Label Distribution Learning for Computer Aided Diagnosis of Multi-class ASD Classification

doi:10.3778/j.issn.1673-9418.2007038

Journal of Frontiers of Computer Science and Technology ›› 2022, Vol. 16 ›› Issue (1): 194-204.DOI: 10.3778/j.issn.1673-9418.2007038

• Artificial Intelligence • Previous Articles Next Articles

Label Distribution Learning for Computer Aided Diagnosis of Multi-class ASD Classification

ZHANG Fengyexin¹, WANG Jun²^,⁺(), JIA Xiuyi³, PAN Xiang¹, DENG Zhaohong¹, SHI Jun², WANG Shitong¹

1. School of Artificial Intelligence and Computer Science, Jiangnan University, Wuxi, Jiangsu 214122, China
2. School of Communication and Information Engineering, Shanghai University, Shanghai 200444, China
3. School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing 210094, China

Received:2020-07-13 Revised:2020-11-03 Online:2022-01-01 Published:2020-11-25
About author:ZHANG Fengyexin, born in 1996, M.S. candidate. His research interests include artificial intelligence and pattern recognition.
WANG Jun, born in 1978, associate professor. His research interests include pattern recognition and artificial intelligence.
JIA Xiuyi, born in 1983, associate professor. His research interests include machine learning, granular computing and data mining.
PAN Xiang, born in 1984, associate professor. His research interest is medical image processing.
DENG Zhaohong, born in 1981, professor. His research interests include uncertainty artificial intelligence and its applications.
SHI Jun, born in 1977, professor. His research interests include uncertainty medical image analysis, medical signal processing, etc.
WANG Shitong, born in 1964, professor. His research interests include artificial intelligence, pattern recognition, etc.
Supported by:
National Natural Science Foundation of China(61773208);Natural Science Foundation of Jiangsu Province(BK20181339);Natural Science Foundation of Jiangsu Province(BK20191287)

面向多分类自闭症辅助诊断的标记分布学习

章枫叶欣¹, 王骏²^,⁺(), 贾修一³, 潘祥¹, 邓赵红¹, 施俊², 王士同¹

1.江南大学人工智能与计算机学院,江苏无锡 214122
2.上海大学通信与信息工程学院,上海 200444
3.南京理工大学计算机科学与工程学院,南京 210094

通讯作者: + E-mail: wangjun_shu@shu.edu.cn
作者简介:章枫叶欣（1996—）,男,浙江丽水人,硕士研究生,主要研究方向为人工智能、模式识别。
王骏（1978—）,男,江苏苏州人,副教授,主要研究方向为模式识别、人工智能。
贾修一（1983—）,男,副教授,主要研究方向为机器学习、粒计算、数据挖掘。
潘祥（1984—）,男,副教授,主要研究方向为医学图像处理。
邓赵红（1981—）,男,安徽蒙城人,教授,主要研究方向为不确定性人工智能及其应用。
施俊（1977—）,男,教授,主要研究方向为不确定性医学图像分析、医学信号处理等。
王士同（1964—）,男,江苏扬州人,教授,主要研究方向为人工智能、模式识别等。
基金资助:
国家自然科学基金(61773208);江苏省自然科学基金(BK20181339);江苏省自然科学基金(BK20191287)

Abstract

Abstract:

Autism spectrum disorder (ASD) is a series of complex neurodevelopmental disorders, including several diseases related to developmental disorders, but most of the existing diagnosis methods for autism are binary classification methods which cannot meet the actual needs. In addition, the label noise contained in ASD data, as well as the characteristics of high dimensionality and data imbalance, has brought great challenges to traditional methods. To this end, a new computer aided diagnosis method of ASD is proposed. This method solves the label noise by introducing label distribution learning (LDL), introduces a cost-sensitive mechanism to solve the data imbalance, uses label distribution support vector regression (SVR) to solve the classification difficulties caused by high-dimensional features by mapping samples to the feature space, and finally realizes the computer aided diagnosis of multi-class ASD. Experimental results show that compared with the existing methods, the proposed method overcomes the imbalance of the influence of the majority class and the minority class on the results, can effectively solve the class imbalance in ASD diagnosis, and has better and stable classification performance, which can assist in the diagnosis of ASD.

