Faculty Publications

Permanent URI for this communityhttps://idr.nitk.ac.in/handle/123456789/18736

Publications by NITK Faculty

Browse

Has files: NoSubject: search.filters.subject.convolutional neural network

Search Results

Now showing 1 - 10 of 29
  • No Thumbnail Available
    Item
    A Lightweight Convolutional Neural Network Model for Tuberculosis Bacilli Detection From Microscopic Sputum Smear Images
    (wiley, 2021) Panicker, R.O.; Pawan, S.J.; Rajan, J.; Sabu, M.K.
    This chapter describes a lightweight convolutional neural network model that automatically detects Tuberculosis (TB) bacilli from sputum smear microscopic images. According to WHO, about onefourth of the population in the universe is infected with TB, and every day five thousand people are killed due to TB disease. There are well-known recommended diagnostics are available for TB detection, among them sputum smear microscopic examination is a primary and most efficient recommended method for most of the developing and moderately developed countries. However, this manual detection method is highly error-prone and time-consuming. In this chapter, we proposed a lightweight CNN model for classifying Tuberculosis bacilli from non-bacilli objects. We adopted a Convolutional Neural Network (CNN) architecture with a skip connection of variable lengths that can identify TB bacilli from sputum smear microscopic images. The performance of the proposed model in terms of accuracy is close to the state-of-the-art. However, the number of parameters in the proposed model is significantly less than other recently proposed models. © 2021 Scrivener Publishing LLC.
  • No Thumbnail Available
    Item
    Classification of Capsule Endoscopy Images based on Feature Concatenation of Deep Neural Networks
    (Institute of Electrical and Electronics Engineers Inc., 2021) Biradher, S.; Aparna., P.
    Wireless capsule endoscopy (WCE) is a noninvasive way of detecting abnormalities in digestive tract. These abnormalities need to be detected at the early stages before they turn malignant. The classification of these abnormalities has put many challenges due to the variations in lesion shape and color, lighting conditions, and other factors. Existing methods based on handcrafted features give less accuracy due to the limited capability of feature representation. This study proposes a new approach for classifying wireless capsule endoscopy images using feature concatenation of deep convolutional neural network models. The features of two pre-trained models are concatenated and tested using a newly created dataset. The dataset is created using images taken from the Kvasir capsule endoscopy and Red lesion endoscopy dataset which is publicly available. This system improves diagnostic efficiency and brings great assistance to the doctor. © 2021 IEEE.
  • No Thumbnail Available
    Item
    Blindness (Diabetic Retinopathy) Severity Scale Detection
    (Institute of Electrical and Electronics Engineers Inc., 2021) Bygari, R.; Naik, R.; Uday Kumar, P.
    Diabetic retinopathy (DR) is a severe complication of diabetes that can cause permanent blindness. Timely diagnosis and treatment of DR are critical to avoid total loss of vision. Manual diagnosis is time consuming and error-prone. In this paper, we propose a novel deep learning based method for automatic screening of retinal fundus images to detect and classify DR based on the severity. The method uses a dual-path configuration of deep neural networks to achieve the objective. In the first step, a modified UNet++ based retinal vessel segmentation is used to create a fundus image that emphasises elements like haemorrhages, cotton wool spots, and exudates that are vital to identify the DR stages. Subsequently, two convolutional neural networks (CNN) classifiers take the original image and the newly created fundus image respectively as inputs and identify the severity of DR on a scale of 0 to 4. These two scores are then passed through a shallow neural network classifier (ANN) to predict the final DR stage. The public datasets STARE, DRIVE, CHASE DB1, and APTOS are used for training and evaluation. Our method achieves an accuracy of 94.80% and Quadratic Weighted Kappa (QWK) score of 0.9254, and outperform many state-of-the-art methods. © 2021 IEEE.
  • No Thumbnail Available
    Item
    Classification of Wireless Capsule Endoscopy Bleeding Images using Deep Neural Network
    (Institute of Electrical and Electronics Engineers Inc., 2022) Biradher, S.; Aparna., P.
    Wireless capsule endoscopy (WCE) is an innovative video technique that allows a non-invasive visual inspection of the whole gastrointestinal system. The digestive tract anomalies must be identified in advance to treat them before they turn into hazardous cancers. Detection and classification of digestive tract associated anomalies have been difficult because of various factors like differences in lesion form and color, level of illumination or lighting, etc. Existing approaches based on non-deep learning methods include a manually designed feature extraction step, which is less efficient since these manually designed features may lose essential information and cannot be optimised because they are not part of an end-to-end learning system. This paper provides a new way of classification between bleeding and non-bleeding classes of wireless capsule endoscopy images. The proposed method makes use of a simple Deep Convolutional Neural Network which consists of six convolutional layers alternated with max-pooling layers and this technique is compared with existing ones in terms of different performance metrics. This approach boosts diagnostic efficiency and gives doctors a huge amount of support. © 2022 IEEE.
