Publications
An investigation of machine learning algorithms and data augmentation techniques for diabetes diagnosis using class imbalanced BRFSS dataset
Authors
Publication Date
2024
Journal
Healthcare Analytics
Volume
5
Pages
100297
Publisher
Elsevier
Description
This study investigates machine learning algorithms for diabetes diagnosis on the imbalanced BRFSS dataset. To address the challenges of class imbalance, we evaluate oversampling (SMOTE-N), undersampling (ENN), and hybrid methods (SMOTE-Tomek, SMOTE-ENN) before classification. The results highlight the importance of carefully applying augmentation techniques to improve prediction accuracy without data leakage, offering a practical pipeline for healthcare practitioners. read more
Brain Tumor Segmentation and Classification using Spatial Fuzzy C mean and Quadratic Support Vector Machine
Authors
Ragib Shahariar Ayon, Jannatul Robaiat Mou, Sharafat Hossain Majed, Rathyatul Rifat.
Publication Date
2019/12/26
Conference
2019 3rd International Conference on Electrical, Computer & Telecommunication Engineering (ICECTE)
Pages
233-236
Publisher
IEEE
Description
Precise detection and classification of brain tumors at a reasonable time is vital to enhance the survival of a patient. But due to the big dimensional and anatomical diversity between brain tumors, this is a difficult job. In this contribution, we are proposing an improved spatial fuzzy c-mean (SFCM) for brain tumor segmentation and a quadratic support vector machine (QSVM) for tumor type classification. When applied to the BraTS 2018 dataset, the acquired segmentation outcomes were shown to be 7.16% and 12.78% better than the conventional fuzzy c-mean (FCM) and k-mean. This method assists radiologists in understanding the medical image and gives a second opinion. read more
A new approach of moving object detection using background subtraction method
Authors
Ragib Shahariar Ayon, Rathyatul Rifat, Jannatul Robaiat Mou, Abid Ahsan
Publication Date
2019/12/26
Conference
2019 3rd International Conference on Electrical, Computer & Telecommunication Engineering (ICECTE)
Pages
256-259
Publisher
IEEE
Description
Moving object detection has introduced a new challenging horizon in the era of image processing. The main challenge to detect moving objects in all environments because of the complex background. This paper proposes a method that detects a moving object more accurately in almost all the environments. Background subtraction is used in our proposed method for detecting moving objects where the Sauvola algorithm is used on a subtracted image for binary conversion. Median filtering and boundary labeling are used to increase the accuracy of moving object detection. Finally, a morphological operation is performed in our proposed method for detecting the moving object. We tested our proposed method on Wallflower datasets and compare the results of moving object detection on the basis of recall, precision and F-measure with other existing methods. read more