Zhang Jiawei, Huang Mingyi, Gu Yong, Chen Aihua, Yu Yuguo
Shanghai Artificial Intelligence Laboratory, Research Institute of Intelligent and Complex Systems and Institute of Science and Technology for Brain-Inspired Intelligence, State Key Laboratory of Medical Neurobiology and MOE Frontiers Center for Brain Science, Human Phenome Institute, Shanghai 200433, China.
Institute of Neuroscience, CAS Center for Excellence in Brain Science and Intelligence Technology, Chinese Academy of Sciences, Shanghai 200031, China.
Brain Sci. 2022 Oct 13;12(10):1387. doi: 10.3390/brainsci12101387.
Numerous studies have demonstrated that animal brains accurately infer whether multisensory stimuli are from a common source or separate sources. Previous work proposed that the multisensory neurons in the dorsal medial superior temporal area (MST-d) serve as integration or separation encoders determined by the tuning-response ratio. However, it remains unclear whether MST-d neurons mainly take a sense input as a spatial coordinate reference for carrying out multisensory integration or separation. Our experimental analysis shows that the preferred tuning response to visual input is generally larger than vestibular according to the Macaque MST-d neuronal recordings. This may be crucial to serving as the base of coordinate reference when the subject perceives moving direction information from two senses. By constructing a flexible Monte-Carlo probabilistic sampling (fMCS) model, we validate this hypothesis that the visual and vestibular cues are more likely to be integrated into a visual-based coordinate rather than vestibular. Furthermore, the property of the tuning gradient also affects decision-making regarding whether the cues should be integrated or not. To a dominant modality, an effective decision is produced by a steep response-tuning gradient of the corresponding neurons, while to a subordinate modality a steep tuning gradient produces a rigid decision with a significant bias to either integration or separation. This work proposes that the tuning response amplitude and tuning gradient jointly modulate which modality serves as the base coordinate for the reference frame and the direction change with which modality is decoded effectively.
大量研究表明，动物大脑能够准确推断多感官刺激是来自共同来源还是不同来源。先前的研究提出，背内侧颞上区（MST-d）的多感官神经元作为整合或分离编码器，由调谐-反应比率决定。然而，MST-d神经元是否主要将一种感官输入作为空间坐标参考来进行多感官整合或分离仍不清楚。我们的实验分析表明，根据猕猴MST-d神经元记录，对视觉输入的偏好调谐反应通常大于前庭反应。当受试者从两种感官感知运动方向信息时，这对于作为坐标参考基础可能至关重要。通过构建灵活的蒙特卡洛概率采样（fMCS）模型，我们验证了这一假设，即视觉和前庭线索更有可能被整合到基于视觉的坐标中，而不是前庭坐标中。此外，调谐梯度的特性也会影响关于线索是否应被整合的决策。对于主导模态，相应神经元的陡峭反应-调谐梯度会产生有效的决策，而对于从属模态，陡峭的调谐梯度会产生偏向整合或分离的显著偏差的刚性决策。这项工作提出，调谐反应幅度和调谐梯度共同调节哪种模态作为参考框架的基础坐标以及有效解码方向变化所依据的模态。
