Joint ictal/interictal SPECT reconstruction for improved epileptic foci localization

Objectives Develop a novel joint ictal/interictal SPECT reconstruction method and demonstrate improved performance in epileptic foci localization as compared to the conventional subtraction method (subtracting reconstructed ictal and interictal images).

Methods In the proposed approach, a seizure image was reconstructed jointly from both ictal and interictal data with Poisson noise modeling. Phantom study: Noise-free 99mTc SPECT projections were acquired on a Hoffman head phantom and on a 3.3-cm spherical “seizure” with identical attenuation background. Ictal/interictal dataset was synthesized from the head phantom projections scaled to match those of a typical clinical study with (ictal) and without (interictal) the addition of the sphere projections (lesion:background contrast=1.5). Nine noise realizations were generated by adding Poisson noise to the projections. Mean and standard deviation (SD) of seizure’s ROI average from the reconstructed seizure image were computed. Patient study: An epilepsy patient dataset was processed as a preliminary test of our approach. The voxel-wise mean/SD ratio in seizure’s ROI was computed.

Results In the phantom study, the difference between means from the proposed joint reconstruction and conventional methods decreased as iteration number increased (9% at 50th iteration). Our proposed approach significantly reduced SD by 52% at 50th iteration compared to the conventional method (p<0.03). Likewise, the mean/SD was substantially improved in the patient study from 0.8 (conventional) to 1.9 (proposed) at 30th iteration.

Conclusions The proposed joint ictal/interictal reconstruction approach yielded improved image quality for the seizure image by circumventing the need to subtract ictal and interictal images (conventional approach), hence yielding better noise control and statistics. It offers a promising alternative to the current SPECT imaging standard for epileptic foci localization.