eduzhai > Applied Sciences > Engineering >

Development and Validation of a Novel Prognostic Model for Predicting AMD Progression Using Longitudinal Fundus Images

  • king
  • (0) Download
  • 20210505
  • Save

... pages left unread,continue reading

Document pages: 10 pages

Abstract: Prognostic models aim to predict the future course of a disease or conditionand are a vital component of personalized medicine. Statistical models make useof longitudinal data to capture the temporal aspect of disease progression;however, these models require prior feature extraction. Deep learning avoidsexplicit feature extraction, meaning we can develop models for images wherefeatures are either unknown or impossible to quantify accurately. Previousprognostic models using deep learning with imaging data require annotationduring training or only utilize a single time point. We propose a novel deeplearning method to predict the progression of diseases using longitudinalimaging data with uneven time intervals, which requires no prior featureextraction. Given previous images from a patient, our method aims to predictwhether the patient will progress onto the next stage of the disease. Theproposed method uses InceptionV3 to produce feature vectors for each image. Inorder to account for uneven intervals, a novel interval scaling is proposed.Finally, a Recurrent Neural Network is used to prognosticate the disease. Wedemonstrate our method on a longitudinal dataset of color fundus images from4903 eyes with age-related macular degeneration (AMD), taken from theAge-Related Eye Disease Study, to predict progression to late AMD. Our methodattains a testing sensitivity of 0.878, a specificity of 0.887, and an areaunder the receiver operating characteristic of 0.950. We compare our method toprevious methods, displaying superior performance in our model. Classactivation maps display how the network reaches the final decision.

Please select stars to rate!

         

0 comments Sign in to leave a comment.

    Data loading, please wait...
×