- In patients with advanced gastric cancer (GC), a nomogram incorporating computed tomography (CT) phenotypes of the primary tumor and the nearby peritoneum combined with Lauren type predicted the presence of occult peritoneal metastasis (PM).
- The technique uses radiomics, which creates an image-based biomarker from medical imaging.
Why this matters
- Laparoscopies are recommended for detecting occult PM, but the procedure is invasive.
- 554 patients were initially diagnosed as PM-negative but later underwent laparoscopy to confirm PM-negative (n=432) or convert to PM-positive (n=122) were used to create the model (n=100), which was tested against an internal cohort (n=226) and 2 external cohorts (n=131; n=97).
- Funding: Study funded by various nonindustry sources in China.
- Radiomic signatures from the primary tumor (RS1), the peritoneum region (RS2), and the Lauren type predicted occult PM (P<.05 for all>
- A nomogram including RS1, RS2, and the Lauren type had an area under the curve of 0.958 for the training group (95% CI, 0.923-0.993), 0.941 for the internal group (95% CI, 0.904-0.977), 0.928 for the first external validation cohort (95% CI, 0.886-0.971), and 0.920 for the second external validation cohorts (95% CI, 0.862-0.978).
- Determination based on 2-dimensional slices.