Better Decision Making
2014 R&D 100 Winner
Humans often perform decision tasks that require visual processing and high-level cognitive analysis of complex visual content. For example, clinicians routinely make medical diagnoses based on their interpretation of images. Great efforts are dedicated to developing clinical decision support systems to help reduce medical errors. However, clinicians make unique, individual errors whereas existing systems offer “one-size-fits-all” supports, ignoring the individual perceptual and cognitive behaviors of each clinician. Oak Ridge National Laboratory has developed iSPM: Intelligent Software Suite for Personalized Modeling of Expert Opinions, Decisions and Errors in Visual Examination Tasks, a novel technology utilizing eye-tracking hardware, an intelligent GUI engine and advanced analytics to predict an individual’s perceptual behavior, cognitive response and risk of error for complex decision tasks such as cancer diagnosis from medical images.
The product leverages advanced machine learning to link computationally visual content, human perception, cognition and decision error. Powered by the three core components, iSPM monitors a user’s case-specific gaze metrics and cognitive behaviors and merges the observation signals with computer-extracted information from the specific image regions to predict human decision-making errors.
The expected benefits of this technology are better outcomes and lower treatment costs due to earlier intervention, and lower malpractice costs. The innovative algorithms support translation of the proposed technology in all domains where personalized modeling of visual and cognitive behavior could improve human performance.
Oak Ridge National Laboratory
|Oak Ridge National Laboratory's iSPM: Intelligent Software Suite for Personalized Modeling of Expert Opinions, Decisions and Errors in Visual Examination Tasks development team (l-r): Songhua Xu, Sophie Voisin, Georgia Tourassi and Hong-Jun Yoon.|
The iSPM: Intelligent Software Suite for Personalized Modeling of Expert Opinions, Decisions and Errors in Visual Examination Tasks Development Team from Oak Ridge National Laboratory
Georgia Tourassi, Principal Developer