Key words: computer aided diagnosis of autism spectrum disorder (ASD), cost-sensitive mechanism, label distribution learning (LDL), support vector regression (SVR)

摘要：

自闭症谱性障碍（ASD）是一系列复杂的神经发展障碍性疾病,其包括若干与发育障碍相关的疾病,但是现有的自闭症辅助诊断方法大多是二分类方法,无法满足现实的需要。此外,ASD数据包含的标记噪声,以及高维度、数据分布不平衡等特点给传统分类方法带来了巨大的挑战。为此,提出一种新型的ASD辅助诊断方法,该方法通过引入标记分布学习（LDL）来解决标记噪声问题,引入代价敏感机制来解决样本不平衡问题,并采用基于支持向量回归（SVR）的标记分布学习方法,通过将样本映射到特征空间,解决高维特征带来的分类困难,最终实现多分类ASD的辅助诊断。实验结果表明,与已有方法比较,所提方法克服了多数类和少数类对结果的影响的不平衡性,可以有效地解决ASD诊断中的不平衡数据问题,拥有更好且稳定的分类性能,可以辅助ASD的诊断。

关键词: 自闭症谱性障碍（ASD）辅助诊断, 代价敏感机制, 标记分布学习（LDL）, 支持向量回归（SVR）

CLC Number:

TP181

ZHANG Fengyexin, WANG Jun, JIA Xiuyi, PAN Xiang, DENG Zhaohong, SHI Jun, WANG Shitong. Label Distribution Learning for Computer Aided Diagnosis of Multi-class ASD Classification[J]. Journal of Frontiers of Computer Science and Technology, 2022, 16(1): 194-204.

章枫叶欣, 王骏, 贾修一, 潘祥, 邓赵红, 施俊, 王士同. 面向多分类自闭症辅助诊断的标记分布学习[J]. 计算机科学与探索, 2022, 16(1): 194-204.

Figures/Tables 9

Fig.1 Flowchart of cost sensitive label distribution support vector regression

Table 1 Evaluation measures

	指标	公式
标记分布指标	Chebyshev↓	$Di s 1 = max j ∈ [1, K] \| d i j - d ˆ i j \|$
	KL↓	$Di s 2 = ∑ j = 1 K d i j ln d i j d ˆ i j$
	Clark↓	$Di s 3 = ∑ j = 1 K (d i j - d ˆ i j) 2 (d i j + d ˆ i j) 2$
	Canberra↓	$Di s 4 = ∑ j = 1 K \| d i j - d ˆ i j \| \| d i j \| + \| d ˆ i j \|$
	Intersection↑	$Si m 1 = ∑ j = 1 K min (d i j, d ˆ i j)$
	Cosine↑	$Si m 2 = d i ⋅ d ˆ i \| d i \| ⋅ \| d ˆ i \|$
多分类指标	Precision↑	$P = 1 N ∑ i = 1 N xnor (l i, l ˆ i)$
多分类指标	mAP↑	$mAP = 1 K ∑ j = 1 K P j$

Table 1 Evaluation measures

	指标	公式
标记分布指标	Chebyshev↓	$Di s 1 = max j ∈ [1, K] \| d i j - d ˆ i j \|$
	KL↓	$Di s 2 = ∑ j = 1 K d i j ln d i j d ˆ i j$
	Clark↓	$Di s 3 = ∑ j = 1 K (d i j - d ˆ i j) 2 (d i j + d ˆ i j) 2$
	Canberra↓	$Di s 4 = ∑ j = 1 K \| d i j - d ˆ i j \| \| d i j \| + \| d ˆ i j \|$
	Intersection↑	$Si m 1 = ∑ j = 1 K min (d i j, d ˆ i j)$
	Cosine↑	$Si m 2 = d i ⋅ d ˆ i \| d i \| ⋅ \| d ˆ i \|$
多分类指标	Precision↑	$P = 1 N ∑ i = 1 N xnor (l i, l ˆ i)$
多分类指标	mAP↑	$mAP = 1 K ∑ j = 1 K P j$

Table 2 Statistics of datasets

数据集	样本数	正常	自闭症	亚斯伯格症候群
NYU	177	103	53	21
UM	144	76	57	11
KKI	48	31	8	9
Leuven	102	62	21	19
UCLA	81	53	16	12

