Improved epileptic foci localization using joint SPECT reconstruction: Phantom and patient studies

Objectives To demonstrate improved performance of joint ictal/interictal brain SPECT reconstruction in epilepsy patients compared to the clinical subtraction method. This method incorporates Poisson noise modeling and registration between ictal and interictal scans to reconstruct the seizure image from both scans simultaneously.

Methods Phantom study: Noise-free 99mTc projection sets were acquired on a 3D Hoffman phantom at two different positions and orientations, and on a 3.3-cm “hot” sphere with a cold background. These datasets were combined to mimic clinical noise-free ictal (3 seizure-to-background contrast ratios: 1.25, 1.4, 1.55) and interictal scans. For each combination, we generated 25 noise realizations by adding noise to the projections. The mean and standard deviation (SD) of seizure-to-background contrast were computed using both the subtraction and joint reconstruction approaches. Patient study: 35 ictal and interictal datasets were processed using both approaches and presented to two physicians to localize a seizure region. The readers’ data were used to calculate the localization-response receiver operating characteristic (LROC) curves.

Results Phantom study: The difference of mean contrast obtained by both methods became less as iteration number increased. Compared with the subtraction approach, the SD reduction by the joint approach after 10 iterations ranged from 54.7% to 64.9% (p<0.0005). Patient study: The joint method yielded increased area under LROC curve for both readers (from 0.24 to 0.34, and 0.15 to 0.20).

Conclusions The proposed joint reconstruction method achieves better image quality, hence holding promise to provide a better alternative to the current clinical standard for epileptic foci localization.