NYU Langone Health recently developed a new AI tool which can read chest X-rays, as well as foresee COVID-19 cases with severity. The accuracy of this new tool is as high as 80%, as reported by experts on May 12, 2021.
The new model is built using over 5,200 radiographs, sourced from approximately 3,000 COVID patients who are seriously ill. Physicians and radiologists are facing an urgent need to find a way ahead of COVID in patients who need additional care.
NYU’s COVID-19 classification test is among the largest AI applications in radiology for addressing emergency needs of caregivers and critical patients during these times. The new artificial intelligence tool also makes use of demographics apart from radiological images. These vitals include test results from lab, age, race, vital signs and gender of a patient.
Certain factors which played a key role in AI tool training included a patient’s ventilator requirement, along with the information whether they could survive or die from the disease.
Yiqiu Shen, NYU’s Center for Data Science’s doctoral student, along with colleagues logged note-worthy results from testing this AI tool on chest X-rays from over 700 patients. The program’s efficacy and promising results are analysed from the fact that it rightfully speculated that 4 out of 5 COVID patients would be needing ICU, ventilator or might end up dying within four days of hospitalization.
Yiqiu Shen’s team is hoping to also include clinical data so as to support and strengthen this model, while testing it in real-time with emergency physicians and radiologists. The team is presently working at drafting of clinical guidelines.