Development of an Efficient Artificial Intelligence based Clinical Assessment System for patients with Lumbar Prolapsed Intervertebral Disc and its Correlation with Magnetic Resonance Imaging findings
Description
The dataset used for developing and validating the AI-based clinical assessment system for lumbar prolapsed intervertebral disc (PIVD) was constructed through a phased and structured process, beginning with clinical assessments of 12 patients with MRI-confirmed PIVD to identify the most pertinent diagnostic parameters. These parameters, including demographic, anthropometric, symptomatic, and neurological findings, served as the foundation for subsequent data compilation. Expert consensus was then obtained from 12 specialists (9 physiotherapists, 2 neurologists, and 1 neurosurgeon) through a Delphi-based survey, where each parameter was rated on a 5-point Likert scale, and those with ≥70% agreement were retained or refined for inclusion. Based on these validated parameters, clinical and demographic data were collected from 100 patients with MRI-confirmed lumbar PIVD to create the primary dataset for training and testing the AI system, with variables such as age, gender, BMI, occupation, limb length discrepancy, duration of symptoms, pain, radicular signs, neurological findings, and standard clinical tests recorded and coded into binary, ordinal, or normalized continuous formats to facilitate machine learning processing. Finally, a dataset of 29 patients, assessed using the developed AI system and their corresponding MRI diagnoses, was compiled to evaluate the agreement between the AI-based diagnosis and MRI findings, using the same coding approach as the primary dataset to ensure consistency. Together, these datasets formed a comprehensive and standardized framework for parameter determination, AI model training, and diagnostic correlation with MRI as the reference standard.