Neural Networks for Classification and Image Generation of Aging in Genetic Syndromes

Abstract

Background: In medical genetics, one application of neural networks is the diagnosis of genetic diseases based on images of patient faces. While these applications have been validated in the literature with primarily pediatric subjects, it is not known whether these applications can accurately diagnose patients across a lifespan. We aimed to extend previous works to determine whether age plays a factor in facial diagnosis as well as to explore other factors that may contribute to the overall diagnostic accuracy.Methods: To investigate this, we chose two relatively common conditions, Williams syndrome and 22q11.2 deletion syndrome. We built a neural network classifier trained on images of affected and unaffected individuals of different ages and compared classifier accuracy to clinical geneticists. We analyzed the results of saliency maps and the use of generative adversarial networks to boost accuracy.Results: Our classifier outperformed clinical geneticists at recognizing face images of these two conditions within each of the age groups (the performance varied between the age groups): 1) under 2 years old, 2) 2–9 years old, 3) 10–19 years old, 4) 20–34 years old, and 5) ≥35 years old. The overall accuracy improvement by our classifier over the clinical geneticists was 15.5 and 22.7% for Williams syndrome and 22q11.2 deletion syndrome, respectively. Additionally, comparison of saliency maps revealed that key facial features learned by the neural network differed with respect to age. Finally, joint training real images with multiple different types of fake images created by a generative adversarial network showed up to 3.25% accuracy gain in classification accuracy.Conclusion: The ability of clinical geneticists to diagnose these conditions is influenced by the age of the patient. Deep learning technologies such as our classifier can more accurately identify patients across the lifespan based on facial features. Saliency maps of computer vision reveal that the syndromic facial feature attributes change with the age of the patient. Modest improvements in the classifier accuracy were observed when joint training was carried out with both real and fake images. Our findings highlight the need for a greater focus on age as a confounder in facial diagnosis.

Privacy, bias and the clinical use of facial recognition technology: A survey of genetics professionals

“…Machine learning approaches to facial analysis have been seen as a means to improve diagnostic capabilities in understudied populations (Kruszka, Tekendo‐Ngongang, & Muenke, 2019). For example, Williams syndrome and 22q11.2 deletion syndrome are typically diagnosed in the pediatric age group, and a facial analysis neural network classifier built to help diagnose these conditions in older individuals outperformed clinical geneticists across five different age groups, with overall accuracy gains over clinical geneticists of 15.5 and 22.7%, respectively (Duong et al, 2022). Another facial analysis application for the autosomal dominant disorder Rubinstein‐Taybi syndrome (RSTS), however, showed poor discrimination efficacy in an African group while demonstrating excellent discrimination efficacy in a European one (Tekendo‐Ngongang et al, 2020).…”

Application of facial analysis Technology in Clinical Genetics: Considerations for diverse populations

“…One of the most common applications of artificial intelligence (AI) in clinical genetics is the use of facial analysis technology. A recent study showed that a neural network classifier outperformed the average clinical geneticists in recognizing two common genetic conditions (Duong et al, 2022). Like any AI platform, accuracy greatly depends on the data set used to train the algorithm.…”

“…The concept of a precise diagnosis is fueling an interest in testing newborns for conditions beyond the current limited newborn screening programs (Gold et al, 2023). Recently, Duong et al using a neural network classifier showed that their algorithm was able to accurately distinguish between Williams syndrome and 22q11.2 deletion syndrome across all ages, including infants (Duong et al, 2022). Research in the accuracy of facial analysis technology in newborns, especially in conditions with actionable findings, would be of value in resource limited settings where other screening modalities are not always available.…”

Perspectives on the future of dysmorphology

“…The last decade has witnessed an unprecedented increase in interest in applications of AI to medicine and biomedical sciences, and a surge in AI translational research initiatives in medical genetics and genomics (Duong et al, 2022; Poplin et al, 2018). Specifically, machine learning methods are being applied successfully for diagnostic processes involving pattern recognition, and the increasing use of computer‐assisted facial images processing for genetic syndromes recognition is a perfect illustration of how machine learning methods are being applied to dysmorphology (Hsieh et al, 2022; Kruszka et al, 2019).…”