Anna Kuzina

• Developed a MONAI/PyTorch pipeline for liver and hepatic lesion segmentation from multiphase CT, covering curation, preprocessing, multiphase input handling, training and evaluation.
• Investigated the persistent failure to learn the lesion class and jointly reworked the overlapping-label representation, loss and class-aware patch sampling, establishing a trainable lesion baseline.
• Replaced single-split evaluation with class-stratified five-fold cross-validation, keeping all phases of each case in the same fold; analysed per-case errors and identified an association between lesion volume and segmentation performance.
• Implemented reproducible cluster training with GitLab CI, SLURM and Singularity, versioned configurations and experiment tracking.

• First author of a conference abstract on hormonal modulation of immune gene expression in HIV-1 infection. Performed RT-qPCR experiments and quantitative analysis contributing to three co-authored peer-reviewed publications.
• Quantified CCR5/CCR8 and TLR2/4/9 gene expression by RT-qPCR in PBMCs from multiple donors across three HIV-1 subtypes (A6, B, G) and varying hormone concentrations; analysed differences across experimental conditions and donor-to-donor variation.
• First author of conference abstracts on innate immune gene expression in oral mucosal repair and in a bacteriophage-treated Pseudomonas aeruginosa wound-infection model.

Apoptosis-regulator gene expression related to spatial-memory performance (Morris water maze); owned the RT-qPCR workflow end to end.