Research Interests
Academic research and real-world solutions focusing on A.I.
My research interests center on machine learning and deep learning for medical image and text analysis, as well as cybersecurity. My work focuses on robust segmentation, classification, anomaly detection, and adaptive learning systems. I am also interested in large language models, natural language processing, and the development of efficient and interpretable models that address complex real-world challenges in healthcare, cybersecurity, and intelligent systems.
Interested in collaboration, research, or project opportunities? Feel free to reach out.
Research
My research extends across machine learning and deep learning for medical image and text analysis, along with cybersecurity. I focus on segmentation, classification, anomaly detection, and adaptive learning systems designed to perform reliably in complex settings. I am also interested in large language models, natural language processing, and developing efficient, interpretable models for real-world challenges in healthcare, cybersecurity, and intelligent systems.
Dmcie: Diffusion model with concatenation of inputs and errors for enhanced brain tumor segmentation in MRI images
Yavari, S., Pandya, R. N., & Furst, J. (2026). DMCIE: Diffusion model with concatenation of inputs and errors for enhanced brain tumor segmentation in MRI images. International Journal of Computer Assisted Radiology and Surgery, 21(2), 389–398. https://doi.org/10.1007/s11548-025-03532-9
DC-IDS: a dynamic continual learning intrusion detection system with uncertainty-aware adaptation
Yavari, S., Tajzadeh, R., & Furst, J. (2026). DC-IDS: A dynamic continual learning intrusion detection system with uncertainty-aware adaptation. In Assurance and security for AI-enabled systems 2026 (Proceedings of SPIE, Vol. 14046, Article 140460N). SPIE. https://doi.org/10.1117/12.3092731
ReCoSeg: Residual-guided cross-modal diffusion for efficient brain tumor segmentation
Yavari, S., Pandya, R. N., & Furst, J. (2025). ReCoSeg: Residual-guided cross-modal diffusion for efficient brain tumor segmentation. In Medical Imaging with Deep Learning 2025., https://openreview.net/forum?id=rvJhX1firb
ReCoSeg++: Extended residual-guided cross-modal diffusion for brain tumor segmentation
Yavari, S., Pandya, R. N., & Furst, J. (2025). ReCoSeg++: Extended residual-guided cross-modal diffusion for brain tumor segmentation [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2508.01058
Mamba-BTS: Boundary-Aware Mamba Segmentation for Multimodal Brain Tumor MRI
Pandya, R. N., Yavari, S., Furst, J., & Tchoua, R. (2026). Mamba-BTS: Boundary-aware Mamba segmentation for multimodal brain tumor MRI. In Machine learning from challenging data 2026. SPIE Defense + Security. Advance online publication. https://doi.org/10.1117/12.3092622
Codebook-guided LLMs for social ecological coding of breast cancer screening focus groups
Chauhan, S., Abdallah, A., Malaviya, D., Kevadiya, R., Pandya, R., Yavari, S., Wang, T., & Tchoua, R. (2026). Codebook-guided LLMs for social ecological coding of breast cancer screening focus groups. In Proceedings of the 22nd IEEE International Conference on eScience. Accepted paper.
Mitigating catastrophic forgetting in the incremental learning of medical images
Yavari, S., & Furst, J. (2025). Mitigating catastrophic forgetting in the incremental learning of medical images [Preprint]. arXiv. https://doi.org/10.48550/arXiv.2504.20033
Biomechanically regularized deep deformable registration for CT and US fusion
Mohanty, S., Yavari, S., & Dakua, S. P. (2025). Biomechanically regularized deep deformable registration for CT and US fusion. Electronics Letters, 61, Article e70351. https://doi.org/10.1049/ell2.70351
Enhancing online intrusion detection systems via attack clustering
Yavari, S., & Oteafy, S. M. A. (2023). Enhancing online intrusion detection systems via attack clustering. In 2023 IEEE Global Communications Conference (GLOBECOM) (pp. 4650–4655). IEEE. https://doi.org/10.1109/GLOBECOM54140.2023.10437054
Demo: Remote heart assessment using deep learning over IoT phonocardiograms
Yavari, S., & Oteafy, S. M. A. (2023). Demo: Remote heart assessment using deep learning over IoT phonocardiograms. In 2023 IEEE 48th Conference on Local Computer Networks (LCN) (pp. 1–4). IEEE. https://doi.org/10.1109/LCN58197.2023.10223343
Machine-learning techniques to design an intrusion detection system in computer networks
Yavari, S., & Mosavi, S. M. (2023). Machine-learning techniques to design an intrusion detection system in computer networks [Poster presentation]. Women in CyberSecurity Conference (WiCyS 2023).
