Identifying congenital generalized lipodystrophy using deep learning-DEEPLIPO.

Area of Science:

• Medical imaging
• Artificial intelligence in medicine
• Genetics and rare diseases

Background:

• Congenital Generalized Lipodystrophy (CGL) is a rare genetic disorder characterized by a near-complete absence of adipose tissue.
• CGL diagnosis relies on clinical features like acromegaloid features and acanthosis nigricans.
• Early identification of CGL is critical due to severe, premature cardiometabolic complications and poor prognosis.

Purpose of the Study:

• To develop and evaluate a deep learning model for identifying the Congenital Generalized Lipodystrophy (CGL) phenotype.
• To assess the feasibility of using image processing with deep learning for routine clinical detection of CGL.

Main Methods:

• A convolutional neural network (CNN) deep learning model was trained and tested on a database of 337 patient photographs.
• The dataset included images of CGL patients, individuals with malnutrition, and eutrophic individuals with athletic builds.
• Fourfold cross-validation was employed, with 75% of data used for training and 25% for testing in each iteration.

Main Results:

• The deep learning model achieved a mean accuracy of 90.85% ± 2.20%.
• Mean sensitivity was 90.63% ± 3.53%, and mean specificity was 91.41% ± 1.10%.
• These results demonstrate high performance in distinguishing CGL phenotype from other conditions.

Conclusions:

• This study presents the first deep learning model capable of identifying the Congenital Generalized Lipodystrophy phenotype.
• The model demonstrates excellent accuracy, sensitivity, and specificity.
• This AI tool holds potential as a strategic method for early CGL detection and management.