Table 3 Comparison algorithms

对比算法名称	对比算法说明
PT-SVM^[21]	基于问题转化的SVM,利用改进的Platt后验概率得到二值向量机的概率,通过逐对耦合多类方法得到预测的标记分布
PT-BAYES^[21]	基于问题转化的BAYES假设每个类都服从Gauss分布,由此计算后验概率作为预测的标记分布
AA-KNN^[21]	基于算法改造的KNN将K个近邻的标记的均值作为预测的标记分布
AA-BP^[21]	基于算法改造的BP神经网络利用softmax激活输出,作为预测的标记分布
SA-IIS^[21]	基于专用算法的IIS使用一种改进迭代尺度算法优化目标函数
LDSVR^[30]	基于专用算法的LDSVR使用了支持向量回归改进目标函数,并用拟牛顿法优化
Decision Tree^[33]	基于实例的归纳学习方法,能从给定的无序的训练样本中,提炼出树型的分类模型
KNN^[34]	基于实例的分类方法,借由样本在特征空间中的K个最相邻的样本来为样本分类

Table 4 Range of parameters

参数名	参数范围
权重系数	0.001,0.01,0.1,1,10,100,1 000
核函数的类型	线性核、多项式核、高斯核
不敏感区大小	0.000 1,0.001,0.01,0.1
高斯核的核带宽	0.01,0.1,1,10,100

