Paper: http://ceur-ws.org/Vol-2882/paper22.pdf Debapriya Banik and Debotosh Bhattacharjee : Deep Conditional Adversarial learning for polyp Segmentation. Proc. of MediaEval 2020, 14-15 December 2020, Online. This approach has addressed the Medico automatic polyp segmentation challenge which is a part of Mediaeval 2020. We have proposed a deep conditional adversarial learning based network for the automatic polyp segmentation task. The network comprises of two interdependent models namely a generator and a discriminator. The generator network is a FCN employed for the prediction of the polyp mask while the discriminator enforces the segmentation to be as similar as the real segmented mask (ground truth). Our proposed model achieved a comparative result on the test dataset provided by the organizers of the challenge.

1. Deep ConditionalAdversarial Learning For Polyp Segmentation
DebapriyaBanik,DebotoshBhattacharjee
DepartmentofComputerScienceandEngineering,JadavpurUniversity,Kolkata-32,India
Paper ID:22
Presented By:
Debapriya Banik
PhD Research Scholar
Dept. of CSE
Jadavpur University, Kolkata, India
MediaEval’20

2. Objective
The task for Medico automatic polyp segmentation challenge[1], a part of Multimedia
Evaluation 2020 (MediaEval 2020) is to develop computer-aided diagnosis systems for
automatic polyp segmentation to detect all types of polyp.
1
MediaEval’20

3. Related Work
• Polyp segmentation is a significant problem for the early diagnosis of Colorectal Cancer(CRC)[2].
• Generative Adversarial Networks (GANs) have shown a significant improvement in the field of medical image
analysis[3].
• Few GANs based approaches for polyp segmentation have been reported in the works of[4-6].
• However, GANs have been less explored in this domain in contrast to other medical imaging task.
• So, main motivation of this work is from the insights of GANs in the domain of medical image analysis (image
translation, super resolution, segmentation, classification, detection).
2
MediaEval’20

4. Contributions
• We have proposed a deep conditional adversarial learning-based network[7] for the automatic polyp
segmentation task.
• A weighted loss function strategy is introduced for the optimization of the network.
• Qualitative and Quantitative evaluation of the method to justify the effectiveness of our proposed approach.
3
MediaEval’20

5. ProposedApproach
• The proposed network is composed of two interdependent models, namely a
generator and a discriminator.
• The generator network is a Fully Convolutional Network (FCN) employed for the
prediction of the polyp mask.
• The discriminator enforces the segmentation to be as similar as the real segmented
mask (ground truth)
4
MediaEval’20

6. ProposedcGAN Architecture
5
MediaEval’20
Fig 1.Proposed framework for the segmentation of polyp

7. ProposedcGAN Architecture
Preprocessing
• The training challenge dataset [8] contains limited number of samples. So to enhance the
dataset, we have perform data augmentation
• Data augmentation is performed by applying 3 general transformation approaches:
Rotation (15◦ to 350◦),
Flipping (vertical and horizontal),
Translation (−5 to 30).
• The training samples are resized to a fixed size of 384 × 288 before they are fed into our
proposed network .
6
MediaEval’20

8. ProposedcGAN Architecture
Generator network
• The generator is an encoder-decoder network responsible to produce the binary
segmented generated mask
• The encoder part consists of 12 convolutional layers with different number of kernels.
• Each convolutional layer undergoes batch normalization and followed by ReLu activation
function.
• After a sequence of 3 convolutional layers, a MaxPooling layer is used.
• The decoder part consists of 12 deconvolutional layers with different strides and number
of kernels.
• Each deconvolutional layer also undergoes batch normalization and followed by ReLu
activation function.
• Skip connections are introduced between the layers in the encoder and the decoder part.
• The last layer is a convolutional layer with a kernel size of 1x1. 7
MediaEval’20

9. ProposedcGAN Architecture
Discriminator network
• The discriminator is a convolutional neural network (CNN).
• The input image and the binary segmented mask containing the polyp region which can be
either be the ground-truth mask or the predicated mask is fed into the discriminator
network.
• The networks has 10 convolutional layers with batch normalization and ReLu activation
function.
• A skip connection is applied between layers ,so that features learned are sustained.
• A sigmoid function is applied to the output which takes real values varying between 0 and
1, where 0 signifies fake, and 1 signifies real.
8
MediaEval’20

