Colon Cancer Detection & Segmentation

Colon cancer detection and segmentation formed the focus of an interdisciplinary medical AI initiative developed during my research work at the Faculty of Mechanical Engineering, University of Belgrade, in collaboration with clinical experts associated with the First Surgical Clinic in Belgrade.

The initiative brought together engineering and medical expertise with the aim of exploring AI-assisted analysis of colorectal cancer imaging data, with the late Academician Zoran Krivokapić, one of Serbia’s most distinguished colorectal surgeons, providing an important connection to the clinical domain.

My work focused primarily on designing the data and annotation infrastructure required for developing machine-learning models from real clinical data. This included defining the complete data preparation protocol, establishing anonymization and processing workflows, specifying dataset structure and formats, and deploying annotation tools configured specifically for experienced radiologists.

A dedicated annotation environment was prepared with predefined labeling conventions and presets in order to make expert annotation consistent, reproducible, and suitable for subsequent computer vision research.

The resulting pipeline enabled the creation of an anonymized medical imaging dataset comprising data from nearly 100 patients, structured from the outset for machine-learning experimentation and the development of detection and segmentation models.

The project provided an opportunity to apply and demonstrate the expertise developed through my doctoral research, particularly in medical image analysis, dataset design, annotation methodology, preprocessing pipelines, and the preparation of clinical imaging data for deep learning-based detection and segmentation.