Radiographic surgical indicators for pediatric orbital abscess Abstract As the most common sequela of acute sinusitis, pediatric orbital cellulitis is the most common cause of inpatient ophthalmic admissions. Unmitigated infection can lead to devastating consequences, such as blindness and death, requiring prompt diagnosis and treatment. Treatment options include the initiation of intravenous antibiotics and the drainage of discrete orbital abscess formation. Unfortunately, the exact role and timing of drainage remain controversial. There is an acute need for a reliable method to rapidly and unambiguously identify pediatric patients who require immediate surgical intervention or just conservative management with medical therapy. Prioritizing patients for surgical intervention will prevent further orbital or intracranial complications of orbital cellulitis and decreased potential morbidities and mortality associated with late interventions. While several radiographic indicators have been associated with surgery, none has been proven satisfactory in clinical practice. In contrast, the size of the abscess (e.g., volume) demonstrates a more robust performance among the identified factors. However, there is no tool available that can consistently and accurately measure the volume of an abscess. Manual delineating an abscess is always time-consuming and associated with high inter- and intra-variability, resulting in different cutoff values and thus widely diverse recommendations for treatment guidelines. In this project, we propose to (1) develop a novel computer tool for automatically, accurately, and consistently quantifying orbital abscess depicted on orbital computed tomography (CT) by leveraging the emerging artificial intelligence (AI) technology; (2) use this tool to comprehensively analyze the radiographical characteristics of the orbital abscess and identify those that are closely associated with the outcome of the therapy; and (3) integrate the radiographical findings with patient demographics as a computer model for predicting the response to medical treatment, by which we expect to identify patients who would benefit from early surgical intervention. We believe the availability of such a tool will be a novel and essential addition to the practice of oculoplastic and orbital surgery. Its availability will facilitate precise diagnosis and decision making, ameliorate a significant portion of these complications associated with prolonged medical treatment, and ultimately decrease the health care burden of pediatric orbital cellulitis.