I-Corps: Standardized MRI Interpretation for Low Back Pain Diagnosis

Suny At Buffalo, Amherst NY

Investigators

Abstract

The proposed project incorporates novel machine learning based methods from a large clinical Magnetic Resonance Imaging (MRI) dataset. These methods incorporate high- and low-level imaging features including intervertebral disc location, shape, and intensity. Moreover, these methods model the disc structure as a Markov chain to enable neighborhood information utilization. In addition to the imaging features, these methods utilize patient meta-data including patient age, height, weight, pain and disability score, patient history, and physical exam findings. The availability of these features provides information for the diagnosis of low back pain. These methods generate unbiased and reproducible interpretation. This technology is capable of providing standardized, unbiased, and reproducible MRI interpretation. Many people are affected by low back pain and it is the most common reason behind job-related disability. It is a prominent chronic disease that causes major disruption in people's lives. An annual estimate of at least $50 billion is spent in the United States on diagnosis and related rehabilitation of low back pain patients. Moreover, the associated individualized treatment and rehabilitation cost is significant and often requires special pre-approval to undergo treatment from the insurance providers responsible for paying for health care costs. The most common current clinically approved standard for low back pain diagnosis is MRI testing and the diagnostic interpretation of MRIs is highly subjective. All subsequent therapeutic recommendations are based on the subjective report. This technology may be able to provide the MRI interpretation based on a standardized protocol will significantly impact the treatment plan outcome and minimize overtreatment.