Seth W. Egger博士於2026年獲選教育部「玉山青年學者」計畫，成為臺北醫學大學第2位玉山學者。他是神經科學家，研究專長為大腦神經迴路如何產生行為與思考。在美國加州大學戴維斯分校取得神經科學博士學位後，曾於麻省理工學院麥戈文腦科學研究所從事博士後研究，並於杜克大學醫學院神經生物學系擔任資深研究員。
他的研究結合神經生理學、行為實驗與計算分析，探討大腦如何在有限的感官資訊下建立「心智表徵」，使人類與動物能預測環境變化並做出精準行動，例如追蹤移動物體或模擬物體運動軌跡等複雜行為。未來，他將於北醫大推動神經迴路與心智表徵相關研究，結合大腦皮質、基底核與小腦等神經系統的研究，深化對大腦功能與神經疾病機制的理解，並期望將研究成果應用於神經退化性疾病與心理健康相關治療策略的發展。以下邀請他分享相關的學術歷程。（文/醫學院）【右圖：北醫大第2位玉山青年學者美籍的Seth W. Egger博士】
人類大腦由相互連結的神經元迴路組合而成，這些迴路將感官體驗轉化為複雜又合理的行為，這一轉化的關鍵構件是心智表徵的建構——亦即脫離直接感官刺激而生成現實世界的思維，例如，在西洋棋中，玩家以棋盤的心智表徵來模擬推算下一步的走法，而非實際移動棋子的動作，儘管神經迴路中的神經群體能夠產生獨立於感官之外的活動，但神經迴路解剖學、單一神經元活動和心智表徵之間的關聯依舊不明，故我的研究計劃目標在於彌合我們理解上的這一段差距。
為此，我將運用並拓展我在系統神經科學方面的訓練。我在加州大學戴維斯分校師從Kenneth Britten，取得神經科學博士學位，並開發了測量行為動力學及將單一神經元活動連結至行為的創新方法；之後，擔任麻省理工學院Mehrdad Jazayeri的博士後研究員，那時運用貝氏推論方法和神經群體分析來研究心理狀態；隨後我前往杜克大學，加入Stephen Lisberger的研究小組，擔任資深研究員，負責開發以大腦迴路解剖結構為基礎的皮質功能模型，在這3個實驗室，我所受的訓練使我能夠設計行為實驗模式，揭示神經迴路的特性並識別其功能背後的原理。
最近，我專注於在概念上類似棒球選手追蹤球路與擊球的研究設計，我和麻省理工學院的同事設計了一個實驗，在這個實驗中，棒球在其投球路徑上只能在兩個位置上被觀測到（圖1）， 要求這位打擊者在兩個觀察點之間，建構一個模擬棒球飛行軌跡的心智表徵，每一次觀察都讓打擊者能夠預估球速，在訓練猴子完成這項實驗後，我們記錄了猴子額葉皮質區域（稱為前輔助運動區preSMA）中數百個神經元的活動（圖2）。【圖1：打擊任務示意圖（上圖）】【圖2：猴腦結構示意圖（右圖），突顯重點區域】
為確定神經動力學是否能反映此種模擬，我們開發了在不同速度條件下，分析和比較神經元群活動的方法，我們應用主成分分析法，將高維度神經元群反應資料簡化為3個維度（PCs），以捕捉資料中最大的變異數，這使得我們可以將投球到擊球之間神經元群反應的軌跡視覺化（圖3）。
在觀察棒球之前，所有速度的軌跡都遵循相同的路徑（淺灰色軌跡），在第1次觀察（色彩圓圈）後，我們觀察到2種效應，首先，軌跡彼此錯開，但遵循相似的路徑（中灰色軌跡），其次，在投慢速球試驗中，軌跡運動減速；在投快速球試驗中，軌跡運動加速；第2次的觀察棒球（色彩圓圈），進一步調整軌跡位置和速度（深灰色軌跡），由於軌跡速度與位置之調整與猴子對投球速度的估計一致，因此我們得出結論，前輔助運動區（preSMA）神經元群活動，執行了對棒球的動態心智表徵。【圖3：前輔助運動區（preSMA）神經元群反應軌跡的視覺化軌跡（左圖）】
【圖4：反復發生迴路（藍色）及其訊號輸入（橘色、綠色和黃色）的示意圖】
大腦神經迴路解剖學如何啟動心智表徵的動態神經元群活動？要回答這個問題，需要剖析額葉內反復的連結動作（圖4/藍色）如何呼應來自系統外部刺激所引發的神經元群反應（圖4/橙色、綠色和黃色），我和杜克大學的同事們透過一項實驗來探討此問題，即是觀察猴子用眼睛追蹤或平順地捕捉到一個完全可視的棒球，數十年的研究顯示，視覺皮層區的MT區域及其下游前額葉皮質靶區，即前額葉眼動區（FEF）對於平順捕捉動作的執行至關重要。（圖2）
為了解視覺皮層MT區域訊號如何啟動前額葉眼動區（FEF）神經元群的活動（圖4/橙色），我們分析了兩個皮層區域神經元群活動的紀錄，進階的迴歸技術分析顯示每個前額葉眼動區神經元如何整合來自視覺皮層MT區域神經元群發出的訊號，雖然視覺皮層區域MT有數百萬個神經元，但其對前額葉眼動區活動的貢獻，僅以視覺皮層區域MT的2個維度，即可做出充分解釋。值得注意的是，這2個維度與我們先前關於視覺皮層區域MT神經元特性與平順捕捉行為之間關係的理論研究相符。
然而，視覺皮層MT區域發出訊號只能解釋棒球運動最初100毫秒前額葉眼動區的活動，隨後，視覺皮層MT區域活動衰減至基線水平，而前額葉眼動區活動和平順捕捉活動則持續存在，根據最新的神經網路理論，我們開發了一種新方法，用於識別反復發生的前額葉眼動區連結活動，如何將視覺皮層MT區域發出的訊號轉換為心智表徵。此分析揭示前額葉眼動區神經元群活動中2種心智表徵的形成和交互作用，一個用於追蹤棒球速度，另一個用於追蹤視覺皮層MT區域訊號，取得其敏感度，反復發生的連結所引發的活動解釋了視覺皮層MT區域訊號消失後，個別的前額葉眼動區神經元仍能持續反應的現象。
在臺北醫學大學，我將運用此支持心智表徵建構和操作的方法，來研究不同神經迴路路徑，包括新的大腦皮質間連接以及與多種神經退化性疾病相關的皮質下迴路、基底核（BG）和小腦（CB，圖2、圖4），此種方法將有助於我們從新的角度理解心智表徵的迴路基礎，以及迴路功能障礙對心理健康的影響，最終，我將運用本人研究專案所發現的原理，開發針對特定神經通路的新型治療方法。
Our brains are made up of interconnected circuits of neurons that transform sensory experiences into complex and intelligent behavior. A key component of this transformation is the construction of mental representations – thoughts about the physical world that can be separated from direct sensory stimulation. In chess, for example, players use a mental representation of the board to simulate upcoming moves rather than physically move the pieces. While populations of neurons in a circuit have the capacity to produce activity independent of sensation, the link between circuit anatomy, individual neuron activity, and mental representations remains unclear. The goal of my research program is to close this gap in our understanding.
