Download to read offline
Uploaded byCHC Connecticut
Implementation of Facial Recognition Software for Clinical Genetics Practice & Education Slides - December 13, 2022
This document discusses the potential uses of facial recognition software in clinical genetics practice and education. It provides 3 examples of how facial recognition software could help in rare disease identification and interpreting genetic testing results. The document also outlines learning objectives about identifying medical uses of facial recognition, using facial grids to match patterns to syndromes, and the importance of diverse training data.
More Related Content
Similar to Implementation of Facial Recognition Software for Clinical Genetics Practice & Education Slides - December 13, 2022
More from CHC Connecticut
Implementation of Facial Recognition Software for Clinical Genetics Practice & Education Slides - December 13, 2022
- 1. Implementation of Facial RecognitionSoftware for Clinical Genetics Practice & Education Stephanie Sacharow, M.D. Division of Genetics and Genomics Boston Children’s Hospital December 13, 2022 12pm EST
- 2. Learning Objectives By theend of this session, attendees will be able to: • Objective 1: Identify uses of AI facial recognition programs including medical uses. • Objective 2: Understand how facial recognition software can be of value in assessing the degree of dysmorphology and utilizing a facial grid to match patterns to specific syndromes. • Objective 3: Note the importance of using people of multiple geographic and ethnic backgrounds to train the software and allow for broader use of these tools. 2
- 3. Discussion Questions • Q1:How can Medical Genetics implement syndrome identification tools into clinical practice? • Q2: What are potential educational applications of facial recognition software for syndrome identification? • Q3: Is there utility for a pediatric or other primary care provider to use AI to help in syndrome recognition or an objective assessment of the degree of dysmorphology?
- 4. Examples of usesof Facial Recognition Software • Studies on Emotion • Social Media • Airports • Law Enforcement • Government • Vanity phone applications
- 5. • Finding lostfamily members – Article describes its use in identifying Holocaust survivors in photos Examples of uses of Facial Recognition Software
- 6. Proposed medical uses •Autism – Autism App: Parents Diagnosing Children (healthline.com) The Autism and Beyond app uses an iPhone self-facing camera to assess child emotional state while viewing various stimuli. The dots are landmarks automatically placed on a video of the child by the software. Picture Credit: Autism & Beyond
- 7. Nature Medicine Identifying facialphenotypes of genetic disorders using deep learning
- 9. Multi-national study onDigeorge/Velocardiofacial Syndrome (22q11.2 deletion Syndrome) • Researchers with the National Institutes of Health’s National Human Genome Research Institute (NHGRI) successfully used facial recognition software to diagnose 22q11 deletion syndrome. • Human malformation syndromes appear different in different parts of the world • Doctors in 11 countries submitted 101 photos – The researchers then used facial analysis technology to compare a group of 156 Caucasians, Africans, Asians and Latin Americans comparing affected and controls • Based on 126 individual facial features, they correctly diagnosed in all ethnic groups 96.6 percent of the time American Journal of Medical Genetics
- 10. Patients with 22q11.2deletion syndrome syndrome (AMERICAN JOURNAL OF MEDICAL GENETICS/NATIONAL HUMAN GENOME RESEARCH INSTITUTE)
- 11. • Their datais now part of the NIH’s Atlas of Human Malformation Syndromes in Diverse Populations, a resource to help clinicians diagnose birth defects and genetic diseases. – www.genome.gov/atlas • Launched by NHGRI and its collaborators in September 2016 • Previously used for Down Syndrome, 2016 – “Down syndrome in diverse populations”
- 13. Syndrome recognition softwarein Guangdong Provincial People's Hospital
- 14. MGeneRx and Children’sNational • Children’s National Hospital has entered into a licensing agreement with MGeneRx Inc. – patented pediatric medical device technology – uses objective digital biometric analysis software – early and non-invasive screening of dysmorphic genetic diseases such as Noonan syndrome
- 18. FDNA Technology Deep Learningand Computer Vision • Deep learning algorithms build syndrome- specific computational-based classifiers (syndrome gestalts) • Technology converts a patient photo into de- identified mathematical facial descriptors • The patient’s facial descriptor is compared to syndrome gestalts to quantify similarity (gestalt scores) resulting in a prioritized list of syndromes with similar morphology • Artificial intelligence suggests likely phenotypic traits and genes to assist in feature annotation and syndrome prioritization
- 19. Phenotype Data • Crowd-sourcedby genetics professionals, labs & bioinformatics from patients • Millions of data points • Learning system uses data for unique health insights and genomics discoveries FDNA Technology
- 20. Implementation of Face2Gene •Boston Children’s Hospital was trial site for Face2Gene software – We trained our physicians (Medical Genetics residents/fellows and attendings) on Face2Gene – Ipads given to attendings and fellows and were available in clinic initially • Transitioned to using the mobile application – It learns from the cases we input – Few cycles of upgrades – Forums • Process involves consent and photography through Face2Gene app
- 22. Case Example 1 Rare-disorderidentification (atypical presentation in a female)
- 23. Case 1 • 1year old female with dysmorphic facial features – prominent supraorbital ridges, wide nasal bridge & broad nasal tip • Conductive hearing loss • Cleft palate • Fingers long, thin and with contractures
