pradesh, 517408, India.
• R Sivakami, Department of Computer Science and Engineering, Sona College of Technology, Salem, 636005, India.
• T R Mahesh, Department of Computer Science and Engineering, JAIN (Deemed-to-Be University), Bengaluru, 562112, India.
• E Mohan, Department of ECE, Saveetha School of Engineering, SIMATS, Chennai, Tamilnadu, India.
• Suresh Guluwadi, Adama Science and Technology University, Adama, 302120, Ethiopia. suresh.guluwadi@astu.edu.et.

Introduction

The diagnosis of Ehlers-Danlos Syndrome (EDS), a group of connective tissue disorders characterized by joint hypermobility, skin hyperextensibility, and tissue fragility, often relies on precise joint angle measurements. Traditional goniometry, although widely used, presents challenges in achieving consistent and accurate measurements due to the subjective nature of manual assessments.

Research Context and Objective

The objective of this research is to introduce and validate HybridPoseNet, a hybrid deep learning model that integrates Convolutional Neural Networks (CNNs) with HyperLSTM units, aimed at addressing the complexities of dynamic joint motion in EDS patients. By enhancing the accuracy of goniometric assessments, this model seeks to provide a more reliable and standardized method for diagnosing joint hypermobility, thereby improving the overall diagnostic process for EDS.

Methodology

HybridPoseNet leverages the strengths of CNNs in feature extraction and the capabilities of HyperLSTM units in capturing temporal dependencies. The CNN component processes input images to extract relevant features related to joint angles, while the HyperLSTM units analyze the sequential data to maintain consistency over time. This hybrid approach allows the model to adapt to the dynamic and often unpredictable nature of joint movements in EDS patients, offering a more comprehensive analysis than traditional methods.

The model was trained and tested on a dataset comprising images and videos of joint movements from EDS patients, with annotations provided by expert clinicians. The performance of HybridPoseNet was compared with traditional manual goniometry techniques to evaluate its accuracy and reliability.

Results

Preliminary results demonstrate a significant improvement in the precision of joint angle measurements using HybridPoseNet. The model achieved approximately a 20% increase in correlation with manual goniometry, indicating a substantial enhancement in diagnostic accuracy. This improvement is particularly noteworthy given the inherent challenges of measuring joint angles in patients with hypermobility disorders.

Furthermore, the model's ability to generalize across different patients and maintain time consistency in measurements underscores its potential as a reliable tool for EDS evaluation. The integration of deep learning techniques in this context represents a significant advancement in the field of medical diagnostics, offering a more objective and reproducible method for assessing joint hypermobility.

Conclusion

HybridPoseNet marks a pioneering step in the application of AI for the diagnosis of connective tissue disorders like Ehlers-Danlos Syndrome. By combining the power of CNNs and HyperLSTM units, this hybrid deep learning model offers a promising solution to the limitations of traditional goniometry. The results of this study highlight the potential of AI-driven approaches in enhancing diagnostic accuracy and standardization, paving the way for more effective and reliable medical assessments in the future.