Table 5 Performance comparison of CSLDSVR and LDL algorithms

评估指标	算法	NYU	UM	Leuven	UCLA	KKI
Chebyshev↓	AA-BP	0.223 7±0.035 6	0.218 4±0.045 8	0.248 0±0.044 6	0.250 6±0.053 5	0.254 7±0.052 9
	AA-KNN	0.144 1±0.011 6	0.154 0±0.021 1	0.157 9±0.026 5	0.142 6±0.031 3	0.157 2±0.029 5
	LDSVR	0.150 1±0.024 3	0.140 0±0.012 8	0.162 9±0.034 4	0.169 4±0.053 4	0.160 2±0.057 0
	SA-IIS	0.147 8±0.011 8	0.153 5±0.023 7	0.174 8±0.021 4	0.145 8±0.032 9	0.162 7±0.049 5
	PT-BAYES	0.381 8±0.111 9	0.205 7±0.009 5	0.206 9±0.007 8	0.213 5±0.009 9	0.215 4±0.008 1
	PT-SVM	0.200 5±0.041 2	0.188 5±0.042 3	0.183 1±0.040 1	0.195 8±0.033 0	0.182 2±0.058 9
	CSLDSVR	0.141 3±0.016 2	0.135 2±0.023 6	0.140 2±0.024 4	0.138 6±0.038 4	0.126 7±0.034 9
Cosine↑	AA-BP	0.873 1±0.034 4	0.881 8±0.035 6	0.862 2±0.049 8	0.839 9±0.057 8	0.843 7±0.058 6
	AA-KNN	0.935 4±0.009 6	0.928 6±0.017 3	0.927 4±0.020 8	0.929 7±0.022 4	0.913 0±0.024 4
	LDSVR	0.937 7±0.019 1	0.944 8±0.013 3	0.932 5±0.029 2	0.928 5±0.052 0	0.932 6±0.047 4
	SA-IIS	0.940 7±0.009 3	0.934 4±0.016 7	0.920 5±0.016 0	0.939 5±0.020 3	0.924 6±0.042 5
	PT-BAYES	0.798 5±0.071 3	0.915 6±0.006 2	0.915 1±0.005 3	0.910 4±0.006 9	0.909 2±0.005 7
	PT-SVM	0.898 7±0.038 5	0.904 3±0.042 8	0.914 5±0.030 9	0.897 4±0.036 5	0.906 8±0.045 8
	CSLDSVR	0.940 5±0.012 1	0.947 3±0.018 3	0.923 4±0.025 5	0.942 8±0.036 8	0.936 3±0.029 4
Clark↓	AA-BP	0.468 1±0.064 8	0.461 3±0.099 0	0.517 0±0.083 8	0.537 1±0.110 1	0.542 7±0.104 6
	AA-KNN	0.263 1±0.020 3	0.282 2±0.036 7	0.287 3±0.047 3	0.261 3±0.053 5	0.283 2±0.053 9
	LDSVR	0.272 9±0.036 4	0.255 7±0.021 8	0.287 2±0.062 6	0.295 6±0.092 0	0.281 9±0.100 8
	SA-IIS	0.266 3±0.019 1	0.278 8±0.039 7	0.311 3±0.033 6	0.262 3±0.055 5	0.293 9±0.088 0
	PT-BAYES	0.893 6±0.359 8	0.352 0±0.014 5	0.352 3±0.012 7	0.363 6±0.016 2	0.366 3±0.013 3
	PT-SVM	0.358 0±0.070 2	0.348 1±0.075 8	0.325 3±0.065 5	0.350 5±0.056 1	0.328 7±0.098 1
	CSLDSVR	0.261 6±0.032 1	0.246 3±0.037 6	0.253 9±0.041 8	0.248 4±0.062 6	0.233 4±0.061 8
Canberra↓	AA-BP	0.717 6±0.102 8	0.710 4±0.152 9	0.810 8±0.125 6	0.811 8±0.169 0	0.831 9±0.160 2
	AA-KNN	0.406 6±0.029 4	0.429 6±0.057 3	0.447 5±0.070 7	0.398 4±0.084 9	0.438 0±0.083 1
	LDSVR	0.432 1±0.064 3	0.399 8±0.039 2	0.469 9±0.097 7	0.493 5±0.129 8	0.464 0±0.159 8
	SA-IIS	0.426 8±0.034 2	0.438 4±0.067 8	0.502 2±0.060 1	0.422 4±0.093 5	0.469 6±0.127 4
	PT-BAYES	1.472 1±0.574 8	0.602 2±0.026 9	0.604 8±0.022 9	0.624 5±0.029 1	0.629 9±0.023 9
	PT-SVM	0.573 6±0.106 2	0.541 1±0.127 7	0.522 7±0.111 0	0.551 0±0.087 4	0.511 5±0.165 9
	CSLDSVR	0.386 3±0.047 1	0.387 5±0.061 7	0.393 8±0.069 5	0.402 3±0.109 2	0.354 4±0.093 9
Intersection↑	AA-BP	0.776 3±0.035 6	0.781 6±0.045 8	0.752 0±0.044 6	0.749 4±0.053 5	0.745 3±0.052 9
	AA-KNN	0.855 9±0.011 6	0.846 0±0.021 1	0.842 1±0.026 5	0.857 4±0.031 3	0.842 8±0.029 5
	LDSVR	0.849 9±0.024 3	0.860 0±0.012 8	0.837 1±0.034 4	0.830 6±0.053 4	0.839 8±0.057 0
	SA-IIS	0.852 2±0.011 8	0.846 5±0.023 7	0.825 2±0.021 4	0.854 2±0.032 9	0.837 3±0.049 5
	PT-BAYES	0.618 2±0.111 9	0.794 3±0.009 5	0.793 1±0.007 8	0.786 5±0.009 9	0.784 6±0.008 1
	PT-SVM	0.799 5±0.041 2	0.811 5±0.042 3	0.816 9±0.040 1	0.804 2±0.033 0	0.817 8±0.058 9
	CSLDSVR	0.858 7±0.041 5	0.864 8±0.023 6	0.859 8±0.024 4	0.861 4±0.038 4	0.873 3±0.034 9
KL↑	AA-BP	0.166 7±0.042 9	0.161 2±0.051 7	0.192 0±0.069 3	0.222 2±0.089 8	0.227 9±0.076 4
	AA-KNN	0.068 5±0.010 1	0.076 0±0.018 4	0.076 6±0.022 1	0.074 6±0.023 2	0.093 2±0.026 6
	LDSVR	0.066 5±0.019 9	0.059 3±0.014 6	0.070 3±0.032 3	0.074 9±0.062 5	0.071 1±0.049 8
	SA-IIS	0.063 9±0.009 3	0.069 8±0.017 8	0.083 7±0.016 6	0.063 9±0.021 0	0.080 0±0.044 1
	PT-BAYES	0.492 9±0.251 0	0.087 9±0.006 7	0.088 0±0.006 1	0.093 5±0.008 0	0.094 8±0.006 6
	PT-SVM	0.108 1±0.041 2	0.105 5±0.047 6	0.090 6±0.032 9	0.110 4±0.040 5	0.100 3±0.048 1
	CSLDSVR	0.060 3±0.041 5	0.056 7±0.019 5	0.069 9±0.024 0	0.060 1±0.046 1	0.068 2±0.030 1