10. Training Strategy
• The training process of the proposed network alternates between 2 steps:
(1) The generator is trained by freezing the discriminator.
(2) The discriminator is trained while freezing the generator.
• The adversarial learning process of the network iteratively optimizes the parameters of
the generator and the discriminator through the loss function.
• To optimize the generator network, we have used a weighted loss function of Binary Cross
Entropy (BCE) and Mean Squared Error (MSE). Similarly, for the discriminator we have
used BCE as a loss function.
• We have trained each of the generator and the discriminator networks for 35 epochs. 9
MediaEval’20

11. ExperimentalResults
• Qualitative Evaluation
• Quantitative Evaluation
10
MediaEval’20
Fig 2. (a) Sample polyp images, and
(b) Corresponding segment masks
generated by the Proposed Approach

12. Conclusion
• In this study, we have proposed an approach for polyp segmentation by a deep
adversarial learning technique.
• Our experimental results on the test dataset justify the effectiveness of our proposed
technique.
• Our work’s future direction would be to implement an effective GAN and to evaluate
our method on datasets of other medical image modalities to justify our method’s
robustness.
MediaEval’20
11

13. Acknowledgements
The first author is grateful to the Council of Scientific and Industrial Research (CSIR) for
providing Senior Research Fellowship (SRF) under the SRF-Direct fellowship program
(ACK No. 143416/2K17/1, File No. 09/096(0922)2K18 EMR-I).
The authors are thankful for the Indo-Austrian joint project grant No.
INT/AUSTRIA/BMWF/P-25/2018 funded by the DST, GOI, and the SPARC project (ID:
231) funded by MHRD, GOI.
MediaEval’20

14. References
[1] Debesh Jha, Steven A. Hicks, Krister Emanuelsen, Håvard D. Johansen, Dag Johansen, Thomas de Lange, Michael A. Riegler, and
Pål Halvorsen. [n.d.]. Medico Multimedia Task at MediaEva 2020:Automatic Polyp Segmentation.
[2] Debapriya Banik, Kaushiki Roy, Debotosh Bhattacharjee, Mita Nasipuri, and Ondrej Krejcar. 2020. Polyp-Net: A Multi-model
Fusion Network for Polyp Segmentation. IEEE Transactions on Instrumentation and Measurement (2020).
[3] Salome Kazeminia, Christoph Baur, Arjan Kuijper, Bram van Ginneken, Nassir Navab, Shadi Albarqouni, and Anirban
Mukhopadhyay. 2020. GANs for medical image analysis. Artificial Intelligence in Medicine (2020), 101938.
[4] Konstantin Pogorelov, Olga Ostroukhova, Mattis Jeppsson, Håvard Espeland, Carsten Griwodz, Thomas de Lange, Dag Johansen,
Michael Riegler, and Pål Halvorsen. 2018. Deep learning and hand-crafted feature based approaches for polyp detection in
medical videos. In 2018 IEEE 31st International Symposium on Computer-Based Medical Systems (CBMS). IEEE, 381–386.
[5] JM Poomeshwaran, Kumar S Santhosh, Keerthi Ram, Jayaraj Joseph, and Mohanasankar Sivaprakasam. 2019. Polyp
Segmentation using Generative Adversarial Network. In 2019 41st Annual International Conference of the IEEE Engineering in
Medicine and Biology Society (EMBC). IEEE, 7201–7204.
[6] Younghak Shin, Hemin Ali Qadir, and Ilangko Balasingham. 2018. Abnormal colon polyp image synthesis using conditional
adversarial networks for improved detection performance. IEEE Access 6 (2018), 56007–56017.
[7] Ian Goodfellow, Jean Pouget-Abadie, Mehdi Mirza, Bing Xu, DavidWarde-Farley, Sherjil Ozair, Aaron Courville, and Yoshua
Bengio. 2014. Generative adversarial nets. In Advances in neural information processing systems. 2672–2680.
[8] Debesh Jha, Pia H Smedsrud, Michael A Riegler, Pål Halvorsen, Thomas de Lange, Dag Johansen, and Håvard D Johansen. 2020.
Kvasir-SEG: A Segmented Polyp Dataset. In Proc. of International Conference on Multimedia Modeling (MMM). 451–462.

15. Thank You
For
Listening…