Mehmet Yigit Avci

I am currently Biomedical Computing M.Sc. student at Technical University of Munich.

I received my BSc degree in Electrical & Electronics Engineering, at Bogazici University, where I have worked in VAVlab with Prof. Burak Acar on Alzheimer Detection using brain connectomes.

I have been working with Prof. Berkin Bilgic and Prof. Qiyuan Tian at Athinoula A. Martinos Center for Biomedical Imaging. on intersection of diffusion MRI and deep learning.

I also worked as part of COMPAI Lab as an intern under supervision of Prof. Julia Schnabel and Veronika Zimmer, PhD and Cosmin Bercea on unsupervised representation learning for Alzheimer's Disease.

I am currently working on my master thesis with these two great groups on zero-shot self-supervised dMRI reconstruction!

Research Interests

Medical Image Analysis, MRI, Deep Learning, Unsupervised Learning

Education

Msc. in Biomedical Computing, TUM 2022-Present

Bsc. in Electrical and Electronics Engineering, Bogazici University 2017-2022

Updates

September 2022 - Started at TUM for my Master's Degree in Biomedical Computing.

September 2022 - Started working as a Machine Learning Intern at deepc, working on benchmarking radiology AI, detection of body part with deep learning and NLP solutions.

June 2022 - Graduated from Bogazici University with a BSc degree Electrical and Electronics Engineering.

May 2022 - Presented my first abstract in ISMRM 2022.

Projects

	MORPHADE: Deformable Autoencoders for Unsupervised Anomaly Detection: Application to Alzheimer's Disease Mehmet Yigit Avci, Cosmin I. Bercea, Benedikt Wiestler, Emily Chan, Daniel Rueckert, Veronika Zimmer, Julia A. Schnabel preprint very soon! We present MORPHADE, an innovative unsupervised anomaly detection system utilizing a multi-head deformable autoencoder for Alzheimer's Disease detection in 3D T1-weighted brain MRI scans. Contributions include improved pseudo-healthy restoration, superior AD detection, precise localization, and severity assessment aligning with clinical evaluations. Our approach offers insights into disease detection and AD progression, supporting the potential to improve patient outcomes considerably.
	Simulation of acquisition shifts in T2 Flair MR images to stress test AI segmentation networks Christiane Posselt, Mehmet Yigit Avci, Mehmet Yigitsoy, Patrick Schuenke, Christoph Kolbitsch, Tobias Schaeffter, Stefanie Remmele Journal of Medical Imaging, 2024 [journal paper] This work's goal is to provide a simulation framework for routine neuroimaging test data, which allows for "stress testing" of deep segmentation networks against acquisition shifts that commonly occur in clinical practice for T2w FLAIR MRI protocols.
	Validation of deep learning techniques for quality augmentation in diffusion MRI for clinical studies Santiago Aja-Fernandez,...Mehmet Yigit Avci, Zihan Li, Berkin Bilgic, Qiyuan Tian,... Tomasz Pieciak ISMRM 2023, NeuroImage: Clinical, 2023 [journal paper] [QUAD results] Participated in QUAD Challenge in MICCAI 2022 and got the 4th place with combination of methods DUnet and deepDTI.
	DUnet: Quantifying the uncertainty of neural networks using Monte Carlo dropout for safer and more accurate deep learning based quantitative MRI M Yigit Avci, Z Li, Q Fan, S Huang, B Bilgic, Q Tian ISMRM 2022 [code] [arXiv] Dropout is conventionally used during the training phase as regularization method and for quantifying uncertainty in deep learning. We propose to use dropout during training as well as inference steps, and average multiple predictions to improve the accuracy, while reducing and quantifying the uncertainty. With our method, accuracy can be improved significantly compared to network outputs without dropout, especially when the training dataset is small. Moreover, confidence maps are generated which may aid in diagnosis of unseen pathology or artifacts.

Template is from Jon Barron's awesome website.