Background: Curie scores have shown prognostic significance in high risk neuroblastoma. However, the potential for significant inter and intra-observer variability exists when determining (and recording) individual patient scores. We have created a deployable and scalable system for recording and tracking MIBG Curie scores that is comparable to current practice, while allowing lesion tracking.
Methods: We designed three computer-assisted methods for recording Curie scores. Method A recapitulates current practice, allowing the radiologist to record a score of 0-3 for each anatomic region. Method B facilitates the marking of lesions and the assignment of a region from a schematic. Method C permits a user-driven segmentation of the image, computer-assisted marking of lesions, and automatic determination of score. For testing, 38 MIBG scans representing a range of ages and extent of disease were selected. Five experienced radiologists from four academic medical centers scored the 38 scans using each method and were surveyed for feedback.
Results: There was no significant difference in total scores among the three methods (p=0.91) or among the observers (p=0.14). Methods B and C took an average of 58 and 111 seconds longer than Method A, respectively, due to marking each lesion. Overall, radiologists felt that tracking lesions would be very useful for oncologists. Of the three methods, marking the lesions and selecting the region (Method B) was preferred.
Conclusions: A computer-assisted approach to determining Curie scores for MIBG scans successfully provides clinicians with a mechanism to track neuroblastoma lesions over time without loss of score reliability or consistency. The software is built on an easily-deployable platform for use at any medical center as both a research tool and clinical instrument. Given the growing necessity to Curie score MIBG scans, longitudinal lesion tracking and score recording could have an important impact on the care and treatment of children with neuroblastoma.