Nature Medicine
Software detected esophageal cancer on existing chest scans
In external testing of 11,466 patients, the model detected 90.0% of cancers and 52.5% of precancerous lesions, while correctly classifying 98.5% without either.In external testing of 11,466 patients, the model detected 90.0% of cancers and 52.5% of precancerous lesions, while correctly classifying 98.5% without either.
Cohort · 11,466 peopleEsophageal cancer
Who was studied11,466 patients
What changedSoftware detected esophageal cancer on existing CT scans with 90.0% sensitivity and 98.5% specificity
The catchFollow-up in the prospective and real-world validations was less than 2 years