Mar 17, 2021 — 4 min read
“AI? I am hearing about AI everywhere I go.”
“But has it reached the level where I can use it in my daily practice? Can I really rely on AI analysis results?”
“Lunit AI for chest x-ray? Has the software performance been validated?”
“What does the clinical evidence say?”
We are going to answer your questions by reviewing the key findings of each study published in major peer-reviewed journals such as JAMA Network Open and European Respiratory Journal.
According to the study published in JAMA Network Open, using Lunit AI solution, non-radiology physicians, radiologists, and even thoracic radiologists can improve their diagnostic performance for chest abnormalities.
Lunit AI algorithm detects pulmonary nodules, pneumothorax, pneumonia, and active tuberculosis. The list goes on.
This finding suggests that Lunit AI solution can be used as a reliable second reader not only in radiology departments, but also in clinical departments.
According to the study published in Radiology, using Lunit AI solution, the number of overlooked lung cancers on chest x-ray exams decreased.
Let’s take a look at a clinical case from the study.
This is the chest x-ray of a 52-year-old woman from her initial exam. There is a small nodule overlapped by the left hilar shadow, which was overlooked during the exam.
After a year and a half, the patient took another exam and was diagnosed with stage four lung cancer.
What if Lunit AI had been used during the initial exam?
According to the study, only two out of nine radiologists could detect this lung nodule. However, with Lunit AI, all nine radiologists successfully found it. If Lunit AI had been used in the reading, the patient could have been diagnosed with early stage lung cancer and received proper treatments a year and a half earlier.
Lunit abnormality score shows how confident the AI is with its detection.
You can sort the exams on the worklist according to this score, up from the exams with the most suspicious abnormal findings, down to the exams with the lowest. You can prioritize images in this reading order.
According to the study published in the European Respiratory Journal, using the AI result-embedded PACS worklist, radiologists could reduce their overall reading time (from 23.5 to 20.5 seconds; 13% reduction).
In particular, they spent a significantly shorter time reading normal cases. (from 17.9 to 13.5 seconds; 33% reduction)
This outcome shows that Lunit AI solution can guide you in deciding which image requires your urgent attention, ultimately reducing your reading time.
These studies demonstrate that by improving the reading accuracy and efficiency, Lunit INSIGHT CXR can contribute to improving the overall patient care outcomes.
Try Lunit AI solution now at insight.lunit.io.
Once you upload DICOM images, you will get the AI results within seconds.