A report-grounded vision-language foundation model for colonoscopy from 280000 routine reports | Digital Library | PAMCET | PAMCET