Academic Research
Our work is grounded in published research. Dr. Dario Sitnik's papers, preprints and datasets focus on computational pathology and hyperspectral imaging for intraoperative cancer diagnosis, much of it with co-authors at the Ruđer Bošković Institute in Zagreb.
// Computational pathology
Semantic Segmentation-Driven Image-Level Diagnosis of Liver Cancers in Hematoxylin and Eosin Histopathology Images
I. Kopriva, D. Sitnik, A. Pačić, K. Krstanac, I. Veliki Dalić, M. Popović Hadžija
arXiv:2607.03253, 2026
A semantic segmentation-based framework for image-level diagnosis of liver cancers from whole-slide H&E images, built to cope with variability in specimen preparation, staining and scanning.
HandE-Liver3C
I. Kopriva, D. Sitnik, A. Pačić, K. Krstanac, I. Veliki Dalić, A. Lukač et al.
Ruđer Bošković Institute, FULIR data repository, 2026
A public dataset of H&E-stained liver histopathology images with pixel-level annotations.
LEFM-Nets: Learnable Explicit Feature Map Deep Networks for Segmentation of Histopathological Images of Frozen Sections
D. Sitnik, I. Kopriva
arXiv:2204.06955, 2022
A framework that embeds existing deep networks into a low-dimensional subspace through a learnable explicit feature map layer, for segmenting histopathological images of frozen sections.
Clustering and classification of low-dimensional data in explicit feature map domain: intraoperative pixel-wise diagnosis of adenocarcinoma of a colon in a liver
D. Sitnik, I. Kopriva
arXiv:2203.03636, 2022
Maps low-dimensional data through an approximate explicit feature map so that interpretable linear algorithms perform better on intraoperative pixel-wise diagnosis, while remaining open to incremental learning.
Deep learning methods for segmentation of images of frozen tissue sections
D. Sitnik
University of Zagreb, Faculty of Electrical Engineering and Computing, 2022
Deep learning methods for segmenting images of frozen tissue sections.
A dataset and a methodology for intraoperative computer-aided diagnosis of a metastatic colon cancer in a liver
D. Sitnik, G. Aralica, M. Hadžija, M. P. Hadžija, A. Pačić, M. M. Periša et al.
Biomedical Signal Processing and Control 66, 102402, 2021
Introduces a dataset and a methodology for intraoperative computer-aided diagnosis of metastatic colon cancer in the liver.
Approximate explicit feature map for computational augmentation of RGB images of hematoxylin and eosin stained histopathological specimens
I. Kopriva, D. Sitnik, G. Aralica, A. Pačić, M. P. Hadžija, M. Hadžija
Medical Imaging 2021: Digital Pathology, Proc. SPIE 11603, 174-185, 2021
Uses an approximate explicit feature map to computationally augment RGB images of H&E-stained histopathological specimens.
Transfer learning approach for intraoperative pixel-based diagnosis of colon cancer metastasis in a liver from hematoxylin-eosin stained specimens
D. Sitnik, I. Kopriva, G. Aralica, A. Pačić, M. P. Hadžija, M. Hadžija
Medical Imaging 2020: Digital Pathology, Proc. SPIE 11320, 57-67, 2020
Transfer learning for intraoperative pixel-based diagnosis of colon cancer metastases in the liver from H&E-stained specimens.
Deep learning approaches for intraoperative pixel-based diagnosis of colon cancer metastasis in a liver from phase-contrast images of unstained specimens
D. Sitnik, G. Aralica, A. Pačić, M. P. Hadžija, M. Hadžija, I. Kopriva
Medical Imaging 2020: Digital Pathology, Proc. SPIE 11320, 47-56, 2020
Deep learning for intraoperative pixel-based diagnosis of colon cancer metastases in the liver from phase-contrast images of unstained specimens.
// Hyperspectral imaging
MERA tensor network extracted pseudo-labels and Grassmann manifolds for unsupervised intraoperative semantic segmentation of hyperspectral images of adenocarcinoma of the colon
I. Kopriva, D. Sitnik
Medical Imaging 2026: Digital and Computational Pathology, Proc. SPIE 13932, 231-243, 2026
Unsupervised intraoperative semantic segmentation of hyperspectral images of colon adenocarcinoma, using pseudo-labels extracted with a MERA tensor network and Grassmann manifolds.
A hyperspectral imaging dataset and Grassmann manifold method for intraoperative pixel-wise classification of metastatic colon cancer in the liver
I. Kopriva, D. Sitnik, L. I. Dion-Bertrand, M. M. Periša, A. Pačić, M. Hadžija et al.
Computers in Biology and Medicine 196, 110841, 2025
Introduces a hyperspectral imaging dataset and a Grassmann manifold method for intraoperative pixel-wise classification of metastatic colon cancer in the liver.
Bright-field dark-field hyperspectral imaging for pixel-wise classification of adenocarcinoma of the colon in the liver
I. Kopriva, L. I. Dion-Bertrand, M. M. Periša, M. P. Hadžija et al.
Medical Imaging 2025: Digital and Computational Pathology, Proc. SPIE 13413, 231-239, 2025
Combines bright-field and dark-field hyperspectral imaging for pixel-wise classification of colon adenocarcinoma in the liver.
// Self-supervised learning and subspace clustering
Robust Self-Supervised Convolutional Neural Network for Subspace Clustering and Classification
D. Sitnik, I. Kopriva
arXiv:2004.03375, 2020
A robust formulation of the self-supervised convolutional subspace clustering network that clusters data from nonlinear manifolds, tolerates data corruptions and classifies out-of-sample data.
// Earlier work
Određivanje poze osoba temeljeno na dubokom učenju
D. Sitnik
University of Zagreb, Faculty of Electrical Engineering and Computing, 2018
Human pose estimation based on deep learning.
System for Video Stitching
D. Sitnik
University of Zagreb, Faculty of Electrical Engineering and Computing, 2016
A system for stitching multiple video recordings into one.
Ready to Get Started?
Book a free consultation to discuss your AI project.