AE CNN Save

Joint Learning Mechanism to Compress and Classify Radiological Images

Source code for ICVGIP 2018 paper: Jointly Learning Convolutional Representations to Compress Radiological Images and Classify Thoracic Diseases in the Compressed Domain

Overview of AE-CNN: Our proposed framework consists of three main blocks namely encoder, decoder, and classifier. The figure shows the autoencoder based convolutional neural network (AE-CNN) model for disease classification. Here, autoencoder reduces the spatial dimension of the imput image of size 1024 × 1024. The encoder produces a latent code tensor of size 224 × 224 and decoder reconstructs back the image. This latent code tensor is passed through a CNN classifier for classifying the chest x-rays. The final loss is the weighted sum of the resconstruction loss by decoder and classification loss by the CNN classifier.

Results

Citation

Please cite the following paper if you found it useful in your work:

@inproceedings{10.1145/3293353.3293408,
author = {Ranjan, Ekagra and Paul, Soumava and Kapoor, Siddharth and Kar, Aupendu and Sethuraman, Ramanathan and Sheet, Debdoot},
title = {Jointly Learning Convolutional Representations to Compress Radiological Images and Classify Thoracic Diseases in the Compressed Domain},
year = {2018},
isbn = {9781450366151},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
url = {https://doi.org/10.1145/3293353.3293408},
doi = {10.1145/3293353.3293408},
booktitle = {Proceedings of the 11th Indian Conference on Computer Vision, Graphics and Image Processing},
articleno = {55},
numpages = {8},
keywords = {compression, X-Ray classification, Convolutional autoencoder},
location = {Hyderabad, India},
series = {ICVGIP 2018}
}

Acknowledgement:

We would like to thank zoozog and arnoweng for open-sourcing their repos which served as the starting point for our work. Their repos can be found here and here respectively.