A computational framework for optimal control of a self-adjustive neural system with activity-dependent and homeostatic plasticity

한글 요약

치료 후 스스로 연결을 재조정하는 뇌의 성질을 반영한 최적 뇌 제어의 계산 틀을 제안한 연구입니다. 기존 뇌 제어 연구는 뇌를 수동적으로 반응하는 시스템으로 보았지만, 이 연구는 치료 후 신경계가 활동 의존 가소성으로 나가는 실효연결망을 조정하고 이어서 들어오는 연결을 항상성에 따라 재조정한다고 모델링했습니다. 원하는 기능을 유도하려면 먼저 이 자기 조정 매개변수를 추정한 뒤 그에 맞춰 치료를 최적화하는 방식입니다. 야생형과 돌연변이 생쥐의 전압감응염료 영상에 동적 인과 모델링(DCM)을 적용해 추정한 해마 회로에서 단일 노드 치료, 약물에 비유한 신경세포 유형별 치료, 두 치료의 병합을 시뮬레이션한 결과, 제한된 치료 조건에서도 비정상 회로를 건강한 회로 쪽으로 이끌 수 있음을 보였습니다. 치료 후 변화하는 뇌를 고려한 맞춤형 치료 설계의 이론적 기반을 제공하는 연구입니다.

The control of the brain system has received increasing attention in the domain of brain science. Most brain control studies have been conducted to explore the brain network's graph-theoretic properties or to produce the desired state based on neural state dynamics, regarding the brain as a passively responding system. However, the self-adjusting nature of neural system after treatment has not been fully considered in the brain control. In the present study, we propose a computational framework for optimal control of the brain with a self-adjustment process in the effective connectivity after treatment. The neural system is modeled to adjust its outgoing effective connectivity as activity-dependent plasticity after treatment, followed by synaptic rescaling of incoming effective connectivity. To control this neural system to induce the desired function, the system's self-adjustment parameter is first estimated, based on which the treatment is optimized. Utilizing this framework, we conducted simulations of optimal control over a functional hippocampal circuitry, estimated using dynamic causal modeling of voltage-sensitive dye imaging from the wild type and mutant mice, responding to consecutive electrical stimuli. Simulation results for optimal control of the abnormal circuit toward a healthy circuit using a single node treatment, neural-type specific treatment as an analogy of medication, and combined treatments of medication and nodal treatment suggest the plausibility of the current framework in controlling the self-adjusting neural system within a restricted treatment setting. We believe the proposed computational framework of the self-adjustment system would help optimal control of the dynamic brain after treatment.