Papers
Published papers and public preprints from ITU BIAI Lab, ordered as the primary section of the site.
A public preprint on prompt-conditioned multimodal quality prediction built on MedSigLIP.
Tolga Demiroglu, Mehmet Ozan Unal, Metin Ertas, Isa Yildirim
A public preprint that unrolls BM3D-style denoising into a trainable network while preserving the logic of non-local collaborative filtering.
Kerem Basim, Mehmet Ozan Unal, Metin Ertas, Isa Yildirim
A public preprint that makes classical total-variation regularization spatially adaptive with a learned lambda map.
Yusuf Talha Basak, Mehmet Ozan Unal, Metin Ertas, Isa Yildirim
A public arXiv preprint studying how large multimodal models can score and explain CT image quality.
Kagan Celik, Mehmet Ozan Unal, Metin Ertas, Isa Yildirim
A public preprint arguing that reconstruction objectives should adapt to the downstream diagnostic task rather than staying fixed.
Necati Sefercioglu, Mehmet Ozan Unal, Metin Ertas, Isa Yildirim
A projection-domain self-supervised reconstruction pipeline that removes the need for paired normal-dose targets.
Mehmet Ozan Unal, Metin Ertas, Isa Yildirim
An IEEE ISBI paper showing that low-dose CT reconstruction can be trained without paired clean targets.
Mehmet Ozan Unal, Metin Ertas, Isa Yildirim
An earlier public preprint extending U-Net style reconstruction into the volumetric domain for low-dose CT.
Doga Gunduzalp, Batuhan Cengiz, Mehmet Ozan Unal, Isa Yildirim
A journal paper that uses a deep generative prior as an unsupervised regularizer for low-dose CT reconstruction.
Mehmet Ozan Unal, Metin Ertas, Isa Yildirim