Table 5 Performance comparison of CSLDSVR and LDL algorithms

评估指标	算法	NYU	UM	Leuven	UCLA	KKI
Chebyshev↓	AA-BP	0.223 7±0.035 6	0.218 4±0.045 8	0.248 0±0.044 6	0.250 6±0.053 5	0.254 7±0.052 9
	AA-KNN	0.144 1±0.011 6	0.154 0±0.021 1	0.157 9±0.026 5	0.142 6±0.031 3	0.157 2±0.029 5
	LDSVR	0.150 1±0.024 3	0.140 0±0.012 8	0.162 9±0.034 4	0.169 4±0.053 4	0.160 2±0.057 0
	SA-IIS	0.147 8±0.011 8	0.153 5±0.023 7	0.174 8±0.021 4	0.145 8±0.032 9	0.162 7±0.049 5
	PT-BAYES	0.381 8±0.111 9	0.205 7±0.009 5	0.206 9±0.007 8	0.213 5±0.009 9	0.215 4±0.008 1
	PT-SVM	0.200 5±0.041 2	0.188 5±0.042 3	0.183 1±0.040 1	0.195 8±0.033 0	0.182 2±0.058 9
	CSLDSVR	0.141 3±0.016 2	0.135 2±0.023 6	0.140 2±0.024 4	0.138 6±0.038 4	0.126 7±0.034 9
Cosine↑	AA-BP	0.873 1±0.034 4	0.881 8±0.035 6	0.862 2±0.049 8	0.839 9±0.057 8	0.843 7±0.058 6
	AA-KNN	0.935 4±0.009 6	0.928 6±0.017 3	0.927 4±0.020 8	0.929 7±0.022 4	0.913 0±0.024 4
	LDSVR	0.937 7±0.019 1	0.944 8±0.013 3	0.932 5±0.029 2	0.928 5±0.052 0	0.932 6±0.047 4
	SA-IIS	0.940 7±0.009 3	0.934 4±0.016 7	0.920 5±0.016 0	0.939 5±0.020 3	0.924 6±0.042 5
	PT-BAYES	0.798 5±0.071 3	0.915 6±0.006 2	0.915 1±0.005 3	0.910 4±0.006 9	0.909 2±0.005 7
	PT-SVM	0.898 7±0.038 5	0.904 3±0.042 8	0.914 5±0.030 9	0.897 4±0.036 5	0.906 8±0.045 8
	CSLDSVR	0.940 5±0.012 1	0.947 3±0.018 3	0.923 4±0.025 5	0.942 8±0.036 8	0.936 3±0.029 4
Clark↓	AA-BP	0.468 1±0.064 8	0.461 3±0.099 0	0.517 0±0.083 8	0.537 1±0.110 1	0.542 7±0.104 6
	AA-KNN	0.263 1±0.020 3	0.282 2±0.036 7	0.287 3±0.047 3	0.261 3±0.053 5	0.283 2±0.053 9
	LDSVR	0.272 9±0.036 4	0.255 7±0.021 8	0.287 2±0.062 6	0.295 6±0.092 0	0.281 9±0.100 8
	SA-IIS	0.266 3±0.019 1	0.278 8±0.039 7	0.311 3±0.033 6	0.262 3±0.055 5	0.293 9±0.088 0
	PT-BAYES	0.893 6±0.359 8	0.352 0±0.014 5	0.352 3±0.012 7	0.363 6±0.016 2	0.366 3±0.013 3
	PT-SVM	0.358 0±0.070 2	0.348 1±0.075 8	0.325 3±0.065 5	0.350 5±0.056 1	0.328 7±0.098 1