- 27. Case Example 2 InterpretingGenetic Testing Results
- 28. Case 2 • 10yofemale with profound developmental delay • Hypotonia • Macroglossia • Coarse facial features • Negative testing for lysosomal storage disease including MPS disorders (mucopolysaccharidoses)
- 30. TBCK Syndrome (“Infantile hypotoniawith psychomotor retardation and characteristic facies 3”) • Coarse facial features • Congenital hypotonia • Global developmental delay, ranging from moderate to severe • Dysphagia • Respiratory insufficiency due to weakness • Epilepsy • MRI features, including white matter changes, cerebellar atrophy and thin corpus callosum • Absent/severely delayed expressive language • Hyporeflexia/areflexia • Hypothyroidism • Osteopenia • Hypercholesterolemia • Frequent urinary tract infections/nephrolithiasis
- 34. Case Example 3 AtypicalPhenotype
- 35. Case 3 • 2year old male – Sprengel anomaly – Butterfly vertebrae – Wide-set eyes
- 37. Waardenburg Syndrome, type1 • 1/40,000 • Causes ~2-5% of congenital hearing loss, often profound • White forelock or early graying of the scalp hair before age 30 years • Heterochromia iridium, partial/segmental heterochromia, or hypoplastic or brilliant blue irides • Congenital leukoderma is frequently seen on the face, trunk, or limbs. • Gene: PAX3
- 38. Case 3 • Onfurther questioning, he has partial heterochromia
- 39. Dysmorphology and Diagnostic DilemmasConferences Educational Applications:
- 43. Familial 16q24.3 microdeletioninvolving ANKRD11 causes a KBG-like syndrome - PubMed (nih.gov)
- 47. Resources and References •Identifying facial phenotypes of genetic disorders using deep learning | Nature Medicine • Atlas of Human Malformation Syndromes in Diverse Populations – www.genome.gov/atlas • Genetic syndromes screening by facial recognition technology: VGG-16 screening model construction and evaluation | Orphanet Journal of Rare Diseases • Commercialization of novel facial analysis technology can improve diagnosis of rare disorders - Children's National (childrensnational.org) • FDNA – Face2Gene.org 47
Editor's Notes
- #5 The spatial distribution of eye movements predicts the (false) recognition of emotional facial expressions | PLOS ONE Facial recognition tech IDs people in archived Holocaust photos (freethink.com)
- #7 Research study through Duke
- #8 Identifying facial phenotypes of genetic disorders using deep learning | Nature Medicine
- #10 DiGeorge syndrome, a rare genetic disease, could be diagnosed with selfies - CBS News Facial recognition software helps diagnose rare genetic disease (genome.gov)
- #12 Down syndrome in diverse populations - Kruszka - 2017 - American Journal of Medical Genetics Part A - Wiley Online Library
- #14 Genetic syndromes screening by facial recognition technology: VGG-16 screening model construction and evaluation | Orphanet Journal of Rare Diseases | Full Text (biomedcentral.com) A total of 228 children with GSs and 228 healthy children were recruited from Guangdong Provincial People's Hospital from Jun 2016 to Jan 2021
- #15 Commercialization of novel facial analysis technology can improve diagnosis of rare disorders - Children's National (childrensnational.org) Children’s National Hospital has entered into a licensing agreement with MGeneRx Inc. for its patented pediatric medical device technology using objective digital biometric analysis software for the early and non-invasive screening of dysmorphic genetic diseases such as Noonan syndrome.
- #24 Filamin A: phenotypic diversity - ScienceDirect
- #27 Other X-linked syndromes presenting atypically in females picked up by Face2Gene includes Coffin Lowry and X-linked Cornelia de Lange
- #29 Mutations in TBCK, Encoding TBC1-Domain-Containing Kinase, Lead to a Recognizable Syndrome of Intellectual Disability and Hypotonia - ScienceDirect
- #36 Malformed vertebrae: a clinical and imaging review | Insights into Imaging | Full Text (springeropen.com) Telecanthus – Wikipedia Sprengel's Deformity of the Shoulder | Best Treatment in Bangalore (bangaloreshoulderinstitute.com)
- #38 Waardenburg syndrome type I (WS1) is an auditory-pigmentary disorder comprising congenital sensorineural hearing loss and pigmentary disturbances of the iris, hair, and skin along with dystopia canthorum (lateral displacement of the inner canthi). The hearing loss in WS1, observed in approximately 60% of affected individuals, is congenital, typically non-progressive, either unilateral or bilateral, and sensorineural. Most commonly, hearing loss in WS1 is bilateral and profound (>100 dB). The majority of individuals with WS1 have either a white forelock or early graying of the scalp hair before age 30 years. The classic white forelock observed in approximately 45% of individuals is the most common hair pigmentation anomaly seen in WS1. Affected individuals may have complete heterochromia iridium, partial/segmental heterochromia, or hypoplastic or brilliant blue irides. Congenital leukoderma is frequently seen on the face, trunk, or limbs. Diagnosis/testing. The diagnosis of WS1 is established in most individuals by physical examination for clinical criteria including: sensorineural hearing loss, pigmentary changes in the hair and eyes, dystopia canthorum identified by calculation of the W index, and specific facial features. Identification of a heterozygous PAX3 pathogenic variant by molecular genetic testing establishes the diagnosis if clinical features are inconclusive Waardenburg syndrome type 1 (escholarship.org) https://www.google.com/url?sa=i&url=https%3A%2F%2Fblue-sea-697d.quartiers047.workers.dev%3A443%2Fhttps%2Fwww.semanticscholar.org%2Fpaper%2FWaardenburg-syndrome%253A-a-rare-genetic-disorder-in-of-Korday-Bhaisara%2F3955bfb484892579f8152d397600e3c894552fac&psig=AOvVaw2awNDXqJ3WeAKOjTl8TxB6&ust=1670638262173000&source=images&cd=vfe&ved=0CAwQjRxqFwoTCKDM5tu66_sCFQAAAAAdAAAAABAJ