  • No Thumbnail Available
    Item
    Transaction fee forecasting in post EIP-1559 Ethereum using 1-D Convolutional Neural Network
    (Institute of Electrical and Electronics Engineers Inc., 2023) Kallurkar, H.S.; Chandavarkar, B.R.
    Cryptocurrencies have established their identity as a healthy alternative to the maintenance of digital assets. Their applications include low-cost money transfers and yield farming. Ethereum is a blockchain that provides the functionality of doing more than a transaction regarding cryptocurrency. Ether is the default cryptocurrency of Ethereum, which is issued to the miners after the successful completion of the consensus mechanism to avoid fraudulent miners gaining profits. Transactions in Ether require that the user should include what is called a 'fee' besides the amount that is sent by the user. The EIP-1559 (Ethereum Improvement Proposals) upgrade to the Ethereum protocol has substantially changed how the transaction fee is calculated. Since this transaction data can be considered time-series data, many prior approaches have been proposed to forecast such a transaction fee using suitable methods effectively. One-dimensional Convolutional Neural Networks have recently been successfully applied to time-series forecasting problems, showing promising results. This paper proposes a univariate 1-D CNN for an effective forecast of transaction fees in the new Ethereum protocol. Furthermore, this paper also compares the proposed method with existing standard approaches, and the results show the superior performance of simple 1-dimensional convolutional neural networks over existing hybrid models. © 2023 IEEE.
  • No Thumbnail Available
    Item
    Automatic detection and localization of Focal Cortical Dysplasia lesions in MRI using fully convolutional neural network
    (Elsevier Ltd, 2019) Bijay Dev, K.M.; Pawan, P.S.; Niyas, S.; Vinayagamani, S.; Kesavadas, C.; Rajan, J.
    Focal cortical dysplasia (FCD) is the leading cause of drug-resistant epilepsy in both children and adults. At present, the only therapeutic approach in patients with drug-resistant epilepsy is surgery. Hence, the quantification of FCD via non-invasive imaging techniques helps physicians to decide on surgical interventions. The properties like non-invasiveness and capability to produce high-resolution images makes magnetic resonance imaging an ideal tool for detecting the FCD to an extent. The FCD lesions vary in size, shape, and location for different patients and make the manual detection time consuming and sensitive to the experience of the observer. Automatic segmentation of FCD lesions is challenging due to the difference in signal strength in images acquired with different machines, noise, and other kinds of distortions such as motion artifacts. Most of the methods proposed in the literature use conventional machine learning and image processing techniques in which their accuracy relies on the trained features. Hence, feature extraction should be done more precisely which requires human expertise. The ability to learn the appropriate features/representations from the training data without any human interventions makes the convolutional neural network (CNN) the suitable method for addressing these drawbacks. As far as we are aware, this work is the first one to use a CNN based model to solve the aforementioned problem using only MRI FLAIR images. We customized the popular U-Net architecture and trained the proposed model from scratch (using MRI images acquired with 1.5T and 3T scanners). FCD detection rate (recall) of the proposed model is 82.5 (33/40 patients detected correctly). © 2019
  • No Thumbnail Available
    Item
    Stack generalized deep ensemble learning for retinal layer segmentation in Optical Coherence Tomography images
    (Elsevier Sp. z o.o., 2020) Anoop, B.N.; Pavan, R.; Girish, G.N.; Kothari, A.R.; Rajan, J.
    Segmentation of retinal layers is a vital and important step in computerized processing and the study of retinal Optical Coherence Tomography (OCT) images. However, automatic segmentation of retinal layers is challenging due to the presence of noise, widely varying reflectivity of image components, variations in morphology and alignment of layers in the presence of retinal diseases. In this paper, we propose a Fully Convolutional Network (FCN) termed as DelNet based on a deep ensemble learning approach to selectively segment retinal layers from OCT scans. The proposed model is tested on a publicly available DUKE DME dataset. Comparative analysis with other state-of-the-art methods on a benchmark dataset shows that the performance of DelNet is superior to other methods. © 2020 Nalecz Institute of Biocybernetics and Biomedical Engineering of the Polish Academy of Sciences
  • No Thumbnail Available
    Item
