JMIR Publications

ABSTRACT

Background:

Low back pain (LBP) stands as a primary worldwide disability factor which impacts people of all ages while showing increasing prevalence among younger demographics. Different patient symptoms and treatment responses exist despite identical MRI results, which makes it difficult to determine between surgical and medical interventions.

Objective:

The study created SPOILS (Software to Predict Outcome in Lumbar Spondylosis) tool as an artificial intelligence-based decision-support tool which merges Clinical Intelligence and Radiomics to generate customized therapy plans for LBP patients.

Methods:

The SPOILS system uses deep learning models to perform automated segmentation which enables the extraction of geometrical parameters that include disc height, disc width, vertebrae height, vertebrae width, canal diameter, disc height index, signal intensity, and disc volume. The creation of a labeled dataset utilized expert-verified Pfirrmann and spondylosis severity gradings to address the clinical issues stemming from manual grading variability and subjectivity. The machine learning algorithms use this combined dataset to predict results and recommend personalized treatment plans.

Results:

The DeepLabV3+ segmentation model with ResNet50 encoder reached 95.5% accuracy, which increased to 98.7% after 8-fold cross-validation and simultaneously improved precision (96.95%), recall (97.1%), Dice coefficient (96.9%), and IoU (94.8%). The CNN with MobileNetV2 achieved 97.84% accuracy and 96.76% IoU for spondylosis severity prediction after cross-validation. The Gradient Boost classifier demonstrated the best results with geometrical data by achieving 91.65% accuracy and 84.59% IoU.

Conclusions:

SPOILS introduces an innovative method to customize LBP treatment through the combination of AI technology with radiological data and clinical expertise.