Cervical cancer remains a significant global health challenge, ranking as a leading cause of cancer related deaths among women. This research addresses the need for improved detection methods by leveraging advanced image processing techniques. Implement Adaptive Median Filtering (AMF) for effective noise reduction while preserving critical details in cervical images, enhancing overall image quality. Utilizing K means clustering for segmentation, accurately isolate regions of interest, and Local Binary Patterns (LBP) for feature extraction enable improved differentiation between normal and cancerous tissues. Additionally, the integration of the Chicken Swarm Optimized algorithm with Artificial Neural Networks (CSOANN) significantly enhances classification accuracy, facilitating earlier detection of cervical cancer. These methodologies not only improve diagnostic accuracy but also contribute to better patient outcomes and more effective screening strategies, ultimately advancing cervical cancer management.
Blog Archives
HD-REGNET: Heart Disease Prediction Using REGNET in Image Processing
Abnormal functionality of the heart due to any cause is known as heart disease. It primarily affects older persons. Every year, almost six out of ten persons diagnosed with heart diseases are above the age of 65. In this paper a novel HD RegNet have been proposed for heart disease prediction using CXR images. Initial the input CXR images are pre processed using Adaptive Gaussian Star Filtering. Then the pre processed images are subjected into RegNet for extracting the features. The best features are selected using Tyrannosaurus optimization Algorithm. Finally the normal and abnormal classes of heart diseases classified utilizing the Deep Neural Network (DNN). The CHD dataset are utilized to evaluate the performance of the developed work interms of performance metrics such as accuracy, recall and precision. The proposed HD RegNet attains an accuracy rate of 99.53% in the normal class and 99.37% in the abnormal class. It achieves an overall accuracy rate 0.37%, 0.52% and 0.75% compared to the existing methods such as RFRF ILM [12], LU Net [17] and CNN & Bi LSTM [18] respectively.
Implementation of a Reverse Carry Propagate Adder Based On Efficient VLSI for Hardware Security and Area Consumption
The various techniques inbuilt in the VLSI (Very Large Scale Integration) designing is originated from the basis of adder circuits. When these adder circuits are processed in the higher digital signal processing systems, the adders have to exhibit high performance, good efficiency with reduced cost and area of implementation. The major objective of designing the VLSI design is minimizing the complexity and to perform the operations more effectively. Whereas in some other methods, the logic circuit designing is made larger in order to achieve higher efficiency. This in turn, results in the reduction of hardware security and makes the computation more complex. This paper utilizes the RCPA adder logic to reduce the area for the implementation and enhancing the hardware security. By implementing RCPA adder circuit, the carry propagates in the opposite direction from the most significant bit to the least significant bit. The usage of shift accumulator improves the speed of process and reduces the area required. By implementing the reverse propagation of the carry in the adder circuit, computation time is optimized and the implementation is made efficient with the reduction in the delay.
A Comprehensive Review of Brain Tumor Detection Using Deep Learning and Machine Learning Classifiers
A brain tumor is a destructive or non destructive anomalous tissue in the brain which affects the lifestyle of peoples. It affects the survival rate of the patients and even causes death. The conventional techniques for brain tumor detection is highly challenging and time consuming procedure. For the purpose of early brain tumor diagnosis, Brain Tumor Categorization (BTC) from Magnetic Resonance Imaging (MRI) images is extremely crucial. This paper provides a comprehensive review of brain tumor analysis using a Deep Learning (DL) and Machine Learning (ML) methods from MRI images. These models considerably upgrade the automated brain tumor diagnosis, and help the physicians and specialists to proficiently identify the tumor affected area. In this study, the DL techniques and the ML techniques are evaluated for the purpose of brain tumor identification using MRI scans. It significantly enhances the precision of medical assessments and treatment procedures and assists in separating healthy cells and tumor affected cells.
Classifying Liver Fibrosis through the Utilization of Transfer Learning and FCNET Classifiers
Abdominal Radiology is used for monitoring patients with liver diseases and to measure severity of liver fibrosis by using Ultrasound (US) images which is obtained by scanning. Scanning uses high frequency sound waves to take images of the affected human organs. Since liver is situated deep into the body the signal get weaker each time it passes through other organs which makes it difficult for diagnosis to overcome this difficult transfer learning technique is proposed in this system. The first step is to remove unwanted disturbances from the US images which is accomplished by pre-processing steps. To ensure the quality of the input images they undergo resizing of resolution, adjusting variations in brightness and contrast. The goal of pre-processing technique is to enhance the quality, uniformity, and relevance of the input images, making them suitable for feature extraction and subsequent classification of liver fibrosis stages. The next step is to apply the Expectation Maximum (EM) algorithm based on Region of Interest (ROI), which is a clustering technique used for image segmentation. The EM algorithm iteratively assigns pixels to different clusters based on their intensities and probabilistic models, aiming to identify distinct regions. The feature extraction process is carried out by Fully Connected Convolutional Neural Network (FCNET) models which enable extracting high level representations from original images and transfer learning classification is used for classification of fibrosis stages with improved accuracy.
Improved OFDM System with Electromagnetic Wave Disturbances using LDPC Coding
Wireless communication systems are able to overcome this in part by using MIMO systems.
