Ciljevi projekta su razvoj algoritama za:

• (ne)linearno robusno samonadzirano grupiranje podataka,
• robusnu raspodjeli šuma prilagođenu regresiju,
• aproksimacije 3D tenzora pomoću nekoliko tenzorskih modela niskog ranga,
• izdvajanje značajki temeljeno na tenzorskim modelima.

Ciljane primjene navedenih algoritama su:

• segmentacija na razini piksela RGB i hiperspektralne slike patohistoloških rezova
• potiskivanje šuma i sažimanje slike u optičkoj koherentnoj tomografiji (OKT)
• dijagnostika bolesti retine iz OKT slike
• detekciji metabolita indikativnih za dijabetes tipa 2 iz ¹H NMR spektara uzoraka urina

Projektni tim:

dr.sc. Ivica Kopriva, Institut Ruđer Bošković (IRB), Zagreb, znanstveni savjetnik u trajnom zvanju, voditelj projekta

dr.sc. Ivanka Jerić, IRB, znanstvena savjetnica

dr.sc. Marijana Popović Hadžija, IRB, viša znanstvena suradnica, članica

Arijana Pačić, Klinička bolnica Dubrava, Zagreb, specijalist patologije

Irena Veliki Dalić, Klinički bolnički centar Zagreb, specijalist patologije

dr.sc. Damir Seršić, Fakultet elektrotehnike i računarstva, Sveučilište u Zagrebu, profesor

dr.sc. Xinjian Chen, Sveučište Soochow, Suzhou, NR Kina, profesor

dr. sc. Fei Shi, Sveučilište Soochow, Suzhou, NR Kina, izv. profesorica

dr.sc. Dario Sitnik, Tehničko sveučilište Minhen, Njemačka, poslijedoktorski student

mag. inf. educ. Lovro Sindičić, Osnovna škola Antuna Mihanovića, Zagreb, doktorski student

Publikacije

Časopisi:

• I. Kopriva, "Pseudo Labels Approach to Interpretable Self-Guided Subspace Clustering," Pattern Recognition 172 (2026) 112618. https://doi.org/10.1016/j.patcog.2025.112618., IF 7.6, Q1: leading 9% (33/368) in Electrical and Electronic Engineering and leading 13% (26/204) in Computer Science and Artificial Intelligence.
• L. Sindičić, I. Kopriva, "Label-independent hyperparameter-free self-supervised single-view deep subspace clustering," Neurocomputing, Vol. 666, article no. 132260, February 2026. IF 6.5, Q1 (leading 18% in Computer Science and Artificial Intelligence), https://doi.org/10.1016/j.neucom.2025.132260
• I.Kopriva, D. Sitnik, L.-I. Dion-Bertrand, M. M. Periša, A. Pačić, M. Hadžija, M. Popović Hadžija, "A Hyerspectral Imaging Dataset and Methodology for Intraoperative Pixel-wise Classification of Metastatic Colon Cancer in the Liver" Computers in Medicine and Biology, Vol. 196, Part B, September, 2025, 110841, https://doi.org/10.1016/j.compbiomed.2025.110841IF 6.3, Q1 (vodećih 15% u kategoriji Computer Science, Interdisciplinary Applications).  

Konferencije:

• L. Sindičić, I. Kopriva, "Mulilayer Graph Approach to Deep Subspace Clustering," na: MedPRAI24 6th Mediterranean Conference on Pattern Recognition and Artificial Intelligence, October 18-19, 2024, Istinye University, Istanbul, Turkey. F. P. García M´arquez et al. (eds.), Pattern Recognition and Artificial Intelligence, Lecture Notes in Networks and Systems 1393, pp. 159-173, 2025. https://doi.org/10.1007/978-3-031-90893-4_11
• I. Kopriva, "Robust Kernel Sparse Subspace Clustering," 2025 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2025), April 6-11, Hyderabad, India. doi:10.1109/ICASSP49660.2025.10888170 (A1 kategorija).
• I. Kopriva, L.-I. Bertrand-Dion, A. Pačić, M. Milković Periša, M. Popović Hadžija, M. Hadžija. D. Sitnik, "Bright field-dark field hyperspectral imaging for pixel-wise classification of adenocarcinoma of the colon in the liver," na: Medical Imaging 2025: Digital and Computational Pathology, edited by John E. Tomaszewski, Aaron D. Ward, Proc. of SPIE Vol. 13413, 134130Y, doi: 10.1117/12.3039121, San Diego, CA, SAD, 16-20, February, 2025.
• I. Kopriva, D. Sitnik, "MERA tensor network extracted pseudo labels and Grassmann manifolds for unsupervised intraoperative semantic segmentation of hyperspectral images of adenocarcinoma of the colon in the liver," SPIE Medical Imaging 2026: Digital and Computational Pathology, edited by John E. Tomaszewski, Scot Doyle, Proc. of SPIE Vol. 13932-33.