Few-Shot Meta-Learning for Recognizing Facial Phenotypes of Genetic Disorders Ömer SÜMERa1 Fabio HELLMANNa Alexander HUSTINXb

2025-05-06 0 0 156.17KB 5 页 10玖币

侵权投诉

Few-Shot Meta-Learning for Recognizing

Facial Phenotypes of Genetic Disorders

Ömer SÜMER a,1, Fabio HELLMANN a, Alexander HUSTINX b,

Tzung-Chien HSIEH b, Elisabeth ANDRÉ a, and Peter KRAWITZ b

aChair for Human-Centered Artiﬁcial Intelligence, University of Augsburg

bInstitute for Genomic Statistics and Bioinformatics, University of Bonn

Abstract. Computer vision-based methods have valuable use cases in precision

medicine, and recognizing facial phenotypes of genetic disorders is one of them.

Many genetic disorders are known to affect faces’ visual appearance and geometry.

Automated classiﬁcation and similarity retrieval aid physicians in decision-making

to diagnose possible genetic conditions as early as possible. Previous work has ad-

dressed the problem as a classiﬁcation problem and used deep learning methods.

The challenging issue in practice is the sparse label distribution and huge class

imbalances across categories. Furthermore, most disorders have few labeled sam-

ples in training sets, making representation learning and generalization essential to

acquiring a reliable feature descriptor. In this study, we used a facial recognition

model trained on a large corpus of healthy individuals as a pre-task and transferred

it to facial phenotype recognition. Furthermore, we created simple baselines of few-

shot meta-learning methods to improve our base feature descriptor. Our quantita-

tive results on GestaltMatcher Database show that our CNN baseline surpasses pre-

vious works, including GestaltMatcher, and few-shot meta-learning strategies im-

prove retrieval performance in frequent and rare classes.

Keywords. Facial genetics, rare genetic disorders, image analysis, few-shot

learning, meta-learning, imbalanced data, deep learning

1. Introduction

Genetic disorders affect more than 5% of the population [1]; in practice, physicians might

fail to spot and clinically diagnose most of them. There is a set of genetic conditions

and 30-40% of them are known to affect craniofacial development and facial morphol-

ogy [2]. The alterations in the face and skull can be recognized by using computer vision.

The output of computer vision-based systems can support physicians in diagnosing rare

syndromes and eventually lead to therapeutic interventions.

The number of samples in real-life situations and databases shows considerable vari-

ation across disorders. This makes training deep convolutional networks not feasible, as

in any object classiﬁcation task. The nature of the problem necessitates addressing data

imbalance and few-shot classiﬁcation in facial phenotype analysis.

This paper presents an approach to improve the baseline for unseen facial genetic

disorders based on a highly imbalanced distribution of disorders.

1Corresponding Author: Ömer Sümer, oemer.suemer@informatik.uni-augsburg.de

arXiv:2210.12705v2 [cs.CV] 24 May 2023

文档加载中……请稍候！
如果长时间未打开，您也可以点击刷新试试。

下载文档到电脑，查找使用更方便

10 玖币 0人已下载

立即下载

摘要：

Few-ShotMeta-LearningforRecognizingFacialPhenotypesofGeneticDisordersÖmerSÜMERa,1,FabioHELLMANNa,AlexanderHUSTINXb,Tzung-ChienHSIEHb,ElisabethANDRÉa,andPeterKRAWITZbaChairforHuman-CenteredArtificialIntelligence,UniversityofAugsburgbInstituteforGenomicStatisticsandBioinformatics,UniversityofBonnAbstr...

展开>> 收起<<

Few-Shot Meta-Learning for Recognizing Facial Phenotypes of Genetic Disorders Ömer SÜMERa1 Fabio HELLMANNa Alexander HUSTINXb.pdf

共5页,预览1页

还剩页未读，继续阅读

声明：本站为文档C2C交易模式，即用户上传的文档直接被用户下载，本站只是中间服务平台，本站所有文档下载所得的收益归上传人(含作者)所有。玖贝云文库仅提供信息存储空间，仅对用户上传内容的表现方式做保护处理，对上载内容本身不做任何修改或编辑。若文档所含内容侵犯了您的版权或隐私，请立即通知玖贝云文库，我们立即给予删除！

Few-Shot Meta-Learning for Recognizing Facial Phenotypes of Genetic Disorders Ömer SÜMERa1 Fabio HELLMANNa Alexander HUSTINXb

相关推荐

开通VIP享超值会员特权

作者详情

相关内容

热门标签

举报选择: