Utku Ozbulak

Adjunct Faculty

Trustworthy AI, Medical imaging, Biomedical imaging

Utku Ozbulak is an Adjunct Professor at George Mason University Korea and a Research Professor at Ghent University Global Campus, specializing in AI for medical and biomedical imaging, as well as explainable and trustworthy AI. He has authored multiple first- and last-author peer-reviewed research papers in top-tier journals and international conferences. His open-source contributions have earned nearly 10,000 GitHub stars and are widely adopted in both academia and industry. He leads international collaborations across Europe and Asia focused on applied AI in areas such as pharmaceutical quality control, surgical video analysis, trustworthy medical imaging, and genomics. In addition to his research, he teaches courses on deep learning, data visualization, and bioinformatics, and actively mentors both undergraduate and graduate students.

Current Research

Trustworthy AI, Medical imaging, Biomedical imaging, AI-assisted pharmaceutical quality control

Selected Publications

As first author

Mutate and Observe: Utilizing Deep Neural Networks to Investigate the Impact of Mutations on Translation Initiation
U. Ozbulak, H.J. Lee, J. Zuallaert, W. De Neve, S. Depuydt, J. Vankerschaver
2023, Q1 SCIE, Bioinformatics, Oxford Press

Know Your Self-Supervised Learning: A Survey on Image-Based Generative and Discriminative Training
U. Ozbulak, H.J. Lee, B. Boga, E.T. Anzaku, H. Park, A. Van Messem, W. De Neve, J. Vankerschaver
2023, SCOPUS, Transactions on Machine Learning Research

Investigating the Significance of Adversarial Attacks and Their Relation to Interpretability for Radar-Based Human Activity Recognition
U. Ozbulak, B. Vandersmissen, A. Jalalvand, I. Couckuyt, A. Van Messem, W. De Neve
2020, Q1 SCIE, Computer Vision and Image Understanding, Elsevier

As last author

SpurBreast: A Curated Dataset for Investigating Spurious Correlations in Real-World Breast MRI Classification
J.B. Won, W. De Neve, J. Vankerschaver, U. Ozbulak
2025, MICCAI, Main track – Early accept

Evaluating Visual Explanations of Attention Maps for Transformer-Based Medical Imaging
M. Chung, J.B. Won, G. Kim, Y. Kim, U. Ozbulak
2024, MICCAI, Workshop track, Oral presentation

Expanded Publication List

First Author Journal Publications

Assessing the Reliability of Point Mutation as Data Augmentation for Deep Learning With Genomic Data
H.J. Lee*, U. Ozbulak*, H. Park, S. Depuydt, W. De Neve, J. Vankerschaver
2024, Q1 SCIE, BMC Bioinformatics, Springer Nature

Perturbation Analysis of Gradient-Based Adversarial Attacks
U. Ozbulak, M. Gasparyan, W. De Neve, A. Van Messem
2020, Q1 SCIE, Pattern Recognition Letters, Elsevier

First Author Conference Proceedings

Revisiting the Evaluation Bias Introduced by Frame Sampling Strategies in Surgical Video Segmentation Using SAM2
U. Ozbulak, S.A. Mousavi, F. Tozzi, N. Rashidian, W. Willaert, W. De Neve, J. Vankerschaver
2025, MICCAI, Workshop track, Oral presentation

Self-Supervised Benchmark Lottery on ImageNet: Do Marginal Improvements Translate to Improvements on Similar Datasets?
U. Ozbulak, E.T. Anzaku, S. Kang, W. De Neve, J. Vankerschaver
2024, IJCNN, Main track, Oral presentation

Utilizing Mutations to Evaluate Interpretability of Neural Networks on Genomic Data
U. Ozbulak, S. Kang, J. Zuallaert, S. Depuydt, J. Vankerschaver
2022, NeurIPS, Workshop track

Evaluating Adversarial Attacks on ImageNet: A Reality Check on Misclassification Classes
U. Ozbulak, M. Pintor, A. Van Messem, W. De Neve
2021, NeurIPS, Workshop track

Selection of Source Images Heavily Influences the Effectiveness of Adversarial Attacks
U. Ozbulak, E.T. Anzaku, W. De Neve, A. Van Messem
2021, BMVC, Main track, Oral presentation

Regional Image Perturbation Reduces Lp Norms of Adversarial Examples While Maintaining Model-to-Model Transferability
U. Ozbulak, J. Peck, W. De Neve, B. Goossens, Y. Saeys, A. Van Messem
2020, ICML, Workshop track

