Taipei Medical University Hospital and Taiwan AI Labs screening for COVID-19 with AI Chest X-ray
Published on 2021-01-04
The screening system applies AI to read chest X-rays uploaded by the hospital, and instantly provides clinicians with reliable values on the location of pneumonia and lung infection.
TMU Hospital Superintendent Dr. Ray-Jade Chen (陳瑞杰) comments that putting suspected cases who were admitted into the hospital through the AI real-time screening helps reduce the number of screening tests and medical costs and further accelerate the hospital’s existing treatment procedure.
Dr. Po-Li Wei (魏柏立), Vice Superintendent of TMU Hospital described that when receiving a suspected case, the hospital will immediately conduct a chest X-ray examination and RT-PCR (Real-Time Polymerase Chain Reaction) screening. After the diagnosis is confirmed, it must be reported to the CDC (Taiwan Centers for Disease Control). The whole procedure requires 2-3 days before a decision is made to admit the patient into a negative pressure isolation ward. Moreover, supposing PCR screening is conducted for 1,000 suspected cases at about NT$3,000 out-of-pocket expense per individual, the total cost would reach NT$3 million. Coupled with the waiting time for the test results, prompt treatment would be compromised.
Therefore, TMU Hospital discussed with Taiwan AI Labs the optimization of current medical procedures. Through an AI machine training model that differs from the past, the specificity of the AI model can be emphasized and True Negative value cases can be filtered out from the large number of suspected patients identified by the thoracic X-ray screening system.
In the collaboration, TMU Hospital provides medical knowledge, ideas and suggestions while the Labs does the technical guidance and increases the difficulty of training models to improve system accuracy and stability. In the future, when a patient completes a chest X-ray, the image will be uploaded to the AI chest X-ray screening system where the reliability value of the patient’s pneumonia characteristics can be detected within a short time. For cases with high reliability values on the AI system, physicians can conduct further testing and treatment, thereby reducing the risk of delayed confirmation and improving the quality of medical treatment.
In addition, jointly developed by both parties in 2019, the AI Sepsis Early Warning System has been used in intensive care units. Through AI big data learning and medical guidance from TMU physicians, specific parameters are identified from tens of thousands of data to predict current sepsis risk of patients and their risk within the next 48 hours.
Based on the prediction, doctors can focus on treating high-risk patients immediately to avoid missing the critical treatment time. Currently, TMU Hospital and Taiwan AI Labs are continuing to work closely together, and hope that the results of AI development can assist doctors in their clinical practice.