Author(s) :

Daud Tariq, Juliana Dellorco, Rayyan Shakkel.

Volume/Issue :

Abstract :

Ehlers-Danlos syndrome (EDS) is a primarily genetic disorder, typically resulting from mutations in collagen-encoding genes. There are thirteen varieties of EDS, each presenting different symptoms and characteristics. Common symptoms include skin elasticity, hypermobility, abnormal scar formation, and bruising. EDS is a complex syndrome and lacks a definite diagnosis method, leading to frequent misdiagnosis, delayed treatment or management, and frequent feelings of resentment among patients. While there is currently no definite diagnosis system, artificial intelligence (AI) is being evaluated to aid in diagnosis, using methods such as AI-based video goniometry; Uniform Manifold Approximation and Projection (UMAP), a dimensionality reduction technique that simplifies complex data while preserving patterns; and Hierarchical Density-Based Spatial Clustering (HDBSCAN), an algorithm that utilizes clustering, grouping similar data points together without requiring predefined categories. These methods were then assessed on their feasibility, including key strengths and weaknesses such as availability and accuracy. Overall, AI interventions in diagnostics are promising innovations that can act as potent preliminary screening tools. However, the data sets that these models are trained on often lead to bias and a lack of generalizability, causing unreliability in the readings. This is further compounded by the fact that these models are “black boxes”, meaning that clinicians cannot access the underlying processing route the machine uses to make a diagnosis. Therefore, while these models show promise, more clinical studies are needed to prove their feasibility in the clinical landscape.