孤独症（autism spectrum disorder）是一种以社交互动、语言沟通、认知加工损伤为主要症状的神经发育障碍。该疾病的临床异质性和发育特性给厘清其病因机制带来了重大挑战。多角度的研究揭示了脑皮层是按照梯度的形式进行组织和架构的，这个模式从单模态的躯体运动/视觉网络区域延伸至多模态的默认模式网络区域，反映了皮层的层级化空间布局，孤独症存在广泛的脑损伤，因此，从整体组织架构的视角探讨其神经基础具有必要性。脑皮层的结构和功能组织是在不断发育的，追踪和量化不同发育阶段皮层组织架构的变化，有助于把握正常脑的发育轨迹，进而深入解析相关疾病的神经基础异常。孤独症作为一种神经发育障碍，功能连接研究显示其存在年龄相关的发育转折，从儿童时期的“过连接”转变为青少年和成年阶段的“失连接”状态。研究孤独症脑的异常发育模式对于揭示其脑网络异常的异质性至关重要。因此，研究不同年龄组孤独症患者与正常对照在脑功能网络梯度上的差异，即探讨年龄和诊断在梯度上的不同交互模式，有助于揭示孤独症脑功能网络的异常机制。

本论文采用国际大型公开数据库孤独症脑影像数据交换（Autism Brain Imaging Data Exchange Dataset, ABIDE）14个中心的691名受试者的静息态脑功能磁共振影像数据，利用扩散映射嵌入的方法，考察了儿童组、青少年组和成人组的孤独症和正常对照的梯度模式，基于发育的框架，从全局指标和区域指标上对年龄与诊断在梯度上的交互模式进行了考察，探究了孤独症脑功能网络梯度的发育模式，并对梯度与临床症状和认知功能之间的关系进行了探索。

研究一考察了孤独症梯度的异常发育模式。儿童孤独症和正常对照的第一梯度为均沿着感觉运动网络渐进到视觉网络（躯体运动—视觉网络），第二梯度则是沿着渐进轴组织的初级视觉/感觉运动网络到默认模式网络（躯体运动/视觉—默认模式网络）。青少年组的梯度模式发生了翻转，第一梯度转变为躯体运动/视觉—默认模式网络，并一直延续到成人阶段。躯体运动/视觉—默认模式网络的全局指标和梯度分数上存在显著的诊断和发育的交互效应，儿童时期呈现梯度的过发育，青少年及成人时期则转为梯度的低发育。

研究二考察了不同年龄组孤独症梯度异常模式的临床与认知表现。首先，基于Neurosynth解码分析，本研究发现了躯体运动/视觉—默认模式网络的梯度分数的不同交互模式分别与自传体记忆、语言生成等认知功能的损伤有关；其次，儿童组孤独症的躯体运动/视觉—默认模式网络这一梯度的全局指标与刻板和重复性行为数呈显著正相关；本研究基于支持向量回归（Support Vector Regression, SVR）发现，躯体运动/视觉—默认模式网络的梯度分数在各个年龄组中都能预测孤独症社会交往缺陷相关的临床症状表现。

总的来说，本研究揭示了不同年龄组孤独症患者脑功能网络梯度的异常，揭示了躯体运动/视觉—默认模式网络这一全脑组织架构在孤独症上不同的诊断与发育的交互模式，并将其与临床表现和认知功能缺陷相关联，为阐明孤独症的脑的异常机制和临床表现的异质性提供了全新视野。

Autism Spectrum Disorder, also known as autism, is a neurodevelopmental disorder characterized by impairments in social interaction, language communication, and cognitive processing. The clinical heterogeneity and developmental features pose significant challenges in elucidating its underlying pathogenic mechanisms. Multiple researches have revealed that the human cerebral cortex is organized in a graded fashion, with a hierarchical spatial layout extending from somatomotor/visual network to default mode network. Autism patients present widespread brain abnormalities, thus necessitating an investigation of its neural underpinnings from the perspective of global organizational architecture. The structural and functional organization of the cerebral cortex is continuously developing. Tracking and quantifying alterations in cortical architectural organization across different developmental stages can help delineate the trajectory of brain maturation in healthy controls and elucidate the neurobiological underpinnings of related disorders. As a neurodevelopmental disorder, autism exhibits developmental shifts in functional connectivity, transiting from "hyper-connectivity" in childhood to "hypo-connectivity" in adolescence and adulthood. Investigating the abnormal developmental patterns of the autism brain is crucial for unraveling its neurobiological mechanisms and symptom origins. Consequently, examining differences in functional network gradients between autism patients and healthy controls across different age groups, specifically exploring the interaction between age and diagnosis on gradient patterns, can shed light on the the abnormal mechanisms underlying autism brain functional networks.

This study utilized resting-state functional magnetic resonance imaging data from 691 participants (children, adolescents, and adults with autism and healthy controls) across 14 centers in the Autism Brain Imaging Data Exchange (ABIDE) dataset. Employing diffusion map embedding, we investigated the gradient patterns in autism and healthy controls across age groups. Within a developmental framework, we investigated the interaction between age and diagnosis on global and regional gradient metrics, explored the developmental patterns of autism functional network gradients, and examined the relationships between gradients and clinical symptoms, and cognitive functions.

Study 1 revealed atypical developmental patterns of gradients in autism. In children, the primary gradient in both autism and controls progressed from the somatomotor network to the visual network , while the second gradient organized from the primary somatomotor/visual network to the default mode network. In adolescents, the gradient pattern underwent a reversal, with the first gradient becoming somatomotor/visual—default mode network. Significant diagnosis-by-age interactions were observed in both global measures and gradient scores of somatomotor/visual—default mode network, exhibiting gradient over-development in childhood autism, which progressively turned into under-development in adolescence and adulthood.

Study 2 investigated the clinical and cognitive manifestations of abnormal gradient patterns in autism across different age groups. Using Neurosynth decoding analysis, the study found that distinct interaction patterns of somatomotor/visual—default mode network gradient scores in autism were associated with impairments in autobiographical memory, language generation, and other cognitive functions. Furthermore, in the children's group, the global metric in autism was positively correlated with restricted and repetitive behavior scores. Additionally, based on support vector regression, the study revealed that somatomotor/visual—default mode network gradient scores could predict social interaction deficits, a core clinical symptom of autism, across all age groups.

In summary, this study uncovered abnormalities in functional network gradients in autism patients across different age groups, revealing distinct diagnosis-by-age interaction patterns in the somatomotor/visual—default mode network gradient, a key aspect of whole-brain organizational architecture. These findings were further linked to clinical manifestations and cognitive deficits, providing a novel perspective on elucidating the abnormal mechanisms and clinical heterogeneity of the autism brain.