Autism spectrum disorder (ASD) is a highly disabling mental disease that
brings significant impairments of social interaction ability to the patients,
making early screening and intervention of ASD critical. With the development
of the machine learning and neuroimaging technology, extensive research has
been conducted on machine classification of ASD based on structural MRI
(s-MRI). However, most studies involve with datasets where participants' age
are above 5. Few studies conduct machine classification of ASD for participants
below 5-year-old, but, with mediocre predictive accuracy. In this paper, we
push the boundary of predictive accuracy (above 0.97) of machine classification
of ASD in children (age range: 0.92-4.83 years), based on s-MRI features
extracted using contrastive variational autoencoder (CVAE). 78 s-MRI, collected
from Shenzhen Children's Hospital, are used for training CVAE, which consists
of both ASD-specific feature channel and common shared feature channel. The ASD
participants represented by ASD-specific features can be easily discriminated
from TC participants represented by the common shared features, leading to high
classification accuracy. In case of degraded predictive accuracy when data size
is extremely small, a transfer learning strategy is proposed here as a
potential solution. Finally, we conduct neuroanatomical interpretation based on
the correlation between s-MRI features extracted from CVAE and surface area of
different cortical regions, which discloses potential biomarkers that could
help target treatments of ASD in the future.

本文探讨使用对比变分自编码器（CVAE）提取结构磁共振成像（s-MRI）特征进行儿童自闭症谱系障碍（ASD）机器分类的方法，数据来自深圳市儿童医院，使用转移学习策略提高预测精度，并针对不同大脑区域进行了神经解剖解释。

基于对比变分自动编码器提取的结构 MRI 特征的儿童孤独症谱系障碍分类

Autism Spectrum Disorder Classification in Children based on Structural  MRI Features Extracted using Contrastive Variational Autoencoder

Structural MRI(S-MRI) is one of the most versatile imaging modality that
revolutionized the anatomical study of brain in past decades. The corpus
callosum (CC) is the principal white matter fibre tract, enabling all kinds of
inter-hemispheric communication. Thus, subtle changes in CC might be associated
with various neurological disorders. The present work proposes the potential of
YOLOv5-based CC detection framework to differentiate atypical Parkinsonian
disorders (PD) from healthy controls (HC). With 3 rounds of hold-out
validation, mean classification accuracy of 92% is obtained using the proposed
method on a proprietary dataset consisting of 20 healthy subjects and 20 cases
of APDs, with an improvement of 5% over SOTA methods (CC morphometry and visual
texture analysis) that used the same dataset. Subsequently, in order to
incorporate the explainability of YOLO predictions, Eigen CAM based heatmap is
generated for identifying the most important sub-region in CC that leads to the
classification. The result of Eigen CAM showed CC mid-body as the most
distinguishable sub-region in classifying APDs and HC, which is in-line with
SOTA methodologies and the current prevalent understanding in medicine.

本文提出了一种基于 YOLOv5 的 CC 检测框架，用于区分非典型帕金森病患者和健康对照组。使用 3 轮留置验证，在由 20 个健康受试者和 20 个 APD 病例组成的专有数据集上，所提出的方法获得了 92% 的平均分类准确率，并且相对于使用同一数据集的 SOTA 方法（CC 形态学和视觉纹理分析）提高了 5%。此外，为了包含 YOLO 预测的可解释性，还生成了基于 Eigen CAM 的热图，用于识别导致分类的最重要的 CC 亚区，结果显示 CC 中体是区分 APDs 和 HC 最显著的亚区，这与医学上的 SOTA 方法和当前普遍的理解相一致。