MIMO systems ensure high throughput, wide coverage, and reliable services by accounting for
multiple transmitter and receiver antenna counts and allowing for spatial dimension. This study
proposes to build an improved OFDM system with electromagnetic wave disturbances by
applying LDPC coding. On the transmitting end, Digitizing impulses to physical form requires
the employment of encoding and decoding systems, depending regarding the receiving end,
analog signals to digital signals. In order to transport data, quadrature phase shift keying
(QPSK) modulators alter the carrier signal’s frequency and amplitude. Using the Inverse
Discrete Fourier Transform (IDFT) method, one can demodulate the modulated signals. One
popular channel model is additive white Gaussian noise (AWGN). Within digital interpersonal
networks, the Minimum Mean Squared Error (MMSE) equalizer lessens the effect of errors
and noise. ML outperforms previous methods by applying the MMSE and ML equalizers on
the AWGN channel. Demodulators of Quadrature Phase Shift Keying (QPSK) demodulate
signals. The trial was successful, so the methodology is being applied to search for possible
enhancements. This project uses the features of the MATLAB software.
Spectral Domain Wavelet Transform Based Channel Estimation for VLC Communication
In the future, future heterogeneous communications networks may use visible light
communications (VLC) as a supplementary technology to WiFi. The signal to noise ratio
(SNR) of the received signal determines a VLC system’s channel capacity, just like it does for
any other communications system. The channel estimate for VLC communication proposed in
this study is based on spectral domain wavelet transform. Due to their excellent spectrum and
energy efficiency for VLC MIMO communication systems, SMTs have become more and more
common. This work focuses on SMTs, or entirely generalized spatial modulation, in which
multiple active optical antennas and a constant number of antennae are used to transmit data
symbols for VLC at any time interval. In this study, an adaptive channel estimation approach
over a time-varying MIMO channel is proposed using the Spectral Domain Wavelet Transform.
Furthermore, the broadcast data and optical antenna indices are identified using a maximum
likelihood (ML) decoder based on the received signal and the predicted VLC-MIMO channel.
Matlab software is used in the implementation of this project.
Pancreatic Cancer Identification using Convolutional Neural Networks
Rarely is pancreatic cancer discovered in its early stages, when it is most treatable. This is because, in many cases, symptoms do not appear until the disease has progressed to other organs. Treatment options for pancreatic cancer are selected according to the cancer’s stage. Options might be radiation treatment, chemotherapy, surgery, or a mix of these. Cancers have fuzzy borders and tiny size, making them challenging to manually annotate and automatically segment. That is why cancer prediction is so important. The convolutional neural network classifier proposed in this study is intended to identify pancreatic cancer. The segmentation method used in the test, Fuzzy C Means, divides the picture into segments. The Gabor based Region Covariance Matrix (GRCM) is used to extract features, and the GW optimization method is used to optimize the process. Using a powerful classifier a Grey wolf Optimization based Convolutional Neural Networks (GWO-based CNN), the outcome is correctly predicted. The findings acquired through the use of CNN Classifier were precise. MATLAB simulation software is used in the implementation of this project. The Accuracy Comparison of the GWO-CNN is 92.5% and Specificity Comparison of GWO-CNN is 93% respectively.
Detection of Intelligent Machine Fault using Deep Learning Classification
Researchers have long been interested in developing fault detection methods for rotating machinery, and engineers and scientists are increasingly focusing on artificial intelligence-based approaches. When using other signal processing techniques to extract fault characteristics or categorize fault features, artificial neural networks especially deep learning-based techniques—are widely employed. The methodologies and these studies are closely related. This deep learning classification to detect intelligent machine errors. This developed a deep learning algorithm’s technique for the classification of machine faults. To facilitate the process, the input dataset is pre-processed. Researching the observed data is part of the data analysis process known as data pre-processing. The dataset is fed into the segmentation block following pre-processing. K-value to precisely identify the size and shape of the faults, morphological operations are employed in conjunction with segmentation for image segmentation. It attempts to maintain as much difference between the clusters as well as much similarity between the intra-cluster data points. The deep learning method includes the classifications of long-short-term memory networks (LSTM). When it comes to fault diagnosis in that area, LSTM classifier is designed to categorize faults according to their type.
Implementation and Prediction of Diabetic Retinopathy Types Based on Deep Convolutional Neural Networks
Diabetic retinopathy (DR) is the primary cause of preventable blindness in those within the working age group in developed nations. The macula, optic discs, and blood vessels that make up a healthy retina are irregularities that indicate a true eye condition. Therefore, retinopathy detection is crucial. Therefore, this study suggests using deep convolutional neural networks to build and predict different kinds of diabetic retinopathy. A pre-processing approach first initiates the input picture. The spatial processing method known as the Gaussian filter, which reduces picture noise and increases the clarity of fuzzy pictures, handles this procedure. The picture is sent to the regions for segmentation in the next stage. The size and form of the illness may be precisely determined using fuzzy C-means (FCM) segmentation. Additionally, it attempts to maintain as much difference between the clusters as well as much similarity between the intra-cluster data points. The next phase in the process is to use deep convolutional neural networks (DCNN). In order to extract features from the input picture, the convolutional layer applies filters. To expedite computation, the pooling layer samples the picture; the fully connected layer then generates the final prediction. The early detection of diabetic retinopathy has been made easier by the use of DCNN algorithms. Within the domain of medical image processing, the DCNN algorithm represents a methodically ordered approach and it shows 94.7%. This work offers a high degree of sensitivity and precision in identifying whether a picture has diabetic retinopathy or not. Creating the output image is the final stage. Python Google Colab software is used in the implementation of this paper.