Sreedhar Kollem
@sru.edu.in
SR University
EDUCATION
B. Tech, M. Tech, Ph.D.
Scopus Publications
Scholar Citations
Scholar h-index
Scholar i10-index
Scopus Publications
M. Naresh, V. Siva Nagaraju, Sreedhar Kollem, Jayendra Kumar, and Samineni Peddakrishna
Elsevier BV
Ch. Rajendra Prasad, Pillalamarri Shivapriya, Naragani Bhargavi, Nagaraj Ravula, Supraja Lakshmi Devi Sripathi, and Sreedhar Kollem
AIP Publishing
Sreedhar Kollem
Springer Science and Business Media LLC
Sreedhar Kollem
Springer Science and Business Media LLC
Sandip Bhattacharya, Mohammed Imran Hussain, John Ajayan, Shubham Tayal, Louis Maria Irudaya Leo Joseph, Sreedhar Kollem, Usha Desai, Syed Musthak Ahmed, and Ravichander Janapati
Wiley
S. Sreejith, L.M.I. Leo Joseph, Sreedhar Kollem, V.T. Vijumon, and J. Ajayan
Elsevier BV
Sreedhar Kollem, Katta Ramalinga Reddy, and Duggirala Srinivasa Rao
Springer Science and Business Media LLC
Sreedhar Kollem, Katta Ramalinga Reddy, Ch. Rajendra Prasad, Avishek Chakraborty, J. Ajayan, S. Sreejith, Sandip Bhattacharya, L. M. I. Leo Joseph, and Ravichander Janapati
Wiley
J. Ajayan, P. Mohankumar, D. Nirmal, L.M.I. Leo Joseph, Sandip Bhattacharya, S. Sreejith, Sreedhar Kollem, Shashank Rebelli, Shubham Tayal, and B. Mounika
Elsevier BV
Sreedhar Kollem, Ch Rajendra Prasad, J. Ajayan, V. Malathy, and Akkala Subbarao
Springer Science and Business Media LLC
Sreedhar Kollem
IEEE
The classification of MRI brain tumors is crucial in medical image processing. However, current classification techniques are ineffective in differentiating benign from malignant tumors, which raises mortality rates. This article proposes a comprehensive methodology that includes data augmentation, a hybrid CNN model, an optimal feature selection method, and a set of classifiers to increase classification accuracy and decrease mortality rates. The proposed method has four steps. First, Lipschitz-based data augmentation is used to reduce data overfitting problems. Second, in order to improve feature extraction, a hybrid CNN model that is composed of three different CNN models, such as AlexNet, VGG-19, and ResNet-50, is utilized. Thirdly, to choose the best features, the Minimum Redundancy Maximum Relevance approach is used. Finally, for the classification task, a set of classifiers comprising SVM and KNN are used. The results obtained using the proposed method are superior to those obtained using the methods that are currently being utilized.
Sreedhar Kollem
IEEE
Multi-Class MRI Brain Tumor Classification is essential for lowering mortality rates in the field of medical imaging. Nevertheless, current techniques do not distinguish between various tumor forms, including glioma, meningioma, no-tumor, and pituitary tumors. Hyper-parameters like batch size and epochs are not optimized by these approaches. As a result, classification accuracy is affected. This paper provides a uniquely optimized CNN method that combines the pre-trained AlexNet CNN model with the Rat Swarm Optimizer technique in order to overcome the aforementioned constraints. There are three stages to this process. First, images of MRI brain tumors are acquired from the BraTS2020 database, which contains images of pituitary, glioma, meningioma, and no-tumor. Second, the images were classified using the pre-trained AlexNet CNN model. Finally, the hyper-parameters like as batch size and epochs are optimized using a rat swarm optimizer to obtain the best values for improved classification. The effectiveness of the suggested technique is evaluated using the F1-score, sensitivity, specificity, and accuracy metrics. When compared to current methods, the proposed approach performs better.
