منصة إيفاد العلمية

أ. د. م. نسرين سليمان

إرسال رسالة

التخصص: هندسة طبية حيوية

الجامعة: جامعة دمشق

النقاط:

معامل الإنتاج البحثي

الخبرات العلمية

أستاذ مشارك في قسم الهندسة الطبية الحيوية جامعة دمشق
إشراف على طلاب الماجستير والدكتوراه
محكم دولي في العديد من المجلات المفهرسة في سكوبس
أستاذ مشارك في قسم الذكاء الصنعي- الجامعة السورية
إشراف على مشاريع التخرج في قسم الهندسة الطبية الحيوية
محكم للأوراق البحثية في العديد من المؤتمرات الدولية

الأبحاث المنشورة

Deep learning-based predictive model of mRNA vaccine deterioration: An analysis of the stanford COVID-19 mRNA vaccine dataset

المجلة: Baghdad Science Journal

سنة النشر: 2023

تاريخ النشر: 2023-08-30

The emergence of SARS-CoV-2, the virus responsible for the COVID-19 pandemic, has resulted in a global health crisis leading to widespread illness, death, and daily life disruptions. Having a vaccine for COVID-19 is crucial to controlling the spread of the virus which will help to end the pandemic and restore normalcy to society. Messenger RNA (mRNA) molecules vaccine has led the way as the swift vaccine candidate for COVID-19, but it faces key probable restrictions including spontaneous deterioration. To address mRNA degradation issues, Stanford University academics and the Eterna community sponsored a Kaggle competition.This study aims to build a deep learning (DL) model which will predict deterioration rates at each base of the mRNA molecule. A sequence DL model based on a bidirectional gated recurrent unit (GRU) is implemented. The model is applied to the Stanford COVID-19 mRNA vaccine dataset to predict the mRNA sequences deterioration by predicting five reactivity values for every base in the sequence, namely reactivity values, deterioration rates at high pH, at high temperature, at high pH with Magnesium, and at high temperature with Magnesium. The Stanford COVID-19 mRNA vaccine dataset is split into the training set, validation set, and test set. The bidirectional GRU model minimizes the mean column wise root mean squared error (MCRMSE) of deterioration rates at each base of the mRNA sequence molecule with a value of 0.32086 for the test set which outperformed the winning models with a margin of (0.02112). This study would help other researchers better understand how to forecast mRNA sequence molecule properties to develop a stable COVID-19 vaccine.

Semi-automatic detection and segmentation algorithm of saccular aneurysms in 2D cerebral DSA images

المجلة: The Egyptian Journal of Radiology and Nuclear Medicine

سنة النشر: 2016

تاريخ النشر: 2016-09-01

Objective To detect and segment cerebral saccular aneurysms (CSAs) in 2D Digital Subtraction Angiography (DSA) images. Patients and methods Ten patients underwent Intra-arterial DSA procedures. Patients were injected with Iodine-containing radiopaque material. A scheme for semi-automatic detection and segmentation of intracranial aneurysms is proposed in this study. The algorithm consisted of three major image processing stages: image enhancement, image segmentation and image classification. Applied to the 2D Digital Subtraction Angiography (DSA) images, the algorithm was evaluated in 19 scene files to detect 10 CSAs. Results Aneurysms were identified by the proposed detection and segmentation algorithm with 89.47% sensitivity and 80.95% positive predictive value (PPV) after executing the algorithm on 19 DSA images of 10 aneurysms. Results have been verified by specialized radiologists. However, 4 false positive aneurysms were detected when aneurysms’ location is at Anterior Communicating Artery (ACA). Conclusion The suggested algorithm is a promising method for detection and segmentation of saccular aneurysms; it provides a diagnostic tool for CSAs.

Developing an Automated Decision-Supporting System to Diagnose Malaria Parasite from Thin Blood Smear Images Using Deep Neural Networks