To do so, I will deploy and build on my training in systems neuroscience. I earned my PhD in neuroscience as a graduate student with Kenneth Britten at the University of California, Davis. There I developed novel methods to measure behavioral dynamics and relate single neuron activity to behavior. I then joined Mehrdad Jazayeri at the Massachusetts Institute of Technology as a postdoctoral associate, where I studied mental states with Bayesian inference methods and neural population analysis. I then moved to Duke University to join Stephen Lisberger’s group as a senior research associate to develop models of cortical function based on the circuit anatomy of the brain. Across all three labs, my training has prepared me to design behavioral tasks that reveal neural circuit properties and identify the principles behind their function.
Recently, I have focused on tasks conceptually similar to a baseball player tracking and hitting a ball. With my colleagues at MIT, I developed a task where the ball was only observable at two locations along its path (Figure 1), requiring the batter to build a mental representation that simulated the ball’s flight between observations. Each observation allowed the hitter to estimate the pitch speed. After training monkeys to do this task, we recorded the activity of hundreds of neurons in the monkey frontal cortical area called preSMA (Figure 2).【Figure 1. Schematic of the hitting task】【Figure 2. Schematic of the monkey brain, highlighting areas of interest】
To determine if neural dynamics reflected this simulation, we developed methods to analyze and compare population activity from each speed condition. We applied principal component analysis to reduce the high dimensional population response data to the three dimensions (PCs) that capture the most variance in the data. This allows visualization of the trajectory of the population response from the time of the pitch to the hit (Figure 3). Before observing the ball, trajectories followed identical paths for all speeds (light gray). After the first observation (lower colored circles) we observed two effects. First, trajectories became offset from one another but followed similar paths (medium gray). Second, trajectory movement decelerated on trials with a slow pitch and accelerated on trials with a fast pitch. The second observation of the ball (upper colored circles) further adjusted trajectory position and speed (dark gray). Because adjustments in trajectory speed and position were consistent with the monkey’s estimates of pitch speed, we concluded that preSMA population activity executed a dynamic mental representation of the ball. 【Figure 3. Visualization of preSMA population response trajectories】
How does the circuit anatomy of the brain give rise to dynamic population activity that supports mental representation? Answering this requires dissecting the contributions of recurrent frontal connections (Figure 4, blue) to the population response from external inputs to the system (Figure 4, orange, green, and yellow). Together with my colleagues at Duke, we approached this question in a task where monkeys tracked, or smoothly pursued, a fully visible ball with their eyes. Decades of research identify MT, a visual cortical area, and its downstream frontal cortical target, FEF, as critical to the execution of smooth pursuit (Figure 2).【Figure 4. Illustration of a recurrent circuit (blue) and its inputs (orange, green, and yellow)】
To understand the contribution of MT inputs to FEF population activity (Figure 4, orange), we analyzed population activity recorded from both cortical areas. Advanced regression techniques revealed how each FEF neuron integrated inputs from across the MT population. Even though MT has millions of neurons, its contribution to FEF activity was best described by only two dimensions of MT activity. Remarkably, these two dimensions agreed with our previous theoretical work relating the properties of MT neurons to smooth pursuit behavior.
However, MT inputs could only explain FEF activity over the first 100 ms of ball movement. After that, MT activity decayed to baseline while FEF activity and smooth pursuit continued to persist. Building on advancements in neural network theory, we developed a novel method to identify how recurrent FEF connections transform MT inputs into mental representations. This analysis revealed the formation and interaction of two mental representations in FEF population activity. One that kept track of the ball’s speed and a second that kept track of pursuit sensitivity to MT input. The activity generated by recurrent connections explained the persistent responses of individual FEF neurons after MT input was extinguished.
At TMU, I will apply the above approaches to the study of different neural circuit pathways that support the construction and manipulation of mental representations. This will include new cortical-cortical connections as well as subcortical circuits implicated in several neurodegenerative diseases, the basal ganglia (BG) and cerebellum (CB, Figures 2 and 4). This approach will allow new understanding of the circuit basis of mental representations and the consequences of circuit dysfunction on mental health. Ultimately, I will apply principles uncovered by my research program to develop novel treatment approaches that target specific neural pathways.
