How Do Sensitivity and Specificity of Imaging Modalities Affect AVM Radiosurgery Planning and Evaluation
P. Zhang,L. Wu,S. Isaacson,T. Liu,G. Kutcher
DOI: https://doi.org/10.1016/j.ijrobp.2007.07.1255
2007-01-01
Abstract:Purpose/Objective(s)Arteriovenous malformation (AVM) is clinically identified using both Magnetic Resonance Angiography (MRA) and orthogonal X-ray angiographies (OXA). However there are discrepancies between the two image findings due to the performance uncertainties of the two imaging modalities. Sensitivity and specificity are used to quantify the performance characteristics of an imaging device. The purpose of this project is to develop a framework that integrates performance characteristics of MRA and OXA into AVM radiosurgery planning and evaluation.Materials/MethodsTen patients with AVM located in critical areas such as occipital or left temple lobe were recruited in this retrospective planning study. The image findings of the 10 patients were first utilized to estimate the sensitivity and specificity of MRA and OXA. Instead of accepting the observation from imaging as a binary (yes or no) mapping of AVM location, our alternative is to translate the image into an AVM probability distribution map by incorporating the sensitivity and specificity of the imaging devices, and use such a map as a basis for treatment planning and evaluation. Three sets of radiosurgery plans, targeting at the MRA and OXA positive overlap, at MRA positive only, and at OXA positive only were optimized for best conformality via the GammaPlan® system. AVM obliteration rate (ORAVM) was calculated using the dose volume histogram and AVM probability distribution. Late toxicity to brain tissue was evaluated using the volume receiving more than 12 Gy (V12). Both ORAVM and V12 served as end points for plan comparison.ResultsIn our 10 patient study, the positive overlap of MRA and OXA accounted for 90.5 ± 8.1% of the MRA positive only volume, and 61.9 ± 7.8% of the OXA positive only volume. The sensitivities and specificities of MRA and OXA were estimated to be (0.95, 0.74) and (0.71, 0.95), respectively. The probability of being true AVM was estimated to be 0.994 ± 0.005 in the MRA and OXA positive overlap zone, 0.560 ± 0.261 in the MRA positive and OXA negative zone, 0.642 ± 0.125 in the MRA negative and XA positive zone, and 0.083 ± 0.071 in the MRA and OXA negative overlap zone, respectively. Compared to plans targeting at the MRA and OXA positive overlap, plans targeting at the MRA positive and at the OXA positive improved ORAVM by 4.1 ± 1.9% and 15.7 ± 8.3%, respectively, while also increasing V12 1.88 ± 1.36 cc and 6.91 ± 2.14 cc, respectively. Attempts of shaping the dose distribution according to the AVM probability map suggested that dose reductions of 2–3 Gy in the zone where MRA and OXA disagree could be applied to spare the surrounding critical brain tissue.ConclusionsThe sensitivity and specificity of MRA and OXA should be incorporated into AVM radiosurgery planning to better assist clinicians' decision making with regard to target definition and plan evaluation. Purpose/Objective(s)Arteriovenous malformation (AVM) is clinically identified using both Magnetic Resonance Angiography (MRA) and orthogonal X-ray angiographies (OXA). However there are discrepancies between the two image findings due to the performance uncertainties of the two imaging modalities. Sensitivity and specificity are used to quantify the performance characteristics of an imaging device. The purpose of this project is to develop a framework that integrates performance characteristics of MRA and OXA into AVM radiosurgery planning and evaluation. Arteriovenous malformation (AVM) is clinically identified using both Magnetic Resonance Angiography (MRA) and orthogonal X-ray angiographies (OXA). However there are discrepancies between the two image findings due to the performance uncertainties of the two imaging modalities. Sensitivity and specificity are used to quantify the performance characteristics of an imaging device. The purpose of this project is to develop a framework that integrates performance characteristics of MRA and OXA into AVM radiosurgery planning and evaluation. Materials/MethodsTen patients with AVM located in critical areas such as occipital or left temple lobe were recruited in this retrospective planning study. The image findings of the 10 patients were first utilized