EFAI's chest x-ray model is an artificial intelligence algorithm that was designed to screen for major findings including: support device, foreign body, fracture, edema, consolidation, pneumonia, emphysema, fibrosis, mass, nodule, pneumothorax, effusion, cardiomegaly, calcification. The software provides fast screening to help reduce missed abnormality detection's and improve clinical workflow.
Clinically, chest x-ray is one of the most commonly used diagnostic test in medicine. Many findings or diseases can be detected through a first glance on chest radiographs. EFAI's chest x-ray model supports physicians in detecting major findings and abnormalities to reduce subjectivity of chest x-ray assessment and increases workflow efficiency.
EFAI's chest x-ray model is a fully-automated AI algorithm that detects abnormalities for posterior-anterior (PA) or anterior-posterior (AP) view chest x-rays. The software is intended to aid medical professionals to detect for support device, foreign body, fracture, edema, consolidation, pneumonia, emphysema, fibrosis, mass, nodule, pneumothorax, effusion, cardiomegaly, calcification. A positive finding other than "no findings" could help notify physicians to double check the location of where the bounding boxes are located. The system is to be used by trained radiologists.
The model works for ICU, ED, and outpatients. Its intended to be used for adults over 20 years old.
Model performance metrics
Macro average AUROC = 0.893
The figure on the left shows an example of the model inference results. The figure on the right shows the performance on each findings.
Ever Fortune.AI was born in 2018 composed of physicians, data scientists, and software engineers. The collaboration between clinical experts and engineers has driven the team to deliver several innovative technologies to the medical community. Ever Fortune has over 12 partnering medical institutions in Taiwan to support the development and commercialization of end-to-end medical AI algorithms with both on-premise and cloud solutions.