    NucleiSegNet: Robust deep learning architecture for the nuclei segmentation of liver cancer histopathology images
    (Elsevier Ltd, 2021) Lal, S.; Das, D.; Alabhya, K.; Kanfade, A.; Kumar, A.; Kini, J.R.
    The nuclei segmentation of hematoxylin and eosin (H&E) stained histopathology images is an important prerequisite in designing a computer-aided diagnostics (CAD) system for cancer diagnosis and prognosis. Automated nuclei segmentation methods enable the qualitative and quantitative analysis of tens of thousands of nuclei within H&E stained histopathology images. However, a major challenge during nuclei segmentation is the segmentation of variable sized, touching nuclei. To address this challenge, we present NucleiSegNet - a robust deep learning network architecture for the nuclei segmentation of H&E stained liver cancer histopathology images. Our proposed architecture includes three blocks: a robust residual block, a bottleneck block, and an attention decoder block. The robust residual block is a newly proposed block for the efficient extraction of high-level semantic maps. The attention decoder block uses a new attention mechanism for efficient object localization, and it improves the proposed architecture's performance by reducing false positives. When applied to nuclei segmentation tasks, the proposed deep-learning architecture yielded superior results compared to state-of-the-art nuclei segmentation methods. We applied our proposed deep learning architecture for nuclei segmentation to a set of H&E stained histopathology images from two datasets, and our comprehensive results show that our proposed architecture outperforms state-of-the-art methods. As part of this work, we also introduced a new liver dataset (KMC liver dataset) of H&E stained liver cancer histopathology image tiles, containing 80 images with annotated nuclei procured from Kasturba Medical College (KMC), Mangalore, Manipal Academy of Higher Education (MAHE), Manipal, Karnataka, India. The proposed model's source code is available at https://github.com/shyamfec/NucleiSegNet. © 2020 Elsevier Ltd
  • No Thumbnail Available
    Item
    A cascaded convolutional neural network architecture for despeckling OCT images
    (Elsevier Ltd, 2021) Anoop, B.N.; Kalmady, K.S.; Udathu, A.; Siddharth, V.; Girish, G.N.; Kothari, A.R.; Rajan, J.
    Optical Coherence Tomography (OCT) is an imaging technique widely used for medical imaging. Noise in an OCT image generally degrades its quality, thereby obscuring clinical features and making the automated segmentation task suboptimal. Obtaining higher quality images requires sophisticated equipment and technology, available only in selected research settings, and is expensive to acquire. Developing effective denoising methods to improve the quality of the images acquired on systems currently in use has potential for vastly improving image quality and automated quantitative analysis. Noise characteristics in images acquired from machines of different makes and models may vary. Our experiments show that any single state-of-the-art method for noise reduction fails to perform equally well on images from various sources. Therefore, detailed analysis is required to determine the exact noise type in images acquired using different OCT machines. In this work we studied noise characteristics in the publicly available DUKE and OPTIMA datasets to build a more efficient model for noise reduction. These datasets have OCT images acquired using machines of different manufacturers. We further propose a patch-wise training methodology to build a system to effectively denoise OCT images. We have performed an extensive range of experiments to show that the proposed method performs superior to other state-of-the-art-methods. © 2021 Elsevier Ltd
  • No Thumbnail Available
    Item
    Multi-Res-Attention UNet: A CNN Model for the Segmentation of Focal Cortical Dysplasia Lesions from Magnetic Resonance Images
    (Institute of Electrical and Electronics Engineers Inc., 2021) Thomas, E.; Pawan, S.J.; Kumar, S.; Horo, A.; Niyas, S.; Vinayagamani, S.; Kesavadas, C.; Rajan, J.
    In this work, we have focused on the segmentation of Focal Cortical Dysplasia (FCD) regions from MRI images. FCD is a congenital malformation of brain development that is considered as the most common causative of intractable epilepsy in adults and children. To our knowledge, the latest work concerning the automatic segmentation of FCD was proposed using a fully convolutional neural network (FCN) model based on UNet. While there is no doubt that the model outperformed conventional image processing techniques by a considerable margin, it suffers from several pitfalls. First, it does not account for the large semantic gap of feature maps passed from the encoder to the decoder layer through the long skip connections. Second, it fails to leverage the salient features that represent complex FCD lesions and suppress most of the irrelevant features in the input sample. We propose Multi-Res-Attention UNet; a novel hybrid skip connection-based FCN architecture that addresses these drawbacks. Moreover, we have trained it from scratch for the detection of FCD from 3 T MRI 3D FLAIR images and conducted 5-fold cross-validation to evaluate the model. FCD detection rate (Recall) of 92% was achieved for patient wise analysis. © 2013 IEEE.