المجلة: Tishreen University Journal-Engineering Sciences Series

سنة النشر: 2023

تاريخ النشر: 2023-07-18

Malaria is a life-threatening disease caused by a parasite called Plasmodium, which is transmitted to humans through the bite of infected female Anopheles mosquitoes. The accurate detection of malaria parasite from thin blood smear images is imperative to improve diagnosis. Purpose: This study aims to investigate the use of a deep neural networks for developing automated clinical decision-making system to improve malaria detection and evaluate its performance. Materials and Methods: A deep neural network was proposed and trained on a dataset of microscopic images of thin blood smears to detect the presence of the malaria parasite. The collection of data comprises 27,558 pictures of red blood cells, where the number of infected and uninfected cells is equal. To test the model's effectiveness, a subset of 2000 images was taken, and the accuracy, precision, recall, and f1-score were used as performance indicators. Results: The results showed that the proposed deep neural network achieved an accuracy of 0.96, indicating its effectiveness in detecting the disease. The precision score of 0.98 indicates that the model has a low rate of false positives, while the recall score of 0.947 indicates that it can detect most cases of malaria. The f1-score of 0.96 shows a good balance between precision and recall. Conclusion: The study demonstrates the potential of using a deep neural network for accurate and efficient malaria detection. The high accuracy, precision, and recall scores suggest that the model is effective in detecting the disease and can minimize the risk of misdiagnosis. Therefore, the proposed deep neural network approach can serve as a promising tool for malaria diagnosis and control.

Predicting Type 2 Diabetes Mellitus using Machine Learning Algorithms

المجلة: Tishreen University Journal-Engineering Sciences Series

سنة النشر: 2022

تاريخ النشر: 2022-11-17

Purpose: to build an effective prediction model based on machine learning (ML) algorithms for the risk of type 2 (non-insulin-dependent) Diabetes Mellitus (T2DM). Methods: I developed two machine learning prediction models based on extreme gradient boosting (XGBoost) and logistic regression (LR). To evaluate the ML prediction models I used the Pima Indian Diabetes dataset (PIDD). The dataset is from the National Institute of Diabetes and Digestive and Kidney Diseases and consists of 500 non-diabetic patients and 268 diabetes patients. Results: Models' performance was evaluated using six performance criteria. XGBoost model outperforms the logistic regression. The XGBoost model achieved: area under receiver operating characteristic curve (AUROC)= 85%, sensitivity= 71%, specificity= 81%, accuracy= 77%, precision= 67%, and F1-score= 69% respectively. Conclusion: This study showed that the XGBoost ML algorithm can be applied to predict individuals at high risk of T2DM in the early phase, which has a strong potential to control diabetes mellitus.

Comparative study for automated coronavirus detection in CT images with transfer learning

المجلة: Damascus University Journa

سنة النشر: 2022

تاريخ النشر: 2022-12-06

o design a computer-aided diagnosis system with transfer learning methods to serve as decision support system for automated coronavirus detection in CT images. Methods Four pre-trained deep convolutional neural networks (ResNet-18, SqueezeNet, ShuffleNet, MobileNet-v2) have been investigated to diagnose coronavirus with CT scans. To evaluate the pre-trained deep convolutional neural network, we used the COVID-CT dataset, which contains 349 CT images of COVID-19 from 216 patients, and 397 CT images obtained from non-COVID-19 subjects. Results Considering binary classification performance results, it has been seen that the pre-trained ResNet-18 model provides the highest classification performance (97.0470 ± 5.5466 accuracy, 98.7342 ± 2.1925 sensitivity, 95.1429 ± 9.3460 specificity, and 0.9737± 0.0489 F1-score) among other three used models. Conclusion ResNet-18 model can be employed as a supportive decision-making system to assist radiologists at clinics and hospitals to screen patients swiftly.

Brain Tumor Detection in MR Images Using K Means Clustering Algorithm

المجلة: Damascus University Journa

سنة النشر: 2023

تاريخ النشر: 2023-05-23

Early detection of brain tumors is essential to preserving human life, it is difficult to manually evaluate magnetic resonance images, and due to the enormous advances in computer graphics processing units, this progress can be used to improve the field of healthcare. In this research, an algorithm was proposed to detect brain tumors and determine their location and size. The research algorithm consists of three basic stages, first the preprocessing stage in which the boundaries of the skull were removed and the contrast was improved, then the segmentation stage using the K means clustering algorithm to separate the tumor from the normal tissue and obtain the region of interest that represents the tumor region, and in the next stage, the shape features were extracted to calculate the area of the tumor in percentage and determine its location. This algorithm was applied to 1100 images equally distributed (550 images of a benign brain tumor and 550 images of a malignant brain tumor). The proposed algorithm achieved high performance when evaluating the segmentation stage and the results were as follows: accuracy 98.57%, sensitivity 89.48%, and specificity 99.07%. Keywords: Magnetic Resonance, Region of Interest, Brain Tumors, K Means Clustering, Machine Learning.