	CSLDSVR	0.261 6±0.032 1	0.246 3±0.037 6	0.253 9±0.041 8	0.248 4±0.062 6	0.233 4±0.061 8
Canberra↓	AA-BP	0.717 6±0.102 8	0.710 4±0.152 9	0.810 8±0.125 6	0.811 8±0.169 0	0.831 9±0.160 2
	AA-KNN	0.406 6±0.029 4	0.429 6±0.057 3	0.447 5±0.070 7	0.398 4±0.084 9	0.438 0±0.083 1
	LDSVR	0.432 1±0.064 3	0.399 8±0.039 2	0.469 9±0.097 7	0.493 5±0.129 8	0.464 0±0.159 8
	SA-IIS	0.426 8±0.034 2	0.438 4±0.067 8	0.502 2±0.060 1	0.422 4±0.093 5	0.469 6±0.127 4
	PT-BAYES	1.472 1±0.574 8	0.602 2±0.026 9	0.604 8±0.022 9	0.624 5±0.029 1	0.629 9±0.023 9
	PT-SVM	0.573 6±0.106 2	0.541 1±0.127 7	0.522 7±0.111 0	0.551 0±0.087 4	0.511 5±0.165 9
	CSLDSVR	0.386 3±0.047 1	0.387 5±0.061 7	0.393 8±0.069 5	0.402 3±0.109 2	0.354 4±0.093 9
Intersection↑	AA-BP	0.776 3±0.035 6	0.781 6±0.045 8	0.752 0±0.044 6	0.749 4±0.053 5	0.745 3±0.052 9
	AA-KNN	0.855 9±0.011 6	0.846 0±0.021 1	0.842 1±0.026 5	0.857 4±0.031 3	0.842 8±0.029 5
	LDSVR	0.849 9±0.024 3	0.860 0±0.012 8	0.837 1±0.034 4	0.830 6±0.053 4	0.839 8±0.057 0
	SA-IIS	0.852 2±0.011 8	0.846 5±0.023 7	0.825 2±0.021 4	0.854 2±0.032 9	0.837 3±0.049 5
	PT-BAYES	0.618 2±0.111 9	0.794 3±0.009 5	0.793 1±0.007 8	0.786 5±0.009 9	0.784 6±0.008 1
	PT-SVM	0.799 5±0.041 2	0.811 5±0.042 3	0.816 9±0.040 1	0.804 2±0.033 0	0.817 8±0.058 9
	CSLDSVR	0.858 7±0.041 5	0.864 8±0.023 6	0.859 8±0.024 4	0.861 4±0.038 4	0.873 3±0.034 9
KL↑	AA-BP	0.166 7±0.042 9	0.161 2±0.051 7	0.192 0±0.069 3	0.222 2±0.089 8	0.227 9±0.076 4
	AA-KNN	0.068 5±0.010 1	0.076 0±0.018 4	0.076 6±0.022 1	0.074 6±0.023 2	0.093 2±0.026 6
	LDSVR	0.066 5±0.019 9	0.059 3±0.014 6	0.070 3±0.032 3	0.074 9±0.062 5	0.071 1±0.049 8
	SA-IIS	0.063 9±0.009 3	0.069 8±0.017 8	0.083 7±0.016 6	0.063 9±0.021 0	0.080 0±0.044 1
	PT-BAYES	0.492 9±0.251 0	0.087 9±0.006 7	0.088 0±0.006 1	0.093 5±0.008 0	0.094 8±0.006 6
	PT-SVM	0.108 1±0.041 2	0.105 5±0.047 6	0.090 6±0.032 9	0.110 4±0.040 5	0.100 3±0.048 1
	CSLDSVR	0.060 3±0.041 5	0.056 7±0.019 5	0.069 9±0.024 0	0.060 1±0.046 1	0.068 2±0.030 1

