Ensemble of Two-Path Capsule Networks for Diagnosis of Turner Syndrome Using Global-Local Facial Images

doi:10.1007/s12204-022-2491-9

Abstract

Abstract: Turner syndrome (TS) is a chromosomal disorder disease that only affects the growth of female patients. Prompt diagnosis is of high significance for the patients. However, clinical screening methods are time-consuming and cost-expensive. Some researchers used machine learning-based methods to detect TS, the performance of which needed to be improved. Therefore, we propose an ensemble method of two-path capsule networks (CapsNets) for detecting TS based on global-local facial images. Specifically, the TS facial images are preprocessed and segmented into eight local parts under the direction of physicians; then, nine two-path CapsNets are respectively trained using the complete TS facial images and eight local images, in which the few-shot learning is utilized to solve the problem of limited data; finally, a probability-based ensemble method is exploited to combine nine classifiers for the classification of TS. By studying base classifiers, we find two meaningful facial areas are more related to TS patients, i.e., the parts of eyes and nose. The results demonstrate that the proposed model is effective for the TS classification task, which achieves the highest accuracy of 0.924 1.

Key words: Turner syndrome (TS), two-path capsule network (CapsNet), ensemble method, few-shot learning

摘要： 特纳综合征是一种染色体异常疾病，对女性病人的成长造成极大的危害。及时诊断对该病患者具有重要意义。然而，现有的临床筛查方法相对耗时且费用昂贵。相关研究人员提出使用机器学习方法进行特纳综合征诊断，但是这些方法的诊断准确率有待提升。因此，面向特纳综合征诊断任务，提出一种基于全局-局部人脸图像的双路径胶囊网络集成方法。具体地，对特纳综合征人脸图像进行预处理，并在医疗专家的指导下，将人脸图像分割为8部分具有医学意义的局部人脸图像；然后，基于完整人脸图像和8部分局部图像进行双路径胶囊网络模型训练，以小样本学习方法解决模型训练过程中面临的样本不足问题；最后，以基于概率的集成方法对9个特纳综合征人脸分类模型进行集成。通过对基础分类模型进行分析，发现眼部区域和鼻子区域的异常面容与特纳综合征疾病具有强相关性。实验结果显示，该集成方法对特纳综合征诊断任务具有一定的有效性，能够取得0.9241的最高诊断准确率。

关键词: 特纳综合征，双路径胶囊网络，集成方法，小样本学习

CLC Number:

LIU Lu (刘璐). Ensemble of Two-Path Capsule Networks for Diagnosis of Turner Syndrome Using Global-Local Facial Images[J]. J Shanghai Jiaotong Univ Sci, 2023, 28(4): 459-.

References

[1] TURNER H H. A syndrome of infantilism, congenitalwebbed neck, and cubitus valgus [J]. Endocrinology,1938, 23(5): 566-574.
[2] GRAVHOLT C H, ANDERSEN N H, CONWAY G S,et al. Clinical practice guidelines for the care of girlsand women with Turner syndrome: Proceedings fromthe 2016 Cincinnati International Turner SyndromeMeeting [J]. European Journal of Endocrinology, 2017,177(3): G1-G70.
[3] MARINO B, FINE K. Blueprints in Pediatrics [M].6th ed. Philadelphia: Lippincott Williams & Wilkins,2013.
[4] LIPPE B. Turner syndrome [J]. Endocrinology andMetabolism Clinics of North America, 1991, 20(1):121-152.
[5] JUNG M P, AMARAL J L, FONTES R G, et al. Diagnosis of Turner’s Syndrome: The experience of the Riode Janeiro State Institute of Diabetes and Endocrinology between 1970 and 2008 [J]. Revista Brasileira DeSa′ude Materno Infantil, 2010, 10(1): 117-124 (in Portuguese).
[6] BONDY C A. New issues in the diagnosis and management of Turner syndrome [J]. Reviews in Endocrine& Metabolic Disorders, 2005, 6(4): 269-280.
[7] TABATABAEI S M, CHALECHALE A. Using DLBPtexture descriptors and SVM for Down syndromerecognition [C]//2014 4th International Conferenceon Computer and Knowledge Engineering. Mashhad:IEEE, 2014: 554-558.
[8] LEITE FILHO H P. Applicability of data mining technique using bayesians network in diagnosis of geneticdiseases [J]. International Journal of Advanced Computer Science and Applications, 2013, 4(1): 47-50.
[9] SONG W A, LEI Y, CHEN S, et al. Multiple facial image features-based recognition for the automatic diagnosis of Turner syndrome [J]. Computers in Industry,2018, 100: 85-95.
[10] YAO G H, LI J Q, PEI Y, et al. An automaticTurner syndrome identification system with facial images [M]//Frontier computing. Singapore: Springer,2020: 105-112.
[11] LIU L, SUN J C, LI J Q, et al. Automatic classifi-cation of Turner syndrome using unsupervised featurelearning [C]//2020 IEEE International Conference onSystems, Man, and Cybernetics. Toronto, ON: IEEE,2020: 1578-1583.
[12] LI J Q, LIU L, SUN J C, et al. Diagnosis and knowledge discovery of Turner syndrome based on facial images using machine learning methods [J]. IEEE Access,2020, 8: 214866-214881.
[13] SABOUR S, FROSST N, HINTON G E. Dynamicrouting between capsules [C]//31st Conference onNeural Information Processing Systems. Long Beach,CA: NIPS, 2017: 1-11.
[14] WONG K C L, SYEDA-MAHMOOD T, MORADI M.Building medical image classifiers with very limiteddata using segmentation networks [J]. Medical ImageAnalysis, 2018, 49: 105-116.
[15] KIM H G, CHOI Y, RO Y M. Modality-bridge transfer learning for medical image classification [C]//201710th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics.Shanghai: IEEE, 2017: 1-5.
[16] YIN S L, BI J. Medical image annotation based ondeep transfer learning [C]//2018 IEEE InternationalConference on Internet of Things (iThings) and IEEEGreen Computing and Communications (GreenCom)and IEEE Cyber, Physical and Social Computing (CPSCom) and IEEE Smart Data. Halifax, NS: IEEE,2018: 47-49.
[17] CAI A H, HU W X, ZHENG J. Few-shot learning formedical image classification [M]//Artificial neural networks and machine learning – ICANN 2020. Cham:Springer, 2020: 441-452.
[18] KIM M, ZUALLAERT J, DE NEVE W. Few-shotlearning using a small-sized dataset of high-resolutionFUNDUS images for glaucoma diagnosis [C]//2ndInternational Workshop on Multimedia for PersonalHealth and Health Care. Mountain View, CA: ACM,2017: 89-92.
[19] WEN G H, HOU Z, LI H H, et al. Ensemble of deepneural networks with probability-based fusion for facial expression recognition [J]. Cognitive Computation,2017, 9(5): 597-610.
[20] FINN C, ABBEEL P, LEVINE S. Model-agnosticmeta-learning for fast adaptation of deep networks[C]//34th International Conference on Machine Learning. Sydney: PMLR, 2017: 1126-1135.
[21] VINYALS O, BLUNDELL C, LILLICRAP T P, etal. Matching networks for one shot learning [C]//30thConference on Neural Information Processing Systems.Barcelona: NIPS, 2016: 3630-3638.