Disorders of the optic nerve (ON) account for a significant percentage of the 20 most impactful ophthalmological conditions. Collectively, diseases of the ON are the number one cause of irreversible blindness worldwide, and present serious public health concerns in the U.S. Consider, for example, that glaucoma impacts more than three million Ameri- cans and costs the U.S. economy almost $3 billion per year. Optic neuritis (i.e., inflammatory demyelination of the ON) is the initial symptom in ~25% of all multiple sclerosis (MS) cases (which impacts over 400 thousand Americans and intro- duces societal health care costs of nearly $30 billion per year). Nearly two thirds of MS patients will experience episodes of optic neuritis in their lifetimes, and 40-60% of patients have visual defects localized to the ON. These disorders irre- versibly damage the ON. Even so, damage to axons in the ON is progressive, defined by a window of opportunity for treatment between loss of function and actual degeneration. The potential for recovery exists because there are treatments that can help prevent progression if administered during this window of opportunity. Yet, we do not have effective means to assess who is in the window and who will benefit from treatment. We propose to translate computational imaging methods from the neuroimaging community to provide ro- bust, quantitative tools for assessing the optic nerve (ON) on clinical and research imaging sequences. These efforts will improve prognostic accuracy, lead to better understanding of patient responses, and enhance targeted interven- tions. The technical hypothesis of this work is that quantitative image processing can robustly and accurately segment, register, and fuse ON data from modern MRI and CT clinical sequences. The central hypothesis of this proposal is that qualitative ON phenotypes on longitudinal clinical imaging will differentiate individuals who respond to treatment versus those who do not. The overall goal of this research is to provide a foundation for image analysis of the ON and its relationships with pathological disorders. We will build upon recent advances in robust medical image computing to segment the ON in clinical CT and MRI acquisitions, develop registration procedures to establish intra- and inter-subject correspondence, and bring together information from the multi-modal battery of imaging studies that are typically used in clinical care (aim 1). With these new methods, we will address the exploratory hypothesis that quantitative use of clinical imaging data can increase prognostic accuracy (aim 2). We note that aim 2 is particularly exploratory and in line with the high- risk/high-reward aspect of this mechanism;many studies have shown that baseline imaging does not conclusively pre- dict long term outcome or treatment response. We hypothesize that this may be because early findings are related to edema and inflammation rather than cellular damage per se. Once this exploratory phase is complete, we will pursue promising prognostic biomarkers using more detailed condition staging criteria and including more than two longitudinal time points in the analysis. Ultimately, these efforts will improve assessment ON disease and, in turn, patient care.
We propose to translate medical imaging computing procedures from the neuroimaging community to provide robust, quantitative tools for assessing the optic nerve (ON) on clinical and research imaging sequences. The technical hypothesis of this work is that quantitative image processing can robustly and accurately segment, register, and fuse ON data from modern MRI and CT clinical sequences. The central hypothesis of this proposal is that qualitative ON phenotypes on longitudinal clinical imaging will differentiate individuals who respond to treatment versus those who do not more effectively than traditional pre-interventional measures.
|Asman, Andrew J; Landman, Bennett A (2014) Hierarchical performance estimation in the statistical label fusion framework. Med Image Anal 18:1070-81|
|Panda, Swetasudha; Asman, Andrew J; Delisi, Michael P et al. (2014) Robust Optic Nerve Segmentation on Clinically Acquired CT. Proc SPIE Int Soc Opt Eng 9034:90341G|
|Asman, Andrew J; Dagley, Alexander S; Landman, Bennett A (2014) Statistical label fusion with hierarchical performance models. Proc SPIE Int Soc Opt Eng 9034:90341E|