Impact of Adversarial Examples on Deep Learning Models for Biomedical Image Segmentation
U. Ozbulak, A. Van Messem, W. De Neve
2019, MICCAI, Main track

Not All Adversarial Examples Require a Complex Defense: Identifying Over-Optimized Adversarial Examples with IQR-Based Logit Thresholding
U. Ozbulak, A. Van Messem, W. De Neve
2019, IJCNN, Main track, Oral presentation

How the Softmax Output Is Misleading for Evaluating the Strength of Adversarial Examples
U. Ozbulak, W. De Neve, A. Van Messem
2018, NeurIPS, Workshop track

Last Author Conference Proceedings

When Tracking Fails: Analyzing Failure Modes of SAM2 for Point-Based Tracking in Surgical Videos
W. Jang, J. Im, J. Choi, U. Ozbulak
2025, MICCAI, Workshop track

Towards Affordable Tumor Segmentation and Visualization for 3D Breast MRI Using SAM2
S. Kang, E. Kim, J. Vankerschaver, U. Ozbulak
2025, MICCAI, Workshop track

Color Flow Imaging Microscopy Improves Identification of Stress Sources of Protein Aggregates in Biopharmaceuticals
M. Cohrs, S. Koak, Y. Lee, Y.J. Sung, W. De Neve, H.L. Svilenov, U. Ozbulak
2024, MICCAI, Workshop track

Exploring Patient Data Requirements in Training Effective AI Models for MRI-Based Breast Cancer Classification
S. Kang, W. De Neve, F. Rameau, U. Ozbulak
2024, MICCAI, Workshop track

Evaluating Visual Explanations of Attention Maps for Transformer-Based Medical Imaging
M. Chung, J.B. Won, G. Kim, Y. Kim, U. Ozbulak
2024, MICCAI, Workshop track, Oral presentation

Identifying Critical Tokens for Accurate Predictions in Transformer-Based Medical Imaging Models
S. Kang, J. Vankerschaver, U. Ozbulak
2024, MICCAI, Workshop track

Other Publications and Preprints

Improved Sub-Visible Particle Classification in Flow Imaging Microscopy via Generative AI-Based Image Synthesis
U. Ozbulak, M. Cohrs, H.L. Svilenov, J. Vankerschaver, W. De Neve
2025, arXiv preprint

One Patient's Annotation is Another One’s Initialization: Towards Zero-Shot Surgical Video Segmentation with Cross-Patient Initialization
S. Mousavi, U. Ozbulak, F. Tozzi, N. Rashidian, W. Willaert, J. Vankerschaver, W. De Neve
2025, arXiv preprint

Less is More? Revisiting the Importance of Frame Rate in Real-Time Zero-Shot Surgical Video Segmentation
U. Ozbulak, S. Mousavi, F. Tozzi, N. Rashidian, W. Willaert, W. De Neve, J. Vankerschaver
2025, arXiv preprint

Balancing Redundancy and Diversity: An In-Depth Analysis of Active Learning for Laparoscopic Video Segmentation
S.A. Mousavi, E.T. Anzaku, U. Ozbulak, R. De Muynck, F. Tozzi, N. Rashidian, W. Willaert, W. De Neve
2025, MICCAI, Workshop track, Oral presentation

BRCA Gene Mutations in dbSNP: A Visual Exploration of Genetic Variants
W. Jang, S. Koak, J. Im, U. Ozbulak, J. Vankerschaver
2023, arXiv preprint

Tryp: a dataset of microscopy images of unstained thick blood smears for trypanosome detection
E.T. Anzaku, M.A. Mohammed, U. Ozbulak, J. Won, H. Hong, J. Krishnamoorthy, S. Van Hoecke, S. Magez, A. Van Messem, W. De Neve
2023, Nature Scientific Data, volume 10, article 716

Exact Feature Collisions in Neural Networks
U. Ozbulak, M. Gasparyan, S. Rao, W. De Neve, A. Van Messem
2022, arXiv preprint

Prevalence of Adversarial Examples in Neural Networks: Attacks, Defenses, and Opportunities
U. Ozbulak
2022, Ghent University PhD thesis

Automatic detection of Trypanosomosis in thick blood smears using image pre-processing and deep learning
T. Jung, E.T. Anzaku, U. Ozbulak, S. Magez, A. Van Messem, W. De Neve
2021, IHCI 2020 – Lecture Notes in Computer Science

Courses Taught

CDS 301/501: Scientific Information and Data Visualization

Education

2022 - PhD in Computer Science, Ghent University, Belgium

2017 - MSc in Data Science, University of Southampton, United Kingdom

2014 - BSc in Computer Engineering, Yasar University, Turkey