Consultation with appropriate specialists improves the quality of healthcare, particularly in patients with complicated cases or chronic illnesses. And for the majority of such patients, specialists use imaging studies (e.g., MR, CT) to objectively document the disease process (e.g., a cancer patient on chemotherapy). However, specialists are generally not available in all communities, tending to be concentrated in academic/specialty centers. Thus, to facilitate the routine use of teleconsultations for patients when specialists are not locally present: 1) the images captured to document the patient's condition must be incorporated into the medical record to enable proper review; and 2) the remote consultant should only receive pertinent parts of the medical record to streamline the consultation process. This proposal is focused on developing and testing a """"""""context-sensitive"""""""" telehealth infrastructure based on: 1) automated incorporation of clinical context (patient presentation and referring physician hypothesis) to focus the consultation process; 2) a knowledge-base derived from data mining of natural language processing (NLP) results, mapping patient presentation to select an appropriate imaging study based on anatomical region and imaging parameters; and 3) automated selection of key anatomical structures in the acquired imaging study through the use of a contrast-customizable atlas and rigid body/deformable registration algorithms. Collectively, these technologies will allow context-sensitive, automated summarization of medical records for telehealth in a real-world environment. The proposed technologies will be implemented for neurological and musculoskeletal domains, two areas that are MR imaging intensive. Technical evaluation will be performed with experts serving as the reference standard and will focus on measuring: 1) the accuracy of the corpus based, NLP-guided knowledge-base in selecting relevant anatomical structures; and 2) the accuracy of anatomical structure delineation using the customizable atlas registration methods. Clinical evaluation will be conducted in a real-world teleconsultation environment in a before/after study design using two performance metrics: 1) the time required for consultations; and 2) the effect on the quality of the consultations.
