Browsing by Author "Mollah, M. M. Imran"

Now showing 1 - 2 of 2

Automatic Brain Tumor Detection Using Feature Selection and Machine Learning from MRI Images.
(Proceedings of the 6th International Conference on Electrical, Control and Computer Engineering. Lecture Notes in Electrical Engineering, vol 842. Springer, Singapore, 2023-03-09) Shafi, A. S. M.; Hasan, Md. Mahmudul; Mollah, M. M. Imran; Alam, Mohammad Khurshed; Islam, Md. Tarequl
A brain tumor is a group of defective cells in the brain. It happens when a cell in the brain develops a dysfunctional structure. Nowadays it becom-ing a crucial factor of death for a large number of people. Among all the varie-ties of tumors, the seriousness of a brain tumor is high. Therefore, instant detec-tion and proper care to be done to save a life from brain tumors. Microscopic examination can separate the tumor cells from healthy cells. They are typically less well separated than normal cells. In modern imaging technology, the de-tection and classification of brain tumors is a primary concern. For a clinical supervisor or radiologist, it is time-consuming and frustrating work. The accu-racy of recognition and classification of tumors executed by radiologists or clin-ical experts is depended on their experience only. Therefore, accurate identifi-cation and classification of brain tumors can be determined by image processing techniques. This research suggests a machine learning module to detect brain tumors using magnetic resonance imaging (MRI) of brain tumors. The method consists of pre-processing of nearly raw raster data (NRRD) of the MRI images, feature extraction, feature selection, and the classification learner to evaluate and construct the final model. The classification learner is designed with a sup-port vector machine (SVM) classifier. The classification method performs well with weighted sensitivity, specificity, precision, and accuracy of 98.81%, 98.88%, 98.82%, and 98.81% respectively. The findings may infer a remarka-ble step for detecting the presence of tumors in neuro-medicine diagnosis.
Feature Selection and Prediction of Heart Disease Using Machine Learning Approaches
(Proceedings of the 6th International Conference on Electrical, Control and Computer Engineering. Lecture Notes in Electrical Engineering, vol 842. Springer, Singapore. https://doi.org/10.1007/978-981-16-8690-0_83, 2023-03-09) Mollah, M. M. Imran; Islam, Md. Sakirul; Shafi, A. S. M.; Alam, Mohammad Khurshed; Islam, Md. Tarequl; Jui, Julakha Jahan
Heart Disease (HD) is the world's most serious illness that seriously impacts human life. The heart does not push blood to other areas of the body in cardiac disease. For the prevention and treatment of cardiac failure, accurate and timely diagnosis of heart disease is critical. The diagnosis of cardiac disease has been considered via conventional medical history. Non-invasive approaches like machine learning are effective and powerful to categorize healthy people and people with heart disease. In the proposed research, by using the cardiovascular disease dataset, we created a machine-learning model to predict cardiac disease. In this paper, it is capable of recognizing and classifying the heart disease patient from healthy people by using three standard machine learning algorithms: Ran-dom Forest (RF), Support Vector Machine (SVM) and K-Nearest Neighbor (KNN). In addition, the ROC/AUC curve is calculated for each classification al-gorithms. In the proposed scheme, we also used the feature selection algorithm to reduce dimensions over a qualified heart disease dataset. After that, the whole structure for the classification of heart disease has been created. On complete features and reduced features, the performance of the proposed approach has been verified. The decrease in features affects the accuracy and time of execution of the classifiers. With the selected features, the highest classification accuracy is obtained for the KNN algorithm is about 93%, with a sensitivity is 0.9750 and specificity is 0.8529. Therefore, with the complete features, the classification ac-curacy is about 91%.