A Deep Learning Based Automatic Severity Detector for Diabetic Retinopathy
المؤلف | AlSaad R. |
المؤلف | Al-Maadeed S. |
المؤلف | Al Mamun M.A. |
المؤلف | Boughorbel S. |
تاريخ الإتاحة | 2020-03-29T12:15:10Z |
تاريخ النشر | 2018 |
اسم المنشور | Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) |
المصدر | Scopus |
الرقم المعياري الدولي للكتاب | 3029743 |
الملخص | Automated Diabetic Retinopathy (DR) screening methods with high accuracy have the strong potential to assist doctors in evaluating more patients and quickly routing those who need help to a specialist. In this work, we used Deep Convolutional Neural Network architecture to diagnosing DR from digital fundus images and accurately classifying its severity. We train this network using a graphics processor unit (GPU) on the publicly available Kaggle dataset. We used Theano, Lasagne, and cuDNN libraries on two Amazon EC2 p2.xlarge instances and demonstrated impressive results, particularly for a high-level classification task. On the dataset of 30,262 training images and 4864 testing images, our model achieves an accuracy of 72%. Our experimental results showed that increasing the batch size does not necessarily speed up the convergence of the gradient computations. Also, it demonstrated that the number and size of fully connected layers do not have a significant impact on the performance of the model. |
راعي المشروع | This work was supported by Sidra Medicine (authors RA and SB), as well as a grant from the Qatar National Research Fund through National Priority Research Program (NPRP) No. 6-249-1-053 (authors SA and MA). |
اللغة | en |
الناشر | Springer Verlag |
الموضوع | Convolutional Neural Networks Deep learning Diabetic retinopathy Medical imaging |
النوع | Conference Paper |
الصفحات | 64-76 |
رقم المجلد | 10934 LNAI |
الملفات في هذه التسجيلة
الملفات | الحجم | الصيغة | العرض |
---|---|---|---|
لا توجد ملفات لها صلة بهذه التسجيلة. |
هذه التسجيلة تظهر في المجموعات التالية
-
علوم وهندسة الحاسب [2342 items ]