iSpine - Spine Disc Prognosis

iSpine

Spinal Imaging and Image Analysis

Mission

To use simple image based biomarkers to perform prognosis of spinal diseases using non-invasive procedures.

Challenges

  1. MRI’s lack of ionizing radiation and high contrast makes it difficult to perform diagnosis.
  2. No current gold standard for diagnosing Degenerative Disc Disease (DDD).[1]

Research

Computer aided diagnosis (CAD): procedures that aid in the interpretation of medical imaging.

  • Scans the image then highlights possible problematic areas that can be used for diagnosis.
  • Allows the causes of lower back pain to be diagnosed with radiographs/ MRI/CT.[1]


Computer aided prognosis (CAP): produces predictions of likely outcomes.

  • Allows one to test for several outcomes from the disease that can be accounted for in treatment.
  • Estimates recovery times/recovery probabilities.[1]


classes of discs classes of discs


Classes of Discs

Approach



Longitudinal Study


We investigated whether a) the image texture features quantified from MRI could be appropriate markers for diagnosis of DDD, b) prognosis of inter-vertebral disc loss, computed and validated in the lumbar region of the spine.[1]

roc curve


Figure 1. ROC for diagnosis of DDD for individual texture markers and combination marker.[1]

roc curve


Figure 2. ROC for prognosis of inter-vertebral disc loss based on individual texture markers and combination marker. [1]

Validation

A longitudinal study was conducted that consisted of 65 subjects in the London, Canada area. A T2 weighted sequence was used to scan all the subjects at both visits. Manual disc segmentation was used to focus on the diagnosis and prognosis of the project. Specific texture features were quantified from the gray level co-occurrence matrix (GLCM) of the segmented disc region. The volume was quantified and the same texture features were quantified from the GLCM of the segmented disc volume as well. Using linear discriminant analysis the prognostic combination marker was developed as a weighted sum of individual texture markers. The area under the ROC curve was used to compute the ability of the GLCM texture features to diagnose DDD and to predict inter-vertebral disc loss.[1]

Literature

1. Shuo Li, Amy Lin, Kengyeow Tay, Walter Romano, and Said Osman. Prognosis of Intervertebral Disc Loss from Diagnosis of Degenerative Disc Disease. SPIE Medical Imaging. 2015.

Collaborators




st. joseph's hospital
lawson health research institute
robarts research
london health sciences centre
general electric
western university
digital imaging group of london