Objectives Ultrasound (US) is the first-choice device to detect different types of facial dysmorphisms. Anyway, at present no standard protocol has been defined for automatic nor semi-automatic diagnosis. Even though the practitioner's contribution is core, steps towards automatism are to be undertaken. We propose a methodology for diagnosing cleft lip on 3D US scans. Methods A bounded Depth Minimum Steiner Trees (D-MST) clustering algorithm is proposed for discriminating groups of 3D US faces relying on the presence/absence of a cleft lip. The analysis of 3D facial surfaces via Differential Geometry is adopted to extract landmarks. Thus, the extracted geometrical information is elaborated to feed the unsupervised clustering algorithm and produce the classification. The clustering returns the probability of being affected by the pathology, allowing physicians to focus their attention on risky individuals for further analysis. Results The feasibility is tested upon the available 3D US scans data and then deeply investigated for a large dataset of adult individuals. 3D facial Bosphorus database is chosen for the testing, which seven cleft lip-affected individuals are added to, by artificially creating the defect. The algorithm correctly separates left and right-sided cleft lips, while healthy individuals create a unique cluster; thus, the method shows accurate diagnosis results. Conclusions Even if further testing is to be performed on tailored datasets made exclusively of fetal images, this techniques gives hefty hints for a future tailored algorithm. This method also fosters the investigation of the scientific formalisation of the "normotype", which is the representative face of a class of individuals, collecting all the principal anthropometric facial measurements, in order to recognise a normal or syndromic fetus.

Antenatal automatic diagnosis of cleft lip via unsupervised clustering method relying on 3D facial soft tissue landmarks / Padula, F.; Vezzetti, Enrico; Marcolin, Federica; Conti, Daniele; Bonacina, L.; Froio, Antonio; Giorlandino, M.; Coco, C.; D'Emidio, L.; Giorlandino, C.; Speranza, D.. - In: ULTRASOUND IN OBSTETRICS & GYNECOLOGY. - ISSN 1469-0705. - ELETTRONICO. - 48:S1(2016), pp. 127-128. [10.1002/uog.16383]

Antenatal automatic diagnosis of cleft lip via unsupervised clustering method relying on 3D facial soft tissue landmarks

VEZZETTI, ENRICO;MARCOLIN, FEDERICA;CONTI, DANIELE;FROIO, ANTONIO;
2016

Abstract

Objectives Ultrasound (US) is the first-choice device to detect different types of facial dysmorphisms. Anyway, at present no standard protocol has been defined for automatic nor semi-automatic diagnosis. Even though the practitioner's contribution is core, steps towards automatism are to be undertaken. We propose a methodology for diagnosing cleft lip on 3D US scans. Methods A bounded Depth Minimum Steiner Trees (D-MST) clustering algorithm is proposed for discriminating groups of 3D US faces relying on the presence/absence of a cleft lip. The analysis of 3D facial surfaces via Differential Geometry is adopted to extract landmarks. Thus, the extracted geometrical information is elaborated to feed the unsupervised clustering algorithm and produce the classification. The clustering returns the probability of being affected by the pathology, allowing physicians to focus their attention on risky individuals for further analysis. Results The feasibility is tested upon the available 3D US scans data and then deeply investigated for a large dataset of adult individuals. 3D facial Bosphorus database is chosen for the testing, which seven cleft lip-affected individuals are added to, by artificially creating the defect. The algorithm correctly separates left and right-sided cleft lips, while healthy individuals create a unique cluster; thus, the method shows accurate diagnosis results. Conclusions Even if further testing is to be performed on tailored datasets made exclusively of fetal images, this techniques gives hefty hints for a future tailored algorithm. This method also fosters the investigation of the scientific formalisation of the "normotype", which is the representative face of a class of individuals, collecting all the principal anthropometric facial measurements, in order to recognise a normal or syndromic fetus.
2016
File in questo prodotto:
File Dimensione Formato  
article_10.1002_uog.16383.pdf

accesso aperto

Descrizione: Extended abstract
Tipologia: Abstract
Licenza: PUBBLICO - Tutti i diritti riservati
Dimensione 66.89 kB
Formato Adobe PDF
66.89 kB Adobe PDF Visualizza/Apri
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11583/2657497
 Attenzione

Attenzione! I dati visualizzati non sono stati sottoposti a validazione da parte dell'ateneo