MUTHUSELVI S

Scopus Publications

An Efficient medical decision-making system for skin cancer classification using SENet
S Muthuselvi, R. Sumathi
A Study on Next Generation Materials and Devicesv, 2025
Quick diagnosis is crucial for identifying and treating skin cancer. A highly efficient medical decision-making system using dermoscopic images is essential for assessing skin malignancies. Recently, significant advancements have been made with Faster Region-Convolutional Neural Network (R-CNN) in detecting skin disease types. Machine learning algorithms, particularly pre-trained convolutional neural networks (CNNs), have shown promise in identifying skin cancer from medical images with minimal data. However, these models often struggle due to the limited availability of malignant tumour images. The goal of this research is to build a highly accurate Faster R-CNN-based model for diagnosing various skin cancers, such as melanoma. We propose enhancing the SENet model by incorporating new data and adding an additional CNN layer. This approach improves the algorithm’s ability to handle disorganized and limited data. Using a dataset of 2638 skin images, we Indicates the effectiveness of our approach, evaluating it based on precision, sensitivity, specificity, F1-score, and the area under the ROC curve (AUC). Our improved SENet Mobile and SENet Large models achieve accuracy ratings of 89.61% and 91.97%, respectively, using the Adam optimization algorithm.
Analysis of Skin Cancer Classification using Deep Learning Techniques
S. Muthuselvi, R. Sumathi
1st International Conference on Innovative Engineering Sciences and Technological Research Iciestr 2024 Proceedings, 2024
Skin cancer is defined as abnormal development and proliferation of skin cells. It is in various steps. Malignant skin cancer is classified into different kinds: carcinoma of the basal cell, carcinoma of the squamous, and melanoma or melanocytic carcinoma. Each of these is called by its name because it occurs in specific cells. Cancer tumors can be observed because skin cancer occurs in the upper layer of the skin, the epidermis. Unlike other cancers, these can be detected and treated at an early stage, which is why the death rate from this type of cancer is relatively low. The primary cause of skin cancer is prolonged exposure to ultraviolet rays emitted by radiation. Melanoma along with additional skin cancers are more prevalent in the general population than lung, colon, breast, and prostate tumors. Melanoma is the deadliest form of malignancy of the skin and has a high death rate compared to other skin cancers, although melanoma is less common in the general population. Among the most common skin cancers are non-melanoma skin cancers. Some people have the possibility of developing melanoma later in life because of a large birthmark (large mole) that appears at birth or later due to melanocytes. Therefore, evidence-based automatic identification of cancer of the skin is useful for improving pathologists' precision and competency in their initial phase. In this work, we suggest a model developed by RCNN deep learning for accurately distinguishing between malignant and benign lesions in the skin. First, we implement a filter or kernel programming to remove noise and artifacts; then, we normalize the source pictures and determine characteristics assisting in precise classification. Furthermore, data enrichment builds up the overall count of images, which gets a better classifying rate its precision. The performance of our proposed RCNN model is comparable to other transfer learning approaches like the algorithm AlexNet, ResNet, The VGG model 16, DenseNet, MobileNet, and more. The approach was tested using the HAM10000 collection of data, with the model achieving the best accuracy during training and testing of 86.16% and 88.93%, accordingly Examining current transferred learning models, the end causes of our proposed RCNN model are precise and potent.
Empowering the Tribal people with the use of big data processing expert system in animal Husbandry and Poultry Farming application
R. Saravanan, V. Nehru, S. Muthuselvi
2023 IEEE International Conference on Research Methodologies in Knowledge Management Artificial Intelligence and Telecommunication Engineering Rmkmate 2023, 2023
The population of the tribal people has been decreased day by day in India due to the lack of awareness in health related issues and there is a series challenges in their sustainable livelihood. The Particularly Vulnerable People from tribal groups (PVTGs) engaged in animal husbandry and poultry farming as their primary source of income, which improves their standard of living. Maintain and safeguard poultry and animals from diseases is a cumbersome process. Providing enough medical facility is still a challenging task due to the geographical location and unavailability of the infrastructure and human resources. The proposed framework uses Apache Kafka-Apache Storm-NoSQL Mongo DB architecture to process enormous volume of sensor data in real time and it receives the sensor data and uses it to create the various disease identification models. The processed data are stored in Mongo DB as a historical data. The system provides a Web-based monitoring system for continuos monitoring the health conditions of cattles and poultry through the Smart Health Care Centre. Smartness in operation is performed through System on Chip (SoC) IoT system, the proposed big data expert system model transcends from the traditional functionalities of disease identification by the real time field visit analysis by the medical professionals. The proposed system is more suitable for the remote hill area. Smart Health Care system improves the disease identification accuracy and provides a powerful Big Data architecture for data analytics and data storage. The big data expert system frame work is underwent successful functional testing of "SoC-IoT smart devices" connected with the network and the performance of the network in terms of CPU, memory usage and the network delay is analyzed. Further the frame work uses the big data processing with the machine learning approach "Hybrid diseases identification Model" with the combination of DBSCAN for outlier detection together with Random Forest classification, which improves the disease identification accuracy of the various disease attacked the cattles and poultry.
2d to 3d conversion with combined texture features anddisparity mapping
S, Scintiaclarinda, , K. Nithya, S. Muthuselvi, A. Rengarajan, , , and
International Journal of Engineering and Advanced Technology, 2019
Image processing is a strategy to change over a picture into advanced structure also, play out specific tasks on it, in order to get an upgraded picture or to extricate some profitable information from it. The transformation procedure of existing 2D pictures to 3D is financially feasible and is satisfying the development of high caliber stereoscopic pictures. A disparity map maybe live of however totallydifferent 2 pictures areaunit, however far similar edge/corner/feature points area unit from one image to another.This idea executes the plan of the programmed 2D to 3D video shading transformation utilizing 2D video and grouping is displayed. The examined structures epitomize along procedure of neighboring casings abuse the ensuing procedure: CIELa*b* shading space transformation, wavelet change (WT) with edge location utilizing HF wavelet sub-groups (HF, LH and HH) or pyramidal plan, shading division through k-implies on a*b* shading plane, up-testing in wavelet case, dissimilarity map (DM) estimation at long last, the dissected 3D scene age.
Applying of a company’s stock price prediction using data mining
S. Muthuselvi, , A. Rengarajan, S, Scintiaclarinda, K. Nithya, , , and
International Journal of Recent Technology and Engineering, 2019
Stock market analysis is a common economic activity that has been an attractive topic to research and used in different forms of day-to-day life in order to predict the stock prices. Techniques like major analysis, Statistical investigation, Time arrangement analysis and so on are reliably worthy forecast device. In this paper, Data mining, Machine learning (ML) and Sentiment analysis are techniques used for analyzing public emotions in order predict the future stock prices. The goal of a project is to review totally different techniques to predict stock worth movement victimization the sentiment analysis from social media, data processing. Sentiment classifiers are designed for social media text like product reviews, blog posts, and email corpus messages. In the company’s communication network, information mining calculation is utilized as to mine email correspondence records and verifiable stock costs. Implementing various Machine learning and Classification models such as Deep Neural network, Random forests, Support Vector Machine, the company can successfully implemented a company-specific model capable of predicting stock price movement with efficient accuracy.