Impact of GAN-based lesion-focused medical image super-resolution on the robustness of radiomic features

Costa de Farias, Erick (author), Di Noia, Christian (author), Han, Changhee (author), Sala, Evis (author), Castelli, Mauro (author), Rundo, Leonardo (author)

Robust machine learning models based on radiomic features might allow for accurate diagnosis, prognosis, and medical decision-making. Unfortunately, the lack of standardized radiomic feature extraction has hampered their clinical use. Since the radiomic features tend to be afected by low voxel statistics in regions of interest, increasing the sample size would improve their robustness in clinical studies. Therefore, we propose a Generative Adversarial Network (GAN)-based lesion-focused framework for Computed Tomography (CT) image Super-Resolution (SR); for the lesion (i.e., cancer) patch-focused training, we incorporate Spatial Pyramid Pooling (SPP) into GAN-Constrained by the Identical, Residual, and Cycle Learning Ensemble (GