Working with tumor classification and segmentation
-Developed framework that is used for multifile classification with a lot of parameters (data balancing ratio, data balancer, feature seletor, heatmap parameters, classification methods selection)
**Framework is tested and was used to write article (accepted) for CIBB conference in 2019. Then used to write extended version of the paper.
**It was also used to write another paper for breast cancer (for medical journal, waiting for journal response)
-Working on computer vision tasks
--3D volume segmentation in medicine
---Developing new models based on existing ones as well as creating new models, because 3D volume segmentation is not that popular as 2D
---Segmentation of volumes based on 2D models with further processing by 3D model for better segmentation results
-Image classification (prostate, breast, biopsy) to reduce burden of doctors
--Applying straight transfer learning to existing data
--Developing new models based on TL. E.g. developed new fused model where we have two parallel path that merge later. This increased convergence speed as well as final accuracy scores.