Improving the efficiency of brain metastases monitoring using a semi-automatic colorimetric longitudinal tool.

BACKGROUND

After stereotactic radiotherapy for brain metastases, MRI follow-up is essential to detect lesion variations. Traditional diagnostic methods, including volumetric changes often face limitations in precision and timeliness.

PURPOSE: To evaluate the performance of an automatic colorimetric segmentation tool (LOBI, IntellisSpace Portal viewer, Version 11, Philips Healthcare) against volumetric analysis in detecting radiological progression (RP) and to explore their potential integration into routine practice.

MATERIAL AND

METHODS

From a retrospective cohort of 125 patients and 298 brain metastases treated at the University Hospital of XXX, 1170 comparisons between two consecutive MRIs were conducted using both the LOBI tool and a volumetric comparison based on manual segmentation performed by four radiologists. Various volumetric cut-offs and the LOBI tool were used and compared to the gold standard (RP) using diagnostic metrics such as balanced accuracy. RP was collected from follow-up data (biopsy, surgery, or multidisciplinary meeting boards) and encompassed both local relapse and radionecrosis. The average analysis times were compared using Student's t-test.

RESULTS

Against the reference diagnosis of radiological progression, LOBI achieved a balanced accuracy of 63.8% with an NPV of 94.0% (Se = 55.3%; Sp = 72.3%). LOBI was significantly faster than manual volumetric monitoring (29svs164s).

CONCLUSIONS

LOBI significantly reduced the measurement times while achieving similar performance results. We propose to combine LOBI as a screening tool and volumetry as a refining tool in routine practice.