Search

Your search keyword '"ŻOŁEK, Norbert"' showing total 22 results
Start Over Author "ŻOŁEK, Norbert"
22 results on '"ŻOŁEK, Norbert"'

1. Performance of Machine Learning Classification in Mammography Images using BI-RADS

Author

Gunawardhana, Malitha and Zolek, Norbert

Subjects
Electrical Engineering and Systems Science - Image and Video Processing, Computer Science - Computer Vision and Pattern Recognition
Abstract

This research aims to investigate the classification accuracy of various state-of-the-art image classification models across different categories of breast ultrasound images, as defined by the Breast Imaging Reporting and Data System (BI-RADS). To achieve this, we have utilized a comprehensively assembled dataset of 2,945 mammographic images sourced from 1,540 patients. In order to conduct a thorough analysis, we employed six advanced classification architectures, including VGG19 \cite{simonyan2014very}, ResNet50 \cite{he2016deep}, GoogleNet \cite{szegedy2015going}, ConvNext \cite{liu2022convnet}, EfficientNet \cite{tan2019efficientnet}, and Vision Transformers (ViT) \cite{dosovitskiy2020image}, instead of traditional machine learning models. We evaluate models in three different settings: full fine-tuning, linear evaluation and training from scratch. Our findings demonstrate the effectiveness and capability of our Computer-Aided Diagnosis (CAD) system, with a remarkable accuracy of 76.39\% and an F1 score of 67.94\% in the full fine-tuning setting. Our findings indicate the potential for enhanced diagnostic accuracy in the field of breast imaging, providing a solid foundation for future endeavors aiming to improve the precision and reliability of CAD systems in medical imaging.

Published
2023

3. Deep meta-learning for the selection of accurate ultrasound based breast mass classifier

Author

Byra, Michal, Karwat, Piotr, Ryzhankow, Ivan, Komorowski, Piotr, Klimonda, Ziemowit, Fura, Lukasz, Pawlowska, Anna, Zolek, Norbert, and Litniewski, Jerzy

Subjects
Electrical Engineering and Systems Science - Image and Video Processing, Computer Science - Computer Vision and Pattern Recognition, Physics - Medical Physics
Abstract

Standard classification methods based on handcrafted morphological and texture features have achieved good performance in breast mass differentiation in ultrasound (US). In comparison to deep neural networks, commonly perceived as "black-box" models, classical techniques are based on features that have well-understood medical and physical interpretation. However, classifiers based on morphological features commonly underperform in the presence of the shadowing artifact and ill-defined mass borders, while texture based classifiers may fail when the US image is too noisy. Therefore, in practice it would be beneficial to select the classification method based on the appearance of the particular US image. In this work, we develop a deep meta-network that can automatically process input breast mass US images and recommend whether to apply the shape or texture based classifier for the breast mass differentiation. Our preliminary results demonstrate that meta-learning techniques can be used to improve the performance of the standard classifiers based on handcrafted features. With the proposed meta-learning based approach, we achieved the area under the receiver operating characteristic curve of 0.95 and accuracy of 0.91., Comment: Work presented at the 2022 IEEE International Ultrasonics Symposium, submission #2078

Published
2022

14. Mutually Orthogonal Golay Complementary Sequences in Medical Ultrasound Diagnostics. Experimental Study.

Author

TROTS, Ihor, ŻOŁEK, Norbert, TASINKEVYCH, Yurij, and WÓJCIK, Janusz

Subjects
*ULTRASONIC imaging, *IMAGE reconstruction, *SIGNAL-to-noise ratio, *SCANNING systems, *SYNTHETIC apertures
Abstract

The objective of this paper is an experimental study of the most crucial parameters of the received acoustic signals (e.g. signal-to-noise ratio (SNR), side-lobes level (SLL), axial resolution) obtained as a result of simultaneous emission of mutually orthogonal Golay complementary sequences (MOGCS) to demonstrate their feasibility of being used in ultrasound diagnostics. Application of the MOGCS in ultrasound measurements allows the image reconstruction time to be shortened without decreasing the resulting quality of reconstructed images in comparison with regular complementary Golay coded sequences (CGCS). In this paper two sets of 16-bits long MOGCS were implemented in the Verasonics Vantage™ (Verasonics Inc., Kirkland, WA, USA) scanner. Ultrasound data were generated using a perfect reflector, a custom-made nylon wire phantom and tissue mimicking phantom. Parameters of the detected MOGCS echoes like SNR, SLL and axial resolution were determined and compared to that of the standard CGCS and the short two-sine cycles pulse. It was evidenced that applying MOGCS did not compromise the parameters of the separated and compressed echoes in comparison to the other types of transmitted signal - the CGCS and the short pulse. Concretely, both the MOGCS and CGCS yield similar SNR increase in comparison to the short pulse. Almost similar values of the axial resolution estimated at the full width at the half maximum level for all types of the transmitted signals were also obtained. At the same time, using the MOGCS the data acquisition speed can be increased twice in comparison with the CGCS signal. [ABSTRACT FROM AUTHOR]

Published
2022
Full Text
View/download PDF

18. Time-resolved detection of fluorescent light during inflow of ICG to the brain—a methodological study

Author

Milej, Daniel, primary, Gerega, Anna, additional, Żołek, Norbert, additional, Weigl, Wojciech, additional, Kacprzak, Michał, additional, Sawosz, Piotr, additional, Mączewska, Joanna, additional, Fronczewska, Katarzyna, additional, Mayzner-Zawadzka, Ewa, additional, Królicki, Leszek, additional, Maniewski, Roman, additional, and Liebert, Adam, additional

Published
2012
Full Text
View/download PDF

22. Performance assessment of time-domain optical brain imagers, part 2: nEUROPt protocol.

Author

Wabnitz H, Jelzow A, Mazurenka M, Steinkellner O, Macdonald R, Milej D, Żołek N, Kacprzak M, Sawosz P, Maniewski R, Liebert A, Magazov S, Hebden J, Martelli F, Di Ninni P, Zaccanti G, Torricelli A, Contini D, Re R, Zucchelli L, Spinelli L, Cubeddu R, and Pifferi A

Subjects
Equipment Design, Europe, Phantoms, Imaging, Reproducibility of Results, Sensitivity and Specificity, Algorithms, Brain cytology, Equipment Failure Analysis methods, Image Interpretation, Computer-Assisted methods, Microscopy instrumentation, Tomography, Optical instrumentation
Abstract

The nEUROPt protocol is one of two new protocols developed within the European project nEUROPt to characterize the performances of time-domain systems for optical imaging of the brain. It was applied in joint measurement campaigns to compare the various instruments and to assess the impact of technical improvements. This protocol addresses the characteristic of optical brain imaging to detect, localize, and quantify absorption changes in the brain. It was implemented with two types of inhomogeneous liquid phantoms based on Intralipid and India ink with well-defined optical properties. First, small black inclusions were used to mimic localized changes of the absorption coefficient. The position of the inclusions was varied in depth and lateral direction to investigate contrast and spatial resolution. Second, two-layered liquid phantoms with variable absorption coefficients were employed to study the quantification of layer-wide changes and, in particular, to determine depth selectivity, i.e., the ratio of sensitivities for deep and superficial absorption changes. We introduce the tests of the nEUROPt protocol and present examples of results obtained with different instruments and methods of data analysis. This protocol could be a useful step toward performance tests for future standards in diffuse optical imaging.

Published
2014
Full Text
View/download PDF

Catalog

Books, media, physical & digital resources
See catalog results