The Algorithm for Preparing a Set of Data for Teaching Neural Networks on the Example of the Task to Analyze the Radiological Images of Lungs
https://doi.org/10.35596/1729-7648-2023-21-1-66-73
Abstract
The methodology for preparing data for teaching neural networks is considered in solving two problems: checking the modality of computed tomography and checking the modality of radiographic images. The algorithm for preparing data for neural networks training is proposed. The influence of the stages (marking of images, normalization of data, determining the dynamic image range, the composition of the training sample) of the algorithm for the learning result is evaluated. The greatest influence in solving the task of modality verification of modality was the choice of optimal values of the dynamic range. The change in the composition of the training sample made it possible to increase the accuracy of the classification by 0.0073. When solving the task of checking the modality of images of computed tomography, the most impact on the result of the training of the neural network had the stage of data normalization. The assumption is put forward that there are special signs of images of this modality.
Keywords
convolutional neural network,
training data preparation,
normalization of data,
modality of radiological images
Supporting agencies: This work was supported by the Belarusian Republican Foundation for Fundamental Research (project No T22MV-026). The author expresses his gratitude and deep gratitude to Cand. of Sci. Eduard Vitalievich Snezhko; Cand. of Sci., Associate Professor Vassili Alekseevich Kovalev and Cand. of Sci., Associate Professor Pavel Viktorovich Kamlach for valuable advice and comments when working on article.
About the Author
A. A. Kosareva
Belarusian State University of Informatics and Radioelectronics
Belarus
Belarus
Kosareva Aleksandra Andreevna, Assistant at the Electronic Engineering and Technology Department
220013, Minsk, P. Brovka St., 6
+375 17 293-88-60
References
1. Clark K., Vendt B., Smith K., Freymann J., Kirby J., Koppel P., Moore S., Phillips S., Maffitt D., Pringle M., Tarbox L., Prior F. (2013) The Cancer Imaging Archive (TCIA): Maintaining and Operating a Public Information Repository. Journal of Digital Imaging. 26 (6), 1045–1057. https://doi.org/10.1007/s10278-0139622-7.
2. Solem J. E. (2012) Programming Computer Vision with Python. 300. https://www.semanticscholar.org/paper/Programming-Computer-Vision-With-Python-Solem/838f0c5eb0958e88c91ced3e743b951b050a5c79.
3. Newton K. M., Peissig P. L., Kho A. N. [et al.] (2013) Validation of Electronic Medical Record-based Phenotyping Algorithms: Results and Lessons Learned from the eMERGE Network. Journal Am Med Inform Assoc. 20 (e1), e147–e154.
4. Willemink M. J., Koszek W. A., Hardell C. [et al.] (2020) Preparing Medical Imaging Data for Machine Learning. Radiology. 295 (1), 4–15. doi: 10.1148/radiol.2020192224.
5. Murphy A., Baba Y. (2022) Windowing (CT). Radiopaedia. Available: https://radiopaedia.org/articles/52108. (Accessed 4 June 2022).
6. Mingxing Tan, Quoc V. Le (2019) EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks. In Proceedings of International Conference on Machine Learning (ICML).
7. Kosareva А. А., Kamlach P. V., Kovalev V. A., Snezhko E. V., Radzhabov A. G. (2022) Finding a Similar Anatomical Area on the Lung`s CT-Image Using EfficientNet. BIG DATA and Advanced Analysis: Collection of Scientific Articles of the VIII Inter. Scient. and Pract. Conf., Minsk, May 11–12, 2022. 476–482 (in Russian).
8. Bhanja S., Das A. (2018) Impact of Data Normalization on Deep Neural Network for Time Series Forecasting. doi: 10.48550/ARXIV.1812.05519.
9. Miroslava Ivko Jordovic Pavlovic, Katarina Djordjevic, Zarko Cojbasic [et al.] (2021) Influence of Data Scaling and Normalization on Overall Neural Network Performances in Photoacoustics. PREPRINT (Version 1) Available at Research Square, 18 Oct. 2021. https://doi.org/10.21203/rs.3.rs-942309/v1.
10. Ramprasaath R. Selvaraju [et al.] (2019) Grad-CAM: Visual Explanations from Deep Networks via Gradientbased Localization. https://arxiv.org/pdf/1610.02391.pdf.
Review
For citations:
Kosareva A.A. The Algorithm for Preparing a Set of Data for Teaching Neural Networks on the Example of the Task to Analyze the Radiological Images of Lungs. Doklady BGUIR. 2023;21(1):66-73. (In Russ.) https://doi.org/10.35596/1729-7648-2023-21-1-66-73
Views: 434
Email this article 