01Signal information
Identify patterns connected to tissue contact and layer transitions.
Krakow, Poland / Computer Science / Experimental AI
Oğuzhan Berke Özdil
PhD researcher at AGH University of Krakow
I design experiments and machine-learning studies for vibroacoustic needle guidance. My work begins with the physical setup, continues through signal processing and ends with validation that reflects how the data were collected.
Current research
Vibroacoustic tissue fingerprinting for needle procedures
Needle contact creates vibrations that travel through the instrument. My PhD investigates whether those signals can identify tissue and layer transitions when the tool, operator, session and sensor conditions change.
Official doctoral topic / ZB 0229/26
Vibroacoustic Tissue Fingerprinting using Machine and Deep Learning for application in Robotic Assisted Needle and Therapy Procedures
Supervised by Prof. Michael Friebe, PhD at the AGH Faculty of Computer Science.
Open the official AGH topic
01
02Generalisation
Test whether those patterns remain useful across tool, operator, session and sensor changes.
03Uncertainty
Recognise unfamiliar conditions instead of returning unsupported confidence.
Projects and laboratory work
Evidence from the systems I helped build and test
These projects shaped how I work: start from the physical system, document how data are produced, and check the result against the experiment.
VIBRONAV research line
VIBRONAV studies signals produced when a medical tool interacts with tissue. The current INNOGLOBO project is developing a wireless clip-on sensing system for real-time support in minimally invasive procedures.
My contribution
I entered the team through website and software work, then moved into experimental setup, manual and robot-assisted acquisition, quality review, and Python/API automation for repeatable Dobot MG400 puncture sequences at AGH.
UltraClear
UltraClear developed a handheld hybrid SPECT and ultrasound imaging system that combines anatomical ultrasound with functional gamma information. The project focused on a portable probe, image fusion and prototype validation.
My contribution
My work covered prototype testing, signal and data software, 3D components, and hardware-software integration. The project is complete and is not presented as an active PhD commitment.
Academic and research path
From software work to independent experimental research
The sequence matters because each stage added a part of the research process I now use in my PhD.
- 01March 2024
Website and software entry
I joined VIBRONAV while studying for my Bachelor's degree and first worked on the research website and supporting software.
- 022024
Experimental measurements
Laboratory work expanded my role into setup preparation, manual measurements, signal acquisition and data-quality review.
- 032024-2025
Dobot MG400 automation
I developed the Python/API workflow for calibration, puncture sequences and acquisition integration. This work became my Bachelor's thesis.
- 042025-2026
Independent Master research
I designed and ran the needle study, prepared and annotated the dataset, compared machine-learning approaches, and used recording-grouped validation. The work led to my first-author BMT contribution.
- 052026
Doctoral research
The PhD now asks whether tissue information remains reliable when experimental conditions change.
Education at AGH University of Krakow
BSc in Computer Science
Faculty of Electrical Engineering, Automatics, Computer Science and Biomedical Engineering, AGH University of Krakow
MSc in Computer Science and Intelligent Systems
Artificial Intelligence and Data Analysis, AGH University of Krakow, final average 4.55/5
PhD in Information and Communication Technology
Faculty of Computer Science (Wydział Informatyki), AGH University of Krakow
Selected publications
Four studies and my contribution to each one
The roles below separate collaborative experimental work from the study I led as first and corresponding author.
Comparative Vibroacoustic Analysis of Spinal Needle Insertion Actions: Quincke vs. Sprotte
First and corresponding author. I led the experiments, dataset, machine learning, validation and manuscript.
Vibroacoustic signatures: proof of concept for simple material characterization during needle interventions
I supported the setup and selected measurements, then reviewed the acquired data and results with the team.
Development of a Test Environment for the Validation of Needle-Mounted Vibroacoustic Sensors
I contributed to prototype-test setup, repeated measurements, data acquisition and signal review.
Initial findings creating a temperature prediction model using vibroacoustic signals originating from tissue needle interactions
I helped develop the setup, acquired audio and coordinated data preparation and annotation. Another author developed the model.
Work with students
Research and thesis collaboration
BSc and MSc students interested in experiments, signal processing or machine learning are welcome to get in touch. You do not need a finished proposal.
The ideas below are starting points, not fixed assignments. We can narrow the scope to your background and the data or equipment available. I prefer direct, relaxed collaboration where questions are welcome.
Formal supervision
I can support the day-to-day research. Formal supervision, topic approval and availability are handled through Prof. Michael Friebe, PhD and AGH.
View the research group student opportunitiesPossible thesis directions
Each idea addresses one part of the research. The exact question, dataset and methods would be agreed before a thesis begins.
BSc or MSc
Lightweight models for vibroacoustic tissue classification
Compare acoustic features and compact machine-learning models for tissue classification. The focus is on careful data splits, reproducible results and practical inference costs.
MSc
Detecting tissue-layer transitions from needle signals
Investigate when a needle crosses a boundary between layers in a controlled phantom. Combine signal analysis with a clearly defined timing or position reference.
MSc
Testing models across changing experimental conditions
Measure what happens when a model encounters a different needle, insertion speed or recording session. Start with a small, controlled comparison and held-out conditions.
MSc
Understanding what a vibroacoustic model uses
Examine which parts of a signal influence a prediction, then test whether those explanations are stable. The goal is to understand model behaviour, not claim biological causality.
About and contact
About me
I am a Turkish researcher living in Krakow. Alongside my PhD, I work as a ServiceNow Developer at EY GDS Poland and hold professional ServiceNow certifications. My software background includes Python, automation systems, REST APIs, React, TypeScript, Java and Spring. I worked more extensively with C# before moving to Java for work. My research also uses signal processing, machine learning and deep learning for experimental data.
Research gives that software work a physical setting. I enjoy building the experiment, seeing how the signal was produced, and checking whether the analysis still makes sense outside one convenient split of the data.
Away from research, I am a devoted Lord of the Rings fan and enjoy fantasy and history books. I have followed PC games since childhood, although there is much less time to play now. Strength training and fitness have been part of my life for 10 years, including periods of coaching and competition. I also enjoy cycling trips, hiking, climbing, canoeing and more.
I used to dance and still enjoy salsa, bachata, tango and modern dance. I am easygoing, enjoy conversation and friendly debate, and am usually happy to discuss almost any subject. I bring the same open atmosphere to day-to-day research collaboration while keeping the work itself clear and serious.
Contact
Get in touch
For research or a thesis idea, my AGH email is the best place to start.
AGH emailozdil@agh.edu.plPersonal emailozdiloguzhanberke@gmail.comLinkedInoguzhanberkeozdilGitHubOguzhanBerkeOzdilGoogle ScholarOğuzhan Berke ÖzdilORCID0009-0004-6967-0587Instagram@berkeozdilFacebookoguzhanberke.ozdil