Looking for Consortium: Dr. Tuğçe Öznacar - Asst. Professor of Biostatistics

Research Focus

I work at the intersection of clinical biostatistics, machine learning and personalised medicine, focusing on predictive modelling for early diagnosis, prognosis estimation and treatment stratification. My research integrates structured clinical data, biomedical indicators and medical imaging to develop explainable, fair and clinically deployable AI-driven decision-support systems, particularly for high-risk clinical conditions.

My ongoing focus includes bridging explainability outputs with clinical decision-support workflows, ensuring that AI reasoning becomes accessible, interpretable and clinically actionable rather than purely algorithmic.

Methodological Expertise

• Machine Learning & Statistical Modelling (Random Forest, XGBoost, LightGBM, SVM, Logistic & Survival analysis)

• Deep Learning for Medical Data (CNNs, MobileNetV2, ResNet, transfer learning, multimodal fusion pipelines)

• Hybrid & Ensemble Approaches (stacking architectures, weighted model fusion, multimodal risk stratification)

• Optimisation & Imbalanced Learning (Bayesian optimisation, GridSearchCV, cost-sensitive learning, SMOTE, ADASYN, threshold optimisation, Memetic optimisation)

• Explainable AI (SHAP, LIME, ICE curves, Grad-CAM for imaging, counterfactual interpretation)

• Validation, Generalisability & Deployment Strategy (internal/external validation, calibration, subgroup performance testing, fairness evaluation, reproducibility standards, TRIPOD-AI / CONSORT-AI alignment)

Clinical Domains of Application

• Oncology (lung, breast, ovarian cancer, brain metastasis survival modelling)

• Cardiometabolic diseases

• Maternal health and obstetric complication prediction

• Multimodal deep learning–based medical imaging classification

Selected Publications

• Öznacar T. et al. (2025). Advanced skin cancer prediction using MobileNetV2 deep learning and optimisation techniques. Scientific Reports (Q1).

• Öznacar T. et al. (2025). Prediction of early diagnosis in ovarian cancer patients using ML with Boruta feature selection. Life (Q1).

• Öznacar T. et al. (2025). Survival prediction in brain metastasis using hybrid machine learning approaches. Brain Sciences (Q3).

• Öznacar T. et al. (2024). A machine learning approach to early detection and malignancy prediction in breast cancer. IJCESEN.

• Öznacar T. et al. (2024). Heart failure prediction using SHAP, LIME and ICE for interpretability evaluation. IJCESEN.

Clinical Domains of Application

• Oncology (lung, breast, ovarian cancer, brain metastasis survival modelling)

• Cardiometabolic diseases

• Maternal health and obstetric complication prediction

• Multimodal deep learning–based medical imaging classification

Selected Publications

• Öznacar T. et al. (2025). Advanced skin cancer prediction using MobileNetV2 deep learning and optimisation techniques. Scientific Reports (Q1).

• Öznacar T. et al. (2025). Prediction of early diagnosis in ovarian cancer patients using ML with Boruta feature selection. Life (Q1).

• Öznacar T. et al. (2025). Survival prediction in brain metastasis using hybrid machine learning approaches. Brain Sciences (Q3).

• Öznacar T. et al. (2024). A machine learning approach to early detection and malignancy prediction in breast cancer. IJCESEN.

• Öznacar T. et al. (2024). Heart failure prediction using SHAP, LIME and ICE for interpretability evaluation. IJCESEN.

Proposed Role

Work Package Leader – AI Modelling, Multimodal Deep Learning, Explainability & Validation

Collaboration & Infrastructure Potential

Data access can be established through hospital networks and ethics processes when required, and methodological capacity is ready for deployment once a consortium dataset is defined.

Contact Details

• Email: yasin.tekin@ankaramedipol.edu.tr - tto@ankaramedipol.edu.tr

• ORCID: https://orcid.org/0000-0003-0364-0401