@book{MTMT:34562492, title = {Proceedings of the Workshop on the Advances of Information Technology 2024}, url = {https://m2.mtmt.hu/api/publication/34562492}, isbn = {9789634219422}, author = {Kiss, Bálint and Szirmay-Kalos, László}, publisher = {Budapesti Műszaki és Gazdaságtudományi Egyetem, Irányítástechnika és Informatika Tanszék}, unique-id = {34562492}, year = {2024}, orcid-numbers = {Szirmay-Kalos, László/0000-0002-8523-2315} } @article{MTMT:33734694, title = {A linear heuristic for multiple importance sampling}, url = {https://m2.mtmt.hu/api/publication/33734694}, author = {Sbert, Mateu and Szirmay-Kalos, László}, doi = {10.1186/s13634-023-00990-8}, journal-iso = {EURASIP J ADV SIG PR}, journal = {EURASIP JOURNAL ON ADVANCES IN SIGNAL PROCESSING}, volume = {2023}, unique-id = {33734694}, issn = {1687-6172}, year = {2023}, eissn = {1687-6180}, orcid-numbers = {Szirmay-Kalos, László/0000-0002-8523-2315} } @book{MTMT:33624046, title = {Proceedings of the Workshop on the Advances of Information Technology 2023}, url = {https://m2.mtmt.hu/api/publication/33624046}, isbn = {9789634218968}, author = {Kiss, Bálint and Szirmay-Kalos, László}, publisher = {Budapesti Műszaki és Gazdaságtudományi Egyetem, Irányítástechnika és Informatika Tanszék}, unique-id = {33624046}, year = {2023}, orcid-numbers = {Szirmay-Kalos, László/0000-0002-8523-2315} } @article{MTMT:33212860, title = {Direct dynamic tomographic reconstruction without explicit blood input function}, url = {https://m2.mtmt.hu/api/publication/33212860}, author = {Szirmay-Kalos, László and Magdics, Milán and Varnyú, Dóra}, doi = {10.1016/j.bspc.2022.104313}, journal-iso = {BIOMED SIGNAL PROCES}, journal = {BIOMEDICAL SIGNAL PROCESSING AND CONTROL}, volume = {80}, unique-id = {33212860}, issn = {1746-8094}, abstract = {With dynamic positron emission tomography (PET), we study the metabolic processes by monitoring the uptake of a radioactive tracer. The goal is to reconstruct the time-activity curves (TACs) of the voxels, from which the relevant kinetic parameters can be obtained. These curves are assumed to have the algebraic form of a pre-determined kinetic model. Plausible algebraic forms have a limited number of free parameters and thus can be used even for low-statistic measurements. The kinetic model typically involves the amount of radiotracer in the blood, which should be determined either by direct measurement or by extracting it from the PET data using image-derived or model-based methods. However, the direct measurement of the blood concentration is complicated, and the results of the image-derived and the model-based methods are also not reliable because of the partial volume effect and the unknown fraction of blood in the voxels. Moreover, in direct dynamic tomography, the kinetic model is fit from the very beginning of the reconstruction, thus the blood input function is needed even before the image-derived or model-based approaches can provide it. In this paper, we propose a method that is based on compartmental modeling and the Feng blood input function model defined by a fourth-order exponential equation with a pair of repeated eigenvalues, but does not require the blood input function and the fraction of blood parameters explicitly. Thus, the model can be used in cases when we wish to have the robustness and advantages of compartmental modeling, but no blood input function measurement is available. The test results show that the added error in the TACs of not knowing the blood input is reduced in comparison with the spline-based TAC representation. The method can be applied in reference tissue analysis and in image-derived input function approaches.}, year = {2023}, eissn = {1746-8108}, orcid-numbers = {Szirmay-Kalos, László/0000-0002-8523-2315; Magdics, Milán/0000-0003-4298-1022; Varnyú, Dóra/0000-0002-9220-5868} } @{MTMT:33550845, title = {Deep learning based regularization of PET reconstruction}, url = {https://m2.mtmt.hu/api/publication/33550845}, author = {Varnyú, Dóra and Szirmay-Kalos, László}, booktitle = {X. Magyar Számítógépes Grafika és Geometria Konferencia}, unique-id = {33550845}, year = {2022}, pages = {112-121}, orcid-numbers = {Varnyú, Dóra/0000-0002-9220-5868; Szirmay-Kalos, László/0000-0002-8523-2315} } @{MTMT:33550817, title = {Virtual Reality Simulation of Surgical Procedures}, url = {https://m2.mtmt.hu/api/publication/33550817}, author = {Szirmay-Kalos, László and Varnyú, Dóra and Fridvalszky, András Máté and Szécsi, László}, booktitle = {X. Magyar Számítógépes Grafika és Geometria Konferencia}, unique-id = {33550817}, year = {2022}, pages = {51-56}, orcid-numbers = {Szirmay-Kalos, László/0000-0002-8523-2315; Varnyú, Dóra/0000-0002-9220-5868; Fridvalszky, András Máté/0000-0002-5570-1280} } @inproceedings{MTMT:33538636, title = {Probabilistic versus fuzzy segmentation of volumetric medical data}, url = {https://m2.mtmt.hu/api/publication/33538636}, author = {Szirmay-Kalos, László}, booktitle = {Proceedings of the Workshop on the Advances in Information Technology 2022}, unique-id = {33538636}, year = {2022}, pages = {105-110}, orcid-numbers = {Szirmay-Kalos, László/0000-0002-8523-2315} } @inproceedings{MTMT:33538623, title = {Volumetric transfer functions with differential rendering}, url = {https://m2.mtmt.hu/api/publication/33538623}, author = {Csenge, Fábián and Szirmay-Kalos, László}, booktitle = {Proceedings of the Workshop on the Advances in Information Technology 2022}, unique-id = {33538623}, year = {2022}, pages = {34-38}, orcid-numbers = {Szirmay-Kalos, László/0000-0002-8523-2315} } @article{MTMT:33456128, title = {A Comparative Study of Deep Neural Networks for Real-Time Semantic Segmentation during the Transurethral Resection of Bladder Tumors}, url = {https://m2.mtmt.hu/api/publication/33456128}, author = {Varnyú, Dóra and Szirmay-Kalos, László}, doi = {10.3390/diagnostics12112849}, journal-iso = {DIAGNOSTICS}, journal = {DIAGNOSTICS}, volume = {12}, unique-id = {33456128}, issn = {2075-4418}, abstract = {Bladder cancer is a common and often fatal disease. Papillary bladder tumors are well detectable using cystoscopic imaging, but small or flat lesions are frequently overlooked by urologists. However, detection accuracy can be improved if the images from the cystoscope are segmented in real time by a deep neural network (DNN). In this paper, we compare eight state-of-the-art DNNs for the semantic segmentation of white-light cystoscopy images: U-Net, UNet++, MA-Net, LinkNet, FPN, PAN, DeepLabv3, and DeepLabv3+. The evaluation includes per-image classification accuracy, per-pixel localization accuracy, prediction speed, and model size. Results show that the best F-score for bladder cancer (91%), the best segmentation map precision (92.91%), and the lowest size (7.93 MB) are also achieved by the PAN model, while the highest speed (6.73 ms) is obtained by DeepLabv3+. These results indicate better tumor localization accuracy than reported in previous studies. It can be concluded that deep neural networks may be extremely useful in the real-time diagnosis and therapy of bladder cancer, and among the eight investigated models, PAN shows the most promising results.}, year = {2022}, eissn = {2075-4418}, pages = {2849}, orcid-numbers = {Varnyú, Dóra/0000-0002-9220-5868; Szirmay-Kalos, László/0000-0002-8523-2315} } @inproceedings{MTMT:33156891, title = {Blood Input Function Estimation in Positron Emission Tomography with Deep Learning}, url = {https://m2.mtmt.hu/api/publication/33156891}, author = {Varnyú, Dóra and Szirmay-Kalos, László}, booktitle = {2021 IEEE Nuclear Science Symposium and Medical Imaging Conference (NSS/MIC)}, doi = {10.1109/NSS/MIC44867.2021.9875543}, unique-id = {33156891}, abstract = {Dynamic positron emission tomography allows in vivo study of metabolic processes by monitoring the tissue uptake of a radioactive tracer. Its aim is to reconstruct the time-activity curves of the voxels of the measured volume, which are sought in the algebraic form of a predetermined kinetic model. This kinetic model typically depends on the blood input function, which describes the amount of radiotracer in the blood that can be absorbed by tissues. The blood input function can be estimated in parallel with the kinetic parameters of the voxels during the iterative reconstruction process. However, because kinetic parameter calculation requires the blood input even at the start of reconstruction, an initial approximation needs to be given. In this paper, a deep CNN-LSTM-DNN network is proposed to estimate the blood input function from the sinogram data.}, year = {2022}, orcid-numbers = {Varnyú, Dóra/0000-0002-9220-5868; Szirmay-Kalos, László/0000-0002-8523-2315} }