Classically, anatomic information provided by MRI has been essential to the neurosurgeon to maximize extent of tumor resection and, as a result, improve survival statistics. That said, it is not routine during resections to make use of similar imaging information that reflect the functional organization of the brain. Task-based fMRI has been employed as a means of pre-operatively localizing function. However, task-based fMRI depends on the patient?s ability to comply with the task paradigm, which frequently is lacking. This problem can be overcome by using the recently developed method of resting state functional magnetic resonance imaging (rsfMRI) to localize function. Moreover, rsfMRI is highly efficient, as multiple resting state networks (RSNs) associated with multiple cognitive domains can be mapped at the same time. With this in mind, the long-term goal of our research is to improve survival and quality of life after surgical resection of brain tumors by improving the identification/preservation of eloquent cortex. The current barrier that prevents the widespread use of rsfMRI is 1) the high degree of advanced imaging expertise currently necessary to create and interpret the images and 2) the necessary IT infrastructure necessary to support the analyses. To address this shortcoming, we propose to create a turnkey system for functional mapping within the brain that resides on a cloud-computing platform. At the heart of our methodology is a multi-layer perceptron (MLP) algorithm that assigns RSN membership to each locus within the brain using supervised classification of rsfMRI data. Current data demonstrate that MLP-based RSN mapping is more reliable than conventional task-based fMRI and is extremely sensitive to sites identified by cortical stimulation (the standard in intraoperative mapping). Translation of the science and techniques created at Washington University will be accomplished by a deep collaboration with Radiologics, an emerging company with strong expertise in cloud computing for clinical imaging. Towards this end, the overall objective of the proposed project is to create an imaging technology package, named Cirrus, that will integrate automatic MLP-based RSN mapping with cloud-based computing.
The Specific Aims of this proposal are to 1) Create Cloud-Based rsfMRI Brain Mapping Capability - Cirrus, 2) Deploy Cloud-Based rsfMRI Capability (Cirrus Platform) to New User Host Institutions, and 3) Optimize the Clinical User Interface (UI) of Cirrus. The expected outcome of this translational strategy will be an integrated imaging/surgical presurgical mapping technology using rsfMRI with clearly defined performance capabilities, well delineated localization outputs, and a clinician friendly user experience that scales to all types of health care settings. Thus, this proposal is innovative because there currently does not exist any comparable system that integrates cutting edge image analysis tools with cloud-based capabilities. This work is significant because it will disseminate technology that fundamentally enhances the surgeon?s understanding of the functional implications of their surgical strategy, thereby enables safer, more tailored approaches to improving outcomes.
The proposed research is an academic industry partnership that will make available, to any neurosurgical practice advanced functional MRI methodology to define critical regions in the brain. This development is relevant to public health because it has the potential to improve functional outcomes following surgical removal of brain tumors. Towards this end, we propose to create an imaging technology package that will integrate 1) advanced machine learning techniques using resting state functional magnetic resonance imaging, and 2) cloud-based computing, to perform advanced brain mapping. Thus, advanced functional localization can be standardized and acquired at any institution by being uniformly processed in the cloud and sent back to the home institution without the need for costly infrastructure. The integrated package will enable clinicians in any practice environment to efficiently and reliably visualize function on the brain?s anatomy prior to surgery.
