Abstract
Introduction
Numerous publications concerning the efficacy of Gamma Knife radiosurgery (GKRS) on vestibular schwannoma (VS) report control rates of approximately 90%. However, it is largely unknown why treatment failed in the other 10% of cases. We started a large project 4 years ago, investigating the factors that may influence the efficacy of GKRS on VS. The objective is to develop a model that can a priori predict the GKRS outcome on an individual basis.
Material and methods
A dataset was created containing all VS patients treated with GKRS, excluding NF2 and previously treated patients, with a follow-up of at least 3 years. We included (1) demographic factors, (2) technical treatment-related factors (dose to 90-100% of the tumor volume, peak dose, number of iso-centers, beam-on time, conformity-, selectivity-, and gradient index), and (3) tumor-related factors (tumor volume, pre-treatment growth rate, MRI characteristics related to tumor texture). Correlation was determined between these factors and the volumetric treatment response using statistical and machine-learning methods.
Results
The dataset contains 795 patients treated between 2002 and 2015, with a median follow-up of 62 months. The 10-year tumor control rate was 88.8%. We did not find correlation between tumor control rate and demographic- and treatment-related factors. These results indicate that differences in tumor biology may be responsible for observed differences in treatment outcome. These differences will most likely be expressed in different growth rates and variations in MRI tumor appearance, i.e. tumor texture. Indeed, pre-treatment growth rates indicate a statistically significant effect: treating slow-growing tumors is more successful than their fast-growing counterparts.
Tumor texture was quantified by calculating mean MRI intensity and features based on gray-level co-occurrence matrices. Using these features, we developed a model to predict volumetric changes such as transient tumor enlargement (TTE) and long-term tumor control. For TTE we obtained a prediction accuracy of 80% and pilot studies evaluating long-term tumor control show promising results, obtaining 85% accuracy.
Conclusion
Results from this large project indicate that variations in tumor biology are responsible for the differences observed in the GKRS outcome on VS. Quantifying MRI tumor texture can lead to powerful tools for predicting treatment outcome on an individual basis.
Numerous publications concerning the efficacy of Gamma Knife radiosurgery (GKRS) on vestibular schwannoma (VS) report control rates of approximately 90%. However, it is largely unknown why treatment failed in the other 10% of cases. We started a large project 4 years ago, investigating the factors that may influence the efficacy of GKRS on VS. The objective is to develop a model that can a priori predict the GKRS outcome on an individual basis.
Material and methods
A dataset was created containing all VS patients treated with GKRS, excluding NF2 and previously treated patients, with a follow-up of at least 3 years. We included (1) demographic factors, (2) technical treatment-related factors (dose to 90-100% of the tumor volume, peak dose, number of iso-centers, beam-on time, conformity-, selectivity-, and gradient index), and (3) tumor-related factors (tumor volume, pre-treatment growth rate, MRI characteristics related to tumor texture). Correlation was determined between these factors and the volumetric treatment response using statistical and machine-learning methods.
Results
The dataset contains 795 patients treated between 2002 and 2015, with a median follow-up of 62 months. The 10-year tumor control rate was 88.8%. We did not find correlation between tumor control rate and demographic- and treatment-related factors. These results indicate that differences in tumor biology may be responsible for observed differences in treatment outcome. These differences will most likely be expressed in different growth rates and variations in MRI tumor appearance, i.e. tumor texture. Indeed, pre-treatment growth rates indicate a statistically significant effect: treating slow-growing tumors is more successful than their fast-growing counterparts.
Tumor texture was quantified by calculating mean MRI intensity and features based on gray-level co-occurrence matrices. Using these features, we developed a model to predict volumetric changes such as transient tumor enlargement (TTE) and long-term tumor control. For TTE we obtained a prediction accuracy of 80% and pilot studies evaluating long-term tumor control show promising results, obtaining 85% accuracy.
Conclusion
Results from this large project indicate that variations in tumor biology are responsible for the differences observed in the GKRS outcome on VS. Quantifying MRI tumor texture can lead to powerful tools for predicting treatment outcome on an individual basis.
| Original language | English |
|---|---|
| Publication status | Published - Jun 2019 |
| Event | 8th Quadrennial Conference on Vestibular Schwannoma and Other CPA Tumors: Advancing Care through Ideas and Innovation 2019 - Mayo Civic Center, Rochester, United States Duration: 18 Jun 2019 → 21 Jun 2019 https://vs2019.mayo.edu/ |
Conference
| Conference | 8th Quadrennial Conference on Vestibular Schwannoma and Other CPA Tumors |
|---|---|
| Country/Territory | United States |
| City | Rochester |
| Period | 18/06/19 → 21/06/19 |
| Internet address |