Fifty-seventh annual meeting of the American association of physicists in medicine
TH-CD-207-10: Effect of Noise On the Optimal B-Value Pairs for Obtaining Perfusion-Insensitive Apparent Diffusion Coefficient in Diffusion-Weighted MRI
Monoexponential model using typically two or three b-values is used in clinical diffusion-weighted MRI (DW-MRI) to compute the apparent diffusion coefficient (ADC) due to computational simplicity and reduced post-processing times compared to the intravoxel incoherent motion (IVIM) model. The purpose of this study is to find the effect of noise level on the optimal b-values to obtain monoexponential ADC values close to the perfusion-insensitive IVIM model ADC (ADCIVIM).
Respiratory triggering was applied to the echo-planar DW-MRI in 1.5T scanner with TR=∼4500 ms, TE=74 ms, eight b-values of 0–1000 µs/µm2, pixel size=1.98×1.98 mm2, slice thickness/gap=6 mm/1.2 mm and 7 axial slices. For one pre-treatment DW-MRI scan of each of ten subjects, DW-MRI signal intensity images with b-values of 0–2000 µs/µm2 were synthesized on a voxel-by-voxel basis from the estimated IVIM parameters. The images were corrupted by Rician noise distribution with σ=0, 5, 10 and 15. Monoexponential ADC value was computed using b-value pairs with low b-value (blow) of 0–500 µs/µm2 and high b-value (bhigh) of 650–2000 µs/µm2. The square root of the mean of squared error percentage (RMSE) of the ADC value relative to the corresponding reference-standard ADCIVIM for the tumor volumes of the scans was computed.
There were wide bands of b-value sets giving small RMSE of ADC value relative to the ground-truth ADCIVIM value. The band of sets with small errors moved from high values of both blow and b high for the lowest noise level towards the lower values with the increase of the noise levels. The minimum error corresponding to the optimal sets increased with the noise level. The range of RMSE for σ=5 was approximately 5–30% whereas that for σ=15 was approximately 10–50%.
The optimal b-value pairs for obtaining ADC values close to reference-standard values depended on the noise level and hence SNR.
This work was supported by Virginia Commonwealth University Massey Cancer Center pilot project A35242 (EW). There are no conflicts of interest.