|
Abstract
|
SAVIR 2022 Conference Abstracts
Statement of Purpose We aim to develop a Machine learning algorithm that is able to predict injuries utilizing trauma series X-Rays through image recognition.
Methods/Approach This will be a cross sectional study conducted at Aga Khan University Hospital, Karachi, Pakistan. Radiological data in the form of X-rays will be extracted retrospectively from past medical records of Adult Road Traffic Crash Victims. We will adopt a two-phase machine learning framework, phase I being the statistical phase and phase II being the machine learning phase to develop an algorithm. The accuracy of the algorithm-based tool will be calculated and reported in the form of Sensitivity, Specificity, Model Accuracy, Precision and Area Under the Receiver Operator Curve (AUROC).
Results/Conclusion This is an abstract for a study protocol. However, the results of this study will be presented in the form of statistical measures and machine learning accuracy predictors.
Significance Road traffic crash injuries contribute to significant morbidity and mortality worldwide with Lower Middle-Income Countries (LMICs) facing the highest burden. Trauma management is a time sensitive matter as it is important to categorize and detect patient's injuries for timely intervention in the golden hour of trauma. Emergency physicians rely heavily on radiological investigations including trauma series (C-spine, Chest and Pelvic X-Rays), apart from their clinical examinations (Primary/secondary survey) in order to reach to a definitive diagnosis but delays in X-ray visualization and reporting is a major hurdle. Hence, developing a machine learning algorithm to detect injuries in trauma patients using simple X-ray trauma series will allow for quicker injury/abnormality detection, surpassing the usual delays, aiding in a faster management of trauma patients, transfer or discharge related decision making and reduce substantial morbidity and mortality, especially in resource limited settings where there is scarcity of trained personnel for radiological readings.
Language: en |