Ch. Rajendra Prasad, Sami Mohammed, P.Ramchander Rao, Sreedhar Kollem, Srinivas Samala, and Srikanth Yalabaka
IEEE
A brain tumour is a dangerous form of cancer that happens when cells divide in an abnormal way. Recent advances in deep learning have helped the medical imaging sector in the diagnosis of numerous diseases. This paper presents Multiclass MRI Brain Tumour Classification with Deep Transfer Learning. In the proposed model, VGG-16 is employed as a deep transfer learning model. The dataset is collected from the Kaggle brain tumour MRI dataset, which is a combination of three popular brain tumour datasets such as figshare, SARTAJ, and Br35H datasets. The data are prepossessed by rescaling and random brightness and/or contrast by ±20% before applying to the modified VGG-16 model. The proposed model employs minimum computational resources and achieves better results in accuracy, precision, recall, and F1 score.
L.M.I. Leo Joseph, J. Ajayan, Sandip Bhattacharya, and Sreedhar Kollem
Wiley
Rajendra Ch Prasad, Sreedhar Kollem, Ravichander Janapati, Srinivas Samala, and Sandip Bhattacharya
CRC Press
K. Rajkumar, Abbady Saivardhan Reddy, Paladugula Sahith, P. Shiva Krishna Kumar, and Sreedhar Kollem
IEEE
Numerous difficulties arise while denoising images in image processing at present. The article indicates a unique deep convolution neural network-based image denoising (DCNN). This article presents a DCNN to generate the noisy image as opposed to earlier learning-based methodologies. Therefore, it is possible to extract the clean latent image from the deformed image by removing the noise. Compared to other scenarios, this strategy is the most effective. Although some of the already existing systems included machine learning techniques in addition to filters such as wiener, median, and mean filters, they were unable to produce reliable results. Experimental results show that the suggested de-noising methodology outperforms the current denoising techniques. We may say that this strategy will aid in obtaining better results by looking at factors such as PSNR, MSE, training loss, and training accuracy. The conclusions also demonstrate that multiple noises with various noise levels can be suppressed using a single denoising model.
Ramyateja Singamshetty, Sangani Sruthi, Kodati Chandhana, Sreedhar Kollem, and Ch Rajendra Prasad
IEEE
This article presents a modern brain tumor recognition method heavily based on deep neural network techniques. Although huge datasets are challenging to train, deep convolutional neural networks have excelled in many computer vision tasks. The deadliest disease in humans, according to some estimates, is a brain tumor. Anormal cells in the body can or may not spread rapidly depending on the location of the tumor and its predicted growth rate. The dataset consists of brain tumor images. To enhance the performance of the image, a data augmentation technique is used to avoid overfitting issues. Overfitting is the process by which a network learns a function with a very high variance in order to perfectly model the training data. Access to big data is limited in many application domains, including medical image analysis. The main focus of this work is data augmentation, a data-space solution to the problem of limited data. Any techniques that increase the size of the dataset can be loosely referred to as data augmentation. For instance, the image can zoom in or out and save the result, alter the brightness of the image, or rotate it. The obtained data are trained with k-fold, a data partitioning technique that enables making the most of the dataset to create a more comprehensive model. Finally, an inception framework facilitates the ability to train much deeper networks than those previously used. Three image augmentation algorithms are covered in this article: data augmentation, the k-fold model, and the inception v3 model that classifies the images by minimizing overfitting issues. Lastly, there is an application page that contains a code in the backend that predicts whether a person is suffering from a brain tumor or not. Specificity, sensitivity, and accuracy are used to assess the performance of the suggested model.
Akula Shravya Sri, Bobbili Varshith Reddy, Kanuri Balakrishna, Vollala Akshitha, Sreedhar Kollem, and Ch Rajendra Prasad
IEEE
This article presents Multi-modal MRI scans used to classify brain tumors according to their size and imaging appearance. Object detection has been significantly improved by utilizing convolutional neural networks and deep learning approaches, resulting in superior performance. Our solution to tackle uncertainty involves a new deep learning method that incorporates pre-processing techniques such as data augmentation, as well as a customized convolutional neural network. The proposed method aims to achieve three objectives. (1) To address overfitting concerns, pre-processing techniques such as data augmentation are employed. (2) A customized deep learning method is used, which includes a convolutional neural network, to classify brain tumors. (3) A Web APP is utilized to provide information related to the tumor. The suggested procedure's performance is evaluated using sensitivity, specificity, and accuracy, and it outperforms traditional deep learning methods. Additionally, the entire process can be completed on the Python platform.
