Related Experiment Video
Updated: May 22, 2026
Comparison of Predictive Performance of Three Lymph Node Staging Systems in Colorectal Signet Ring Cell Carcinoma Based on Machine Learning Model
Published on: April 18, 2025
MSCT follow-up in malignant lymphoma: comparison of manual linear measurements with semi-automated lymph node
1Department of Clinical Radiology, University of Münster. weslingj@uni-muenster.de
Purpose:
Assignment of semi-automated lymph node analysis compared to manual measurements for therapy response classification of malignant lymphoma in MSCT.
Materials And Methods:
MSCT scans of 63 malignant lymphoma patients before and after 2 cycles of chemotherapy (307 target lymph nodes) were evaluated. The long axis diameter (LAD), short axis diameter (SAD) and bi-dimensional WHO were determined manually and semi-automatically. The time for manual and semi-automatic segmentation was evaluated. The ref. standard response was defined as the mean relative change across all manual and semi-automatic measurements (mean manual/semi-automatic LAD, SAD, semi-automatic volume). Statistical analysis encompassed t-test and McNemar's test for clustered data.
Results:
Response classification per lymph node revealed semi-automated volumetry and bi-dimensional WHO to be significantly more accurate than manual linear metric measurements. Response classification per patient based on RECIST revealed more patients to be correctly classified by semi-automatic measurements, e. g. 96.0 %/92.9 % (WHO bi-dimensional/volume) compared to 85.7/84.1 % for manual LAD and SAD, respectively (mean reduction in misclassified patients of 9.95 %). Considering the use of correction tools, the time expenditure for lymph node segmentation (29.7 ± 17.4 sec) was the same as with the manual approach (29.1 ± 14.5 sec).
Conclusion:
Semi-automatically derived "lymph node volume" and "bi-dimensional WHO" significantly reduce the number of misclassified patients in the CT follow-up of malignant lymphoma by at least 10 %. However, lymph node volumetry does not outperform bi-dimensional WHO.