Reproducibility of measurements with a semi-automatic software package for the evaluation of rectal cancer

Background: Staging of rectal cancer with MRI has major impact on treatment choice and may be of importance in new cancer management strategies such as “wait-and-see” policy. Purpose: To assess the reproducibility of a software package recently developed at our department to measure volumes, apparent diffusion coefficient, and the skewness of apparent diffusion coefficient in lymph nodes and tumors in rectal cancer patients before and after chemoradiation treatment. Material and Methods: This study included 20 consecutive patients with biopsy-verified rectal cancer, in whom MRI staging had been performed both before and after chemoradiation treatment. The diffusion-weighted images were transferred to the software. The volume, apparent diffusion coefficient, and skewness were determined for 93 lymph nodes and 40 tumors. The volumes were compared with manual measurements of the volume of the same lymph nodes and tumors. Results: The agreement in semi-automatic measurements of lymph nodes was very good (ICC = 0.99), and in tumors good (ICC = 0.88). The agreement in manual measurements of lymph nodes was very good (ICC = 0.95) when all lymph nodes were included, but low (ICC = 0.52) if three outliers were excluded. Bland–Altman plots showed clear agreement between manual and semi-automatic measurements in the lymph nodes, but not in measurements of tumors. The values of apparent diffusion coefficient and skewness in tumors differed before and after treatment but did not differ in lymph nodes as a group. Conclusion: The software package showed a high degree of reproducibility in measurements on lymph nodes but requires further development to improve the reproducibility of tumor measurements.