Ch.Rajendra Prasad, Ramya Bandi, Devulapally Aashrith, Anjali Sampelly, Maraboina Sai Chand, and Sreedhar Kollem
IEEE
Human Activities Recognition is the process of automatically identifying a person’s physical activities in order to create a secure environment for everyone, even elderly people, in their daily lives. In this paper, the classification of human activities using Conventional Neural networks with Principal Component Analysis with presented. In the proposed method, Principal Component Analysis is employed for dimensionality reduction and Conventional Neural networks are employed for classification. The Human Activities Recognition dataset from Kaggle is used in the suggested model. The effectiveness of the proposed model is assessed in terms of accuracy. The proposed model achieved an accuracy of about 96.71%.
Ch.Rajendra Prasad, Banothu Arun, Soma Amulya, Preethi Abboju, Sreedhar Kollem, and Srikanth Yalabaka
IEEE
Breast cancer is the deadliest and most common cancer in the world. Early treatment of this cancer can help to nip it in the bud. In present medical setting, this cancer is identified by manual clinical procedures, which can lead to human errors and further delay the treatment procedure. So, we propose a Convolutional Neural Network (CNN) model employed with transfer learning approach with RESNET50, VGG19 and InceptionV3 algorithms. The histopathological image dataset is used to detect cancer cells in the tissues of the breast. We examine the performance of different models based on their accuracy, by varying different optimizers (Adam, SGDM and RMSProp) for each transfer learning model. The results show that the Inception-V3 model with Adam optimizer outperforms VGG19 and RESNET-50 in terms of accuracy.
Ch. Rajendra Prasad, Polaiah Bojja, Sreedhar Kollem, and P. Ramchandar Rao
Springer Nature Singapore
Sandip Bhattacharya, J. Ajayan, D.Nirmal, Shubham Tayal, Sreedhar Kollem, and L. M. I. Leo Joseph
Springer Science and Business Media LLC
S. Sreejith, J. Ajayan, Sreedhar Kollem, and B. Sivasankari
Springer Science and Business Media LLC
Sreedhar Kollem, Katta Ramalinga Reddy, Duggirala Srinivasa Rao, Chintha Rajendra Prasad, V. Malathy, J. Ajayan, and Deboraj Muchahary
Wiley
Various image denoising algorithms have been developed for medical imaging. But some disadvantages have been found, including the block effect, which increases smoothing, and the loss of image detail. Using the statistical properties of observed noisy images, we propose a new diffusivity function‐based partial differential equation method for image denoising. This model incorporates a Quaternion Wavelet Transform for the generation of the various noisy image coefficients, an improved generalized cross‐validation function for generating the optimal threshold value via the soft threshold function, and a new diffusivity function for controlling the diffusion process. The fourth‐order PDE diffusivity function, which is presented in this article, is a novel diffusion coefficient that is more effective at removing noise and preserving edges than previous approaches. Finally, the performance of the proposed method is evaluated using the peak signal‐to‐noise ratio, mean square error, structural similarity index, and comparisons to other traditional methods.
Ravichander Janapati, Vishwas Dalal, Rakesh Sengupta, Usha Desai, P. V. Raja Shekar, and Sreedhar Kollem
World Scientific Pub Co Pte Ltd
Currently, the operational electroencephalography (EEG)-based brain–computer interfaces (BCIs) suffer from problems of BCI latency/lag issues, which restricts the use of interfaces impractical scenarios. One of the reasons behind the present challenges is the application of a purely data-driven approach to the BCI pipeline. Although BCI applications have improved significantly with the research in the fields of artificial intelligence (AI) and machine learning (ML), fundamental issues of data-driven training restrict the latency that can be achieved under current BCI paradigms. This work explores the possibility of future BCI using a combination of data-driven and theory-driven methods. In this study, an EEG-BCI dataset from steady-state visually evoked potentials (SSVEPs) is applied, where the SSVEP signals contain, source components from the occipital, parietal and frontal regions of the brain. Source reconstruction is done with the combination of independent component analysis (ICA) and low-resolution electromagnetic tomography analysis (LORETA). This method was able to predict BCI classification labels 5[Formula: see text]s earlier, based on pre-recorded signals from the scalp. The novelty of the current contribution lies in utilizing the source reconstructed EEG time-series for BCI classification, which allows for retention of classification accuracy up to 70% while working with the reduced data dimensionality. Implementation of this algorithm will allow a significant reduction in lag in online BCIs.