Security analysis of ultra-wideband communication systems for telemedicine and IoT-based healthcare
Yavari, S. (2018). Security analysis of ultra-wideband communication systems for telemedicine and IoT-based healthcare. Yasin Publishing.
Providing a fuzzy logic control system to reduce the frequency deviation in independent hybrid power system
Yavari, S. (2017). Providing a fuzzy logic control system to reduce the frequency deviation in independent hybrid power system. In 2017 IEEE 4th International Conference on Knowledge-Based Engineering and Innovation (KBEI) (pp. 723–728). IEEE. https://doi.org/10.1109/KBEI.2017.8324892
Teaching
“A teacher affects eternity; he or she can never tell where their influence stops.” -Henry Adams
From ancient scholars to modern university faculty, teachers shape minds and inspire generations. My goal is to join this community of educators and contribute to meaningful learning, discovery, and innovation.
Teaching and Mentoring the next generation of students
I value the opportunity to support students as they develop confidence, critical thinking, and the skills needed to succeed in both research and practice.
My approach to mentoring is grounded in patience, clarity, and encouragement. I aim to create a learning environment where students feel supported, challenged, and inspired to ask questions, explore ideas, and build independence.
Beyond instruction, I see teaching as a lasting contribution to the development of others. Whether in the classroom, in supervision, or in informal academic guidance, I strive to help students realize their potential and take meaningful steps toward their goals.
Projects
I have developed and evaluated machine learning and deep learning solutions across medical imaging and cybersecurity, with a focus on robust predictive and segmentation models. My projects include brain tumor segmentation, CT- and MRI-based cancer classification and segmentation, ophthalmic and prostate image analysis, and intrusion detection and anomaly-detection methods. This work incorporates rigorous data preprocessing, model optimization, and performance evaluation.
Healthcare Text Classification Using BioClinicalBERT
Fine-tuned and evaluated BioClinicalBERT for healthcare text classification, incorporating clinical-text preprocessing and systematic evaluation using precision, recall, F1 score, and confusion-matrix analysis. Synthetic, non-sensitive clinical narratives were used to demonstrate the pipeline without exposing protected health information.
Text-to-Image Generation Using Diffusion Models
Developed an experimental text-conditioned diffusion pipeline to investigate prompt-guided image synthesis through iterative denoising. Implemented model training, text conditioning, sampling, and visualization in PyTorch.
Vision Transformer for Lung Cancer Classification
Adapted and evaluated a Vision Transformer for lung-cancer classification from CT images. The workflow included image preprocessing, model training, and performance analysis using standard classification metrics.
U-Net-Based Medical Image Segmentation
Applied U-Net-based architectures to pixel-level segmentation of ophthalmic abnormalities, breast tumors, and prostate cancer. Addressed challenges including irregular boundaries, small lesions, and class imbalance.
Liver Tumor Segmentation Using Deep-Learning Architectures
Compared U-Net and a segmentation architecture with a ResNet-50 encoder for liver-tumor delineation in CT scans. Evaluated segmentation performance using Dice score and intersection over union.
Machine-Learning-Based Network Intrusion Detection
Applied and compared supervised-learning and anomaly-detection algorithms, including random forest, support vector machine (SVM), Isolation Forest, and autoencoders, to detect malicious network traffic and distinguish it from normal activity. Performed data preprocessing, feature selection, hyperparameter optimization, and cross-validation. Evaluated performance using precision, recall, F1 score, area under the precision–recall curve (AUPRC), and false-positive rate.
Machine-Learning-Based Lung Cancer Risk Prediction
Applied and compared classification algorithms, including random forest, support vector machine (SVM), and multilayer perceptron (MLP), to predict lung cancer risk using patients’ demographic, behavioral, and clinical attributes. Performed data preprocessing, feature selection, class-imbalance handling, hyperparameter optimization, and stratified cross-validation. Evaluated performance using accuracy, precision, recall, F1 score, and AUPRC.