arXiv subjects

Ilangko Balasingham

Publications and source records attributed to Ilangko Balasingham.

25 records · Page 2Linked to original sources

arXiv2019

Polyp Detection and Segmentation using Mask R-CNN: Does a Deeper Feature Extractor CNN Always Perform Better?

Automatic polyp detection and segmentation are highly desirable for colon screening due to polyp miss rate by physicians during colonoscopy, which is about 25%. However, this computerization is still an unsolved problem due to various polyp-like structures in the colon and high interclass polyp variations in terms of size, color, shape, and texture. In this paper, we adapt Mask R-CNN and evaluate its performance with different modern convolutional neural networks (CNN) as its feature extractor for polyp detection and segmentation. We investigate the performance improvement of each feature extractor by adding extra polyp images to the training dataset to answer whether we need deeper and more complex CNNs or better dataset for training in automatic polyp detection and segmentation. Finally, we propose an ensemble method for further performance improvement. We evaluate the performance on the 2015 MICCAI polyp detection dataset. The best results achieved are 72.59% recall, 80% precision, 70.42% dice, and 61.24% Jaccard. The model achieved state-of-the-art segmentation performance.

cs.CV↗

arXiv2019

Automatic Colon Polyp Detection using Region based Deep CNN and Post Learning Approaches

Automatic detection of colonic polyps is still an unsolved problem due to the large variation of polyps in terms of shape, texture, size, and color, and the existence of various polyp-like mimics during colonoscopy. In this study, we apply a recent region based convolutional neural network (CNN) approach for the automatic detection of polyps in images and videos obtained from colonoscopy examinations. We use a deep-CNN model (Inception Resnet) as a transfer learning scheme in the detection system. To overcome the polyp detection obstacles and the small number of polyp images, we examine image augmentation strategies for training deep networks. We further propose two efficient post-learning methods such as, automatic false positive learning and off-line learning, both of which can be incorporated with the region based detection system for reliable polyp detection. Using the large size of colonoscopy databases, experimental results demonstrate that the suggested detection systems show better performance compared to other systems in the literature. Furthermore, we show improved detection performance using the proposed post-learning schemes for colonoscopy videos.

cs.CV↗

arXiv2019

Abnormal Colon Polyp Image Synthesis Using Conditional Adversarial Networks for Improved Detection Performance

One of the major obstacles in automatic polyp detection during colonoscopy is the lack of labeled polyp training images. In this paper, we propose a framework of conditional adversarial networks to increase the number of training samples by generating synthetic polyp images. Using a normal binary form of polyp mask which represents only the polyp position as an input conditioned image, realistic polyp image generation is a difficult task in a generative adversarial networks approach. We propose an edge filtering-based combined input conditioned image to train our proposed networks. This enables realistic polyp image generations while maintaining the original structures of the colonoscopy image frames. More importantly, our proposed framework generates synthetic polyp images from normal colonoscopy images which have the advantage of being relatively easy to obtain. The network architecture is based on the use of multiple dilated convolutions in each encoding part of our generator network to consider large receptive fields and avoid many contractions of a feature map size. An image resizing with convolution for upsampling in the decoding layers is considered to prevent artifacts on generated images. We show that the generated polyp images are not only qualitatively realistic but also help to improve polyp detection performance.

eess.IV↗

arXiv2014

Heart-induced movements in the thorax as detected by MRI

In order to provide information for the use of radar in diagnostics a qualitative map of movements in the thorax has been obtained. This map was based on magnetic resonance image sequences of a human thorax during suspended respiration. The movements were measured using two distinct techniques. Segmentation provided measures of aorta dilatation and displacements, and image edge detection indicated other movements. The largest heart movements were found in the anterior and left regions of the heart with in-plane displacements on the order of 1 cm and which caused lung vessels displacements on the order of 2-3mm especially on the left side due to the heart ventricular. Mechanical coupling between the heart and aorta caused aorta displacements and shape distortions. Despite this coupling, aorta dilatations most likely reflected blood pressure variations.

physics.med-ph↗

arXiv2014

Zero-Delay Joint Source-Channel Coding for a Multivariate Gaussian on a Gaussian MAC

In this paper, communication of a Multivariate Gaussian over a Gaussian Multiple Access Channel is studied. Distributed zero-delay joint source-channel coding (JSCC) solutions to the problem are given. Both nonlinear and linear approaches are discussed. The performance upper bound (signal-to-distortion ratio) for arbitrary code length is also derived and Zero-delay cooperative JSCC is briefly addressed in order to provide an approximate bound on the performance of zero-delay schemes. The main contribution is a nonlinear hybrid discrete-analog JSSC scheme based on distributed quantization and a linear continuous mapping named Distributed Quantizer Linear Coder (DQLC). The DQLC has promising performance which improves with increasing correlation, and is robust against variations in noise level. The DQLC exhibits a constant gap to the performance upper bound as the signal-to-noise ratio (SNR) becomes large for any number of sources and values of correlation. Therefore it outperforms a linear solution (uncoded transmission) in any case when the SNR gets sufficiently large.

cs.IT↗

arXiv2012

Reconstruction mechanism with self-growing equations for hyper, improper and proper fractional chaotic systems through a novel non-Lyapunov approach

Identification of the unknown parameters and orders of fractional chaotic systems is of vital significance in controlling and synchronization of fractional-order chaotic systems. However there exist basic hypotheses in traditional estimation methods, that is, the parameters and fractional orders are partially known or the known data series coincide with definite forms of fractional chaotic differential equations except some uncertain parameters and fractional orders. What should I do when these hypotheses do not exist? In this paper, a non-Lyapunov novel approach with a novel united mathematical model is proposed to reconstruct fractional chaotic systems, through the fractional-order differential equations self-growing mechanism by some genetic operations ideas independent of these hypotheses. And the cases of identifying the unknown parameters and fractional orders of fractional chaotic systems can be thought as special cases of the proposed united mathematical reconstruction method in non-Lyapunov way. The problems of fractional-order chaos reconstruction are converted into a multiple modal non-negative special functions' minimization through a proper translation, which takes fractional-order differential equations as its particular independent variables instead of the unknown parameters and fractional orders. And the objective is to find best form of fractional-order differential equations such that the objective function is minimized. Simulations are done to reconstruct a series of hyper and normal fractional chaotic systems. The experiments' results show that the proposed self-growing mechanism of fractional-order differential equations with genetic operations is a successful methods for fractional-order chaotic systems' reconstruction, with the advantages of high precision and robustness.

nlin.CD↗

arXiv2012

Zero-Delay Joint Source-Channel Coding for a Bivariate Gaussian on a Gaussian MAC

In this paper, delay-free, low complexity, joint source-channel coding (JSCC) for transmission of two correlated Gaussian memoryless sources over a Gaussian Multiple Access Channel (GMAC) is considered. The main contributions of the paper are two distributed JSCC schemes: one discrete scheme based on nested scalar quantization, and one hybrid discrete-analog scheme based on a scalar quantizer and a linear continuous mapping. The proposed schemes show promising performance which improve with increasing correlation and are robust against variations in noise level. Both schemes exhibit a constant gap to the performance upper bound when the channel signal-to-noise ratio gets large.

cs.IT↗