Fig.2 Comparison of classification performance of CSLDSVR and label distribution algorithms

Table 6 Performance comparison of CSLDSVR and multi-classification algorithms

数据集	Decision Tree		KNN		CSLDSVR
数据集	Precision	mAP	Precision	mAP	Precision	mAP
NYU	0.548 8±0.142 3	0.409 3±0.070 3	0.614 4±0.152 5	0.364 7±0.052 7	0.655 4±0.057 1	0.451 7±0.039 8
UM	0.576 7±0.132 5	0.385 9±0.087 2	0.528 5±0.121 4	0.374 0±0.086 1	0.701 4±0.070 8	0.497 1±0.125 0
Leuven	0.617 1±0.226 1	0.424 2±0.208 6	0.608 5±0	0.333 3±0	0.617 6±0.072 5	0.448 2±0.086 1
UCLA	0.605 2±0.183 3	0.442 0±0.208 6	0.654 3±0	0.333 3±0	0.665 2±0.150 4	0.443 4±0.165 9
KKI	0.559 8±0.256 7	0.395 4±0.294 1	0.646 5±0	0.333 3±0	0.687 5±0.123 7	0.447 6±0.101 6

Fig.3 Sensitivity analysis of parameters Cand ε on 5 datasets

References 36

[1]	DUAN X, CHEN H, HE C, et al. Resting-state functional under-connectivity within and between large-scale cortical networks across three low-frequency bands in adolescents with autism[J]. Progress in Neuro-Psychopharmacology and Biological Psychiatry, 2017, 79(B):434-441. DOI URL
[2]	DICKIE E W, AMEIS S H, SHAHAB S, et al. Personalised intrinsic network topography mapping and functional con-nectivity deficits in autism spectrum disorder[J]. Biological Psychiatry, 2018, 84(4):278-286. DOI URL
[3]	WOZNIAK R H, LEEZENBAUM N B, NORTHRUP J B, et al. The development of autism spectrum disorders: varia-bility and causal complexity: development of autism spectrum disorders[J]. Wiley Interdisciplinary Reviews Cognitive Sci-ence, 2016, 8(1):27906524.
[4]	ALI E M, AL-ADWAN F E Z, AL-NAIMAT Y M. Autism spectrum disorder (ASD); symptoms, causes, diagnosis, in-tervention, and counseling needs of the families in Jordan[J]. Modern Applied Science, 2019, 13(5):48-56.
[5]	MINSHEW N J, KELLER T A. The nature of brain dysfunc-tion in autism: functional brain imaging studies[J]. Current Opinion in Neurology, 2010, 23(2):124-130. DOI URL
[6]	STIGLER K A, MCDONALD B C, ANAND A, et al. Struc-tural and functional magnetic resonance imaging of autism spectrum disorders[J]. Brain Research, 2011, 1380(12):146-161. DOI URL
[7]	PUERTO E, AGUILAR J, LÓPEZ C, et al. Using multilayer fuzzy cognitive maps to diagnose autism spectrum disorder[J]. Applied Soft Computing, 2018, 75:58-71. DOI URL
[8]	LI X X, DVORNEK N C, ZHUANG J T, et al. Brain bio-marker interpretation in ASD using deep learning and fMRI[C]// LNCS 11072: Proceedings of the 21st International Con-ference on Medical Image Computing and Computer-Assisted Intervention, Granada, Sep 16-20, 2018. Cham: Springer, 2018: 206-214.
[9]	ECKER C, MARQUAND A, MOURAO-MIRANDA J, et al. Describing the brain in autism in five dimensions-magnetic resonance imaging-assisted diagnosis of autism spectrum disorder using a multiparameter classification approach[J]. Journal of Neuroscience, 2010, 30(32):10612-10623. DOI URL
[10]	WANG J, QIAN W, HAN Z, et al. Sparse multiview task-centralized ensemble learning for ASD diagnosis based on age- and sex-related functional connectivity patterns[J]. IEEE Transactions on Cybernetics, 2019, 49(8):3141-3154. DOI URL
[11]	张春香, 王骏, 张嘉旭, 等. 面向自闭症辅助诊断的联合组稀疏TSK建模方法[J]. 计算机科学与探索, 2020, 14(12):2083-2093.
	ZHANG C X, WANG J, ZHANG J X, et al. Novel TSK modeling method with joint group sparse learning for au-tism aided diagnosis[J]. Journal of Frontiers of Computer Science and Technology, 2020, 14(12):2083-2093.
[12]	张英, 王骏, 鲍国强, 等. 面向自闭症辅助诊断的无监督模糊特征学习新方法[J]. 智能系统学报, 2019, 14(5):882-888.
	ZHANG Y, WANG J, BAO G Q, et al. A novel unsupervised fuzzy feature learning method for computer-aided diagnosis of autism[J]. CAAI Transactions on Intelligent Systems, 2019, 14(5):882-888.
[13]	CHEN X, ZHANG H, GAO Y, et al. High-order resting-state functional connectivity network for MCI classification[J]. Human Brain Mapping, 2016, 37(9):3282-3296. DOI URL
[14]	AGGARWAL P, GUPTA A. Multivariate graph learning for detecting aberrant connectivity of dynamic brain networks in autism[J]. Medical Image Analysis, 2019, 56:11-25. DOI URL
[15]	HEINSFELD A S, FRANCO A R, CRADDOCK R C. Iden-tification of autism spectrum disorder using deep learning and the ABIDE dataset[J]. NeuroImage: Clinical, 2018, 17:16-23. DOI URL
[16]	American Psychiatric Association. Diagnostic and statistical manual of mental disorders[M]. Washington: American Psy-chiatric Association, 1994.
