Browsing by Author "Mondol, Chaity"

Now showing 1 - 3 of 3

Automated Breast Tumor Ultrasound Image Segmentation With Hybrid UNet and Classification Using Fine-Tuned CNN Model
(Elsevier, 2023-10-20) Hossain, Shahed; Azam, Sami; Montaha, Sidratul; Karim, Asif; Chowa, Sadia Sultana; Mondol, Chaity; Hasan, Md Zahid; Jonkman, Mirjam
Introduction Breast cancer stands as the second most deadly form of cancer among women worldwide. Early diagnosis and treatment can significantly mitigate mortality rates. Purpose The study aims to classify breast ultrasound images into benign and malignant tumors. This approach involves segmenting the breast's region of interest (ROI) employing an optimized UNet architecture and classifying the ROIs through an optimized shallow CNN model utilizing an ablation study. Method Several image processing techniques are utilized to improve image quality by removing text, artifacts, and speckle noise, and statistical analysis is done to check the enhanced image quality is satisfactory. With the processed dataset, the segmentation of breast tumor ROI is carried out, optimizing the UNet model through an ablation study where the architectural configuration and hyperparameters are altered. After obtaining the tumor ROIs from the fine-tuned UNet model (RKO-UNet), an optimized CNN model is employed to classify the tumor into benign and malignant classes. To enhance the CNN model's performance, an ablation study is conducted, coupled with the integration of an attention unit. The model's performance is further assessed by classifying breast cancer with mammogram images. Result The proposed classification model (RKONet-13) results in an accuracy of 98.41 %. The performance of the proposed model is further compared with five transfer learning models for both pre-segmented and post-segmented datasets. K-fold cross-validation is done to assess the proposed RKONet-13 model's performance stability. Furthermore, the performance of the proposed model is compared with previous literature, where the proposed model outperforms existing methods, demonstrating its effectiveness in breast cancer diagnosis. Lastly, the model demonstrates its robustness for breast cancer classification, delivering an exceptional performance of 96.21 % on a mammogram dataset. Conclusion The efficacy of this study relies on image pre-processing, segmentation with hybrid attention UNet, and classification with fine-tuned robust CNN model. This comprehensive approach aims to determine an effective technique for detecting breast cancer within ultrasound images.
Early Prediction of Chronic Kidney Disease
(Daffodil International University, 22-08-29) Mondol, Chaity; Shamrat, F. M. Javed Mehedi; Hasan, Md. Robiul; Alam, Saidul; Ghosh, Pronab; Tasnim, Zarrin; Ahmed, Kawsar; Bui, Francis M.; Ibrahim, Sobhy M.
Chronic kidney disease (CKD) is one of the most life-threatening disorders. To improve survivability, early discovery and good management are encouraged. In this paper, CKD was diagnosed using multiple optimized neural networks against traditional neural networks on the UCI machine learning dataset, to identify the most efficient model for the task. The study works on the binary classification of CKD from 24 attributes. For classification, optimized CNN (OCNN), ANN (OANN), and LSTM (OLSTM) models were used as well as traditional CNN, ANN, and LSTM models. With various performance matrixes, error measures, loss values, AUC values, and compilation time, the implemented models are compared to identify the most competent model for the classification of CKD. It is observed that, overall, the optimized models have better performance compared to the traditional models. The highest validation accuracy among the tradition models were achieved from CNN with 92.71%, whereas OCNN, OANN, and OLSTM have higher accuracies of 98.75%, 96.25%, and 98.5%, respectively. Additionally, OCNN has the highest AUC score of 0.99 and the lowest compilation time for classification with 0.00447 s, making it the most efficient model for the diagnosis of CKD.
Lung and Colon Cancer Classification Using Medical Imaging Through Transfer Learning
(Daffodil International University, 23-02-12) Fahim, Kayes Uddin; Mondol, Chaity
Carcinomas of the lung and colon are among the most common sources of invasive cancer and are the two most common causes of cancer deaths in worldwide. Detecting and treating cancer at an initial stage can diminish death rates worldwide. At present, transfer learning is the very extensively used, compelling and successful imaging approach for the recognition of lung and colon cancer. As part of this study, 10 CNN architectures are analyzed to find ascertain which model provides the best performance for detecting colon and lung cancer with the minimum amount of data loss and completion time: VGG16, VGG19, MobileNet, MobileNetV2, InceptionV3, ResNet50, ResNet50V2, ResNet101, DenseNet201 and Xception. Performance, data loss, and completion time are compared using an evaluation matrix. MobileNet performs with the highest accuracy, VGG19 performs with the second-highest accuracy, and VGG16 performs with third highest accuracy. The dataset contains 25,000 images. MobileNet achieved the best 99.90% training accuracy, 99.88% validation accuracy, and 99.58% test accuracy. It takes the lowest completion time, 45 seconds per epoch, with 0.3323 % of data loss and gives the highest result with the least epoch. Based on image processing and transfer learning, the recommended method yields the highest accuracy, the least completion time, and the least data loss.