Individualized neuromodulation improves visual perception in a cortically blind patient: a treatment that can be applied to optic glioma patients
Introduction
The Papageorgiou lab has developed an individualized closed-loop fMRI-neuromodulation (iNM) with the goal to rehabilitate visual perception in cortically blind (CB) patients with lesions downstream of the optic chiasm. These lesions result in visual field (VF) deficits called hemianopias, which negatively impact patients’ quality of life and productivity. The etiological factors of cortical blindness are multifactorial; the most common one is stroke, but also optic nerve glioma tumors. Our goal is to develop iNM that can rehabilitate CB following optic glioma resection. We present a CB patient, with superior and inferior right homonymous hemianopia after a left posterior cerebral infarct.
Methods
Our patient (R.S.) had discriminate 100% or 33% motion coherence, interleaved with periods of random motion, during the behavioral (via a computer, outside MRI) and iNM scans. The right upper VF was trained behaviorally for three years. Training did not generalize to the inferior VF quadrant (Q). R.S. then, underwent iNM in the lower right VFQ for three sessions, during which he was asked to discriminate up and down direction at 100% and 33% coherence. Linear support vector machine (SVM) decoded direction discrimination. We quantified the area under the curve (AUC) for each area’s oxygenated hemoglobin (HbO2) magnitude and variance across directions and coherences, as a function of time. Causal modeling deciphered the brain states associated with direction and coherence selectivity, and the crosstalk and magnitude within and across networks.
Results
Behavioral training in the upper RVF Q reached normal levels of direction discrimination at 14-18K trials, which was maintained after a 9- and 20-month interval of no-training. iNM in the lower RVF Q reached normal levels of direction discrimination at 3-3.5K trials, also maintained at 20 months. SVM classification of motion direction versus random motion generated greater performance under iNM at 78.9% (100% coherence) and 69.8% (33% coherence) compared to 57.7% (100%) and 58.7% (33%) under control-no iNM. iNM enhanced the HbO2 associated with motion perception: 1) 125% and 187% increase in the magnitude of the HbO2’s AUC for up and down at 100%, and 1236% increase for up motion at 33% (p<0.01), while down motion increased by 245% at 33% (p=0.05); and 2) the variance of the HbO2’s magnitude for up motion decreased by 50% at 100% and 203% at 33% coherence, and for down motion decreased by 231% and 29%, respectively (p<0.01). Causal modeling identified two brain states in the lesioned left hemisphere (LH) associated with the discrimination performance: 1) a dominant state, associated with up motion at 100% coherence in 93% of the iNM trials, while at 33% it occurred in 67% of the iNM trials, while 60% of the control trials at 100% coherence and 58% were associated with 33% coherence (p=0.01); and 2) a non-dominant state associated with random motion greater in the control (5% at 100% and 10% at 33%) compared to iNM (0% at 100% and 4% at 33%) trials (p=0.01).
Conclusion
iNM strengthened sensory (cerebellum; ACC; cingulate), perceptual decision-making (orbitofrontal; basal ganglia, parahippocampal) and motion perception (precuneus) networks in R.S.’s LH for direction selectivity. These can serve as biomarkers in the rehabilitation of visual perception in CB patients as a sequalae of optic glioma tumor resection.