to estimate the sensitivity and specificity of MRA and OXA. Instead of accepting the observation from imaging as a binary (yes or no) mapping of AVM location, our alternative is to translate the image into an AVM probability distribution map by incorporating the sensitivity and specificity of the imaging devices, and use such a map as a basis for treatment planning and evaluation. Three sets of radiosurgery plans, targeting at the MRA and OXA positive overlap, at MRA positive only, and at OXA positive only were optimized for best conformality via the GammaPlan® system. AVM obliteration rate (ORAVM) was calculated using the dose volume histogram and AVM probability distribution. Late toxicity to brain tissue was evaluated using the volume receiving more than 12 Gy (V12). Both ORAVM and V12 served as end points for plan comparison. Ten patients with AVM located in critical areas such as occipital or left temple lobe were recruited in this retrospective planning study. The image findings of the 10 patients were first utilized to estimate the sensitivity and specificity of MRA and OXA. Instead of accepting the observation from imaging as a binary (yes or no) mapping of AVM location, our alternative is to translate the image into an AVM probability distribution map by incorporating the sensitivity and specificity of the imaging devices, and use such a map as a basis for treatment planning and evaluation. Three sets of radiosurgery plans, targeting at the MRA and OXA positive overlap, at MRA positive only, and at OXA positive only were optimized for best conformality via the GammaPlan® system. AVM obliteration rate (ORAVM) was calculated using the dose volume histogram and AVM probability distribution. Late toxicity to brain tissue was evaluated using the volume receiving more than 12 Gy (V12). Both ORAVM and V12 served as end points for plan comparison. ResultsIn our 10 patient study, the positive overlap of MRA and OXA accounted for 90.5 ± 8.1% of the MRA positive only volume, and 61.9 ± 7.8% of the OXA positive only volume. The sensitivities and specificities of MRA and OXA were estimated to be (0.95, 0.74) and (0.71, 0.95), respectively. The probability of being true AVM was estimated to be 0.994 ± 0.005 in the MRA and OXA positive overlap zone, 0.560 ± 0.261 in the MRA positive and OXA negative zone, 0.642 ± 0.125 in the MRA negative and XA positive zone, and 0.083 ± 0.071 in the MRA and OXA negative overlap zone, respectively. Compared to plans targeting at the MRA and OXA positive overlap, plans targeting at the MRA positive and at the OXA positive improved ORAVM by 4.1 ± 1.9% and 15.7 ± 8.3%, respectively, while also increasing V12 1.88 ± 1.36 cc and 6.91 ± 2.14 cc, respectively. Attempts of shaping the dose distribution according to the AVM probability map suggested that dose reductions of 2–3 Gy in the zone where MRA and OXA disagree could be applied to spare the surrounding critical brain tissue. In our 10 patient study, the positive overlap of MRA and OXA accounted for 90.5 ± 8.1% of the MRA positive only volume, and 61.9 ± 7.8% of the OXA positive only volume. The sensitivities and specificities of MRA and OXA were estimated to be (0.95, 0.74) and (0.71, 0.95), respectively. The probability of being true AVM was estimated to be 0.994 ± 0.005 in the MRA and OXA positive overlap zone, 0.560 ± 0.261 in the MRA positive and OXA negative zone, 0.642 ± 0.125 in the MRA negative and XA positive zone, and 0.083 ± 0.071 in the MRA and OXA negative overlap zone, respectively. Compared to plans targeting at the MRA and OXA positive overlap, plans targeting at the MRA positive and at the OXA positive improved ORAVM by 4.1 ± 1.9% and 15.7 ± 8.3%, respectively, while also increasing V12 1.88 ± 1.36 cc and 6.91 ± 2.14 cc, respectively. Attempts of shaping the dose distribution according to the AVM probability map suggested that dose reductions of 2–3 Gy in the zone where MRA and OXA disagree could be applied to spare the surrounding critical brain tissue. ConclusionsThe sensitivity and specificity of MRA and OXA should be incorporated into AVM radiosurgery planning to better assist clinicians' decision making with regard to target definition and plan evaluation. The sensitivity and specificity of MRA and OXA should be incorporated into AVM radiosurgery planning to better assist clinicians' decision making with regard to target definition and plan evaluation.