Registry and Central Database for MRI Lower Back Pain Scan Datasets

Architecture and Tools for an Open Source Database Registry of Magnetic Resonance Imaging (MRI) Scans for Automated Lower Back Pain Differential Diagnosis

Specific Aims

More than 50 million Americans suffer from chronic back pain and of those, 25 million live with daily chronic lower back pain (cLBP) and lack effective and safe non-opioid options for pain management. Despite the pipeline for new pharmacologic treatments for chronic lower back pain (cLBP) which includes agents targeting inflammation, there are currently no consistently effective and durable pharmacologic interventions for cLBP that work. Modic changes (MC) are one of the leading causes of low back pain after spinal stenosis or disc herniation. Studies suggest that there is a strong relationship between Modic changes and back pain, especially Modic change type I. Elucidation of Modic Changes etiology is hindered by the dynamic clinical presentation and multifactorial pathophysiology. MC1 and MC2 are interconvertible over time and can eventually convert to MC3 1. We intend to provide comprehensive toolkits for Lower Back Pain MRI scans processing and computational data analysis and functional testing. Datasets will provide detailed photographs of MRI scans as well as various instrumental readout from a broad range of Lower Back Pain (LBP) conditions that will be integrated onto a proposed universal Back Pain Database representative framework. At present, much of the MRI scanned data uses proprietary data formats to transform existing data into the proposed generic template via open source methods.

Specific Aim 1: Establish a central database registry. 50 to 100 persons in each of approximately 10 centers will be followed as a means to analyze Lower Back Pain patients that have had MRI scans. Data annotation, curation, and appropriate image analysis tools, image processing, querying, browsing, and downloading will be seamlessly applied in the metadata of the radiological image datasets. Application specific custom tools for querying orthopedic and anatomical information will be developed and provided.

Specific Aim 2: Document and record all patients before and after MRI scans with (a) physical examination, (b) radiographic studies, and (c) outcome measures. We intend to use a Lower Back Pain assessing score and associated inclusive and exclusive characteristics that will be noted in the patient history. Detailed history of Patient data, Imaging, Diagnostic measures and Outcome measures will be recorded at each patient visit.

Specific Aim 3: Architecture and tools for an MRI Scan of LBP Database representations with Annotations allowing for detailed querying across multiple resources. We intend to build a database architecture capable of allowing data sharing and detailed querying across multiple resources that is cloud hosted with smart- phone connectivity capabilities. 50 terabytes (TB) of storage (data is usually photos and videos), and appropriate hardware and hard-drive based installed equipment will be integrated into in-house existing servers. All data will be de-identified to meet HIPPA requirements.

Final Outcome of Success: Automated MRI Assessment to detect End Plate Damage in Lower Back Pain (LBP) diagnosis will be generated. The utilization of deep learning will help radiologists make faster and more accurate diagnoses of Lower Back Pain MRI Scan results using VoxelMorph and AUTOMAP Open Source available algorithms that have been recently demonstrated to make the process of comparing 3- D scans up to 1,000 times faster. The ability to utilize unsupervised Machine Learning will assist in automating radiological outcome and will assist with patient clinical diagnosis in a quantitative and predictable manner.