A dysmorphic feature is an abnormal difference in body structure. It can be an isolated finding in an otherwise normal individual, or it can be related to a congenital disorder, genetic syndrome or birth defect. Dysmorphology is the study of dysmorphic features, their origins and proper nomenclature. One of the key challenges in identifying and describing dysmorphic features is the use and understanding of specific terms between different individuals. Clinical geneticists and pediatricians are usually those most closely involved with the identification and description of dysmorphic features, as most are apparent during childhood.
Dysmorphic features can vary from isolated, mild anomalies such as clinodactyly or synophrys to severe congenital anomalies, such as heart defects and holoprosencephaly. In some cases, dysmorphic features are part of a larger clinical picture, sometimes known as a sequence, syndrome or association. Recognizing the patterns of dysmorphic features is an important part of a geneticist's diagnostic process, as many genetic disease present with a common collection of features. There are several commercially available databases that allow clinicians to input their observed features in a patient to generate a differential diagnosis. These databases are not infallible, as they require on the clinician to provide their own experience, particularly when the observed clinical features are general. A male child with short stature and hypertelorism could have several different disorders, as these findings are not highly specific. However a finding such as 2,3-toe syndactyly raises the index of suspicion for Smith-Lemli-Opitz Syndrome.
Most open source projects that perform phenotype-driven disease or gene prioritization work with the terminology of the Human Phenotype Ontology. This controlled vocabulary can be used to describe the clinical features of a patient and is suitable for machine learning approaches. Publicly accessible databases that labs use to deposit their diagnostic findings, such as ClinVar, can be used to build knowledge graphs to explore the clinical feature space.
Dysmorphic features are invariably present from birth, although some are not immediately apparent upon visual inspection. They can be divided into groups based on their origin, including malformations (abnormal development), disruptions (damage to previously normal tissue), deformations (damage caused by an outside physical force) and dysplasias (abnormal growth or organization within a tissue).
Dysmorphology is the discipline of using dysmorphic features in the diagnostic workup and delineation of syndromic disorders. In the recent years advances in computer vision have also resulted in several deep learning approaches that assist geneticists in the study of the facial gestalt. Since deep learning requires big data, but most dysmorphic features are rare, transfer learning is crucial for achieving high performance. Training data for medical imaging such as portraits, x-rays and funduscopies can found at e.g. Kaggle and GestaltMatcher, which is a community driven effort to annotate medical images from the scientific literature for machine learning.
- Reardon, W.; Donnai, D. (2007). "Dysmorphology demystified". Archives of Disease in Childhood: Fetal and Neonatal Edition. 92 (3): F225–F229. doi:10.1136/adc.2006.110619. PMC 2675338. PMID 17449858.
- Maitra, Anirban; Kumar, Vinay (2004). "Diseases of Infancy and Childhood". In Kumar, Vinay; Abbas, Abul L.; Fausto, Nelson (eds.). Robbins and Coltran Pathologic Basis of Disease (7th ed.). Philadelphia: Elsevier. pp. 469–508. ISBN 978-0-7216-0187-8.
- j.-p., F.; De Ravel, T. D. (2002). "London Dysmorphology Database, London Neurogenetics Database and Dysmorphology Photo Library on CD-ROM \Version 3] 2001". Human Genetics. 111 (1): 113. doi:10.1007/s00439-002-0759-6. PMID 12136245. S2CID 20083700.
- Nowaczyk, M. J.; Waye, J. S. (2001). "The Smith-Lemli-Opitz syndrome: A novel metabolic way of understanding developmental biology, embryogenesis, and dysmorphology". Clinical Genetics. 59 (6): 375–386. doi:10.1034/j.1399-0004.2001.590601.x. PMID 11453964. S2CID 9146017.
- Peng, Chengyao; Dieck, Simon; Schmid, Alexander; Ahmad, Ashar; Knaus, Alexej; Wenzel, Maren; Mehnert, Laura; Zirn, Birgit; Haack, Tobias; Ossowski, Stephan; Wagner, Matias; Brunet, Teresa; Ehmke, Nadja; Danyel, Magdalena; Rosnev, Stanislav; Kamphans, Tom; Nadav, Guy; Fleischer, Nicole; Fröhlich, Holger; Krawitz, Peter (2021). "CADA: Phenotype-driven gene prioritization based on a case-enriched knowledge graph". medRxiv. doi:10.1101/2021.03.01.21251705. S2CID 232085750.
- Ferry, Quentin; Steinberg, Julia; Webber, Caleb; FitzPatrick, David R; Ponting, Chris P; Zisserman, Andrew; Nellåker, Christoffer (2014-06-24). Tollman, Stephen (ed.). "Diagnostically relevant facial gestalt information from ordinary photos". eLife. 3: e02020. doi:10.7554/eLife.02020. ISSN 2050-084X. PMC 4067075. PMID 24963138.
- Gurovich, Yaron; Hanani, Yair; Bar, Omri; Nadav, Guy; Fleischer, Nicole; Gelbman, Dekel; Basel-Salmon, Lina; Krawitz, Peter M.; Kamphausen, Susanne B.; Zenker, Martin; Bird, Lynne M. (January 2019). "Identifying facial phenotypes of genetic disorders using deep learning". Nature Medicine. 25 (1): 60–64. doi:10.1038/s41591-018-0279-0. ISSN 1546-170X. PMID 30617323. S2CID 57574514.
- Hsieh, Tzung-Chien; Bar-Haim, Aviram; Moosa, Shahida; Ehmke, Nadja; Gripp, Karen W.; Pantel, Jean Tori; Danyel, Magdalena; Mensah, Martin Atta; Horn, Denise; Fleischer, Nicole; Bonini, Guilherme (2021-01-04). "GestaltMatcher: Overcoming the limits of rare disease matching using facial phenotypic descriptors". medRxiv: 2020.12.28.20248193. doi:10.1101/2020.12.28.20248193. S2CID 230350187.