[17]	FRÉNAY B, VERLEYSEN M. Classification in the presence of label noise: a survey[J]. IEEE Transactions on Neural Networks and Learning Systems, 2014, 25(5):845-869. DOI URL
[18]	WU F, JING X Y, SHAN S G. Multiset feature learning for highly imbalanced data classification[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2021, 43(1):139-156. DOI URL
[19]	GAO B B, XING C, XIE C W, et al. Deep label distribution learning with label ambiguity[J]. IEEE Transactions on Image Processing, 2016, 26(6):2825-2838. DOI URL
[20]	JIA X Y, REN T T, CHEN L, et al. Weakly supervised label distribution learning based on transductive matrix completion with sample correlations[J]. Pattern Recognition Letters, 2019, 125:453-462. DOI URL
[21]	GENG X. Label distribution learning[J]. IEEE Transactions on Knowledge and Data Engineering, 2016, 28(7):1734-1748. DOI URL
[22]	GIBAJA E, VENTURA S. A tutorial on multilabel learning[J]. ACM Computing Surveys, 2015, 47(3):1-38.
[23]	ZHANG M L, ZHOU Z H. A review on multi-label learn-ing algorithms[J]. IEEE Transactions on Knowledge and Data Engineering, 2014, 26(8):1819-1837. DOI URL
[24]	ZHOU D Y, ZHANG X, ZHOU Y, et al. Emotion distribu-tion learning from texts[C]// Proceedings of the 2016 Con-ference on Empirical Methods in Natural Language Process-ing, Austin, Nov 1-4, 2016. Stroudsburg: ACL, 2016: 638-647.
[25]	GENG X, XIA Y. Head pose estimation based on multivariate label distribution[C]// Proceedings of the 2014 IEEE Confer-ence on Computer Vision and Pattern Recognition, Columbus, Jun 23-28, 2014. Washington: IEEE Computer Society, 2014: 1837-1842.
[26]	XU N, TAO A, GENG X. Label enhancement for label distri-bution learning[C]// Proceedings of the 27th International Joint Conference on Artificial Intelligence, Stockholm, Jul 13-19, 2018: 2926-2932.
[27]	SHAO R F, XU N, GENG X. Multi-label learning with label enhancement[C]// Proceedings of the 2018 IEEE Interna-tional Conference on Data Mining, Singapore, Nov 17-20, 2018. Piscataway: IEEE, 2018: 437-446.
[28]	KLIR G J, YUAN B. Fuzzy sets and fuzzy logic-theory and applications[M]. Upper Saddle River: Prentice Hall, 1995.
[29]	BUDA M, MAKI A, MAZUROWSKI M A. A systematic study of the class imbalance problem in convolutional neural networks[J]. Neural Networks, 2017, 106:249-259. DOI URL
[30]	GENG X, HOU P. Pre-release prediction of crowd opinion on movies by label distribution learning[C]// Proceedings of the 24th International Joint Conference on Artificial Intelli-gence, Buenos Aires, Jul 25-31, 2015. Menlo Park: AAAI, 2015: 3511-3517.
[31]	PÉREZ-CRUZ F, ALARCÓN-DIANA P L, NAVIA-VÁZQUEZ Á, et al. Fast training of support vector classi-fiers[C]// Proceedings of the Advances in Neural Informa-tion Processing Systems 13, Denver, Nov 27-Dec 2, 2000. Red Hook: Curran Associates, 2000: 734-740.
[32]	SCHLKOPF B, SMOLA A J. Learning with kernels[M]. Cambridge: MIT Press, 2001.
[33]	OLANOW C W, KOLLER W C. An algorithm (decision tree) for the management of Parkinson’s disease: treatment guidelines[J]. Neurology, 1998, 50(3):S1.
[34]	LIN X T, CHEN X W. Mr.KNN: soft relevance for multi-label classification[C]// Proceedings of the 19th ACM Con-ference on Information and Knowledge Management, Tor-onto, Oct 26-30, 2010. New York: ACM, 2010: 349-358.
[35]	THAI-NGHE N, GANTNER Z, SCHMIDT-THIEME L. Cost-sensitive learning methods for imbalanced data[C]// Proceedings of the 2010 International Joint Conference on Neural Networks, Barcelona, Jul 18-23, 2010. Piscataway: IEEE, 2010: 1-8.
[36]	CASTRO C L, BRAGA A P. Novel cost-sensitive approach to improve the multilayer perceptron performance on im-balanced data[J]. IEEE Transactions on Neural Networks and Learning Systems, 2013, 24(6):888-899. DOI URL

Label Distribution Learning for Computer Aided Diagnosis of Multi-class ASD Classification

面向多分类自闭症辅助诊断的标记分布学习

RichHTML

PDF

Knowledge

Abstract

Cite this article

share this article

Figures/Tables 9

References 36

Related Articles 2

Recommended Articles

Metrics

[1]	SANG Qingbing, GAO Shuang. No-Reference Image Quality Assessment via Quaternion Wavelet Transform [J]. Journal of Frontiers of Computer Science and Technology, 2017, 11(4): 633-642.
[2]	GAO Xinyu, YANG Zhixia, LU Zhenxing. Nonparallel ε-Band Support Vector Regression [J]. Journal of Frontiers of Computer Science and Technology, 2013, 7(7): 649-658.