A Novel Optimized Recurrent Network-Based Automatic System for Speech Emotion Identification

Koppula, Neeraja; Rao, Koppula Srinivas; Nabi, Shaik Abdul; Balaram, Allam

doi:10.1007/s11277-022-10040-5

Cited by 14 publications

(1 citation statement)

References 70 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The ROI is then adjusted using an automated threshold approach and threshold selection criteria. Segmentation examples of magnetic resonance imaging (MRI) brain scan at T1 & T2 weights, highlight the subtler characteristics of brain tissue [19]. It is the goal of a feature selection strategy, such as the class labels in classification, to pick a limited group of characteristics that minimizes duplication and increases the relevance to the objective.…”

Section: Literature Surveymentioning

confidence: 99%

Cloud-Based LeNet-5 CNN for MRI Brain Tumor Diagnosis and Recognition

Srinivasarao,

Rajesh,

Saikumar

et al. 2023

View full text Add to dashboard Cite

Early and accurate diagnosis of brain tumors is crucial in the medical field, as undetected or misdiagnosed tumors can lead to sudden death. Traditional models for diagnosing brain tumors suffer from low Time of Conversion (ToC) and low accuracy, contributing to a high mortality rate among the 5 million people affected by brain disease annually, as reported by the World Health Organization (WHO). Previous methods, such as Elastic Net Regression (ENR), Logistic Regression (LR), and other machine learning models, struggle to accurately locate and identify brain lesions. Moreover, these models are not suited for cloud-based platforms. To address this issue, we developed a sophisticated, cloud-based brain abnormality detection application using the LeNet-5 Convolutional Neural Network (CNN) on the DriveHQ platform. The pre-trained LeNet-5 model extracts features from ADNI-1, ADNI-2, and MIRIAD datasets. Real-time MRI brain images were collected from Manipal Hospital in Vijayawada, Andhra Pradesh, India. The LeNet-5 model employs hidden layers, flattened layers, max-pooling layers, dense layers, and ReLu layers for optimal performance. Our 2D-LeNet-5 CNN approach preprocesses images using split and merge techniques of binary mask segmentation. The Python 3.7 software tool was used to train and test datasets to identify abnormalities in MRI brain images. The proposed application achieved remarkable performance metrics, including 99.65% accuracy, 99.59% sensitivity, 99.72% F1-score, 99.25% recall, 59.32 PSNR, and 0.9929 MCC. These results demonstrate the superiority of our methodology in comparison to existing models, making it a promising solution for cloud-based brain tumor diagnosis and recognition.

show abstract

Section: Literature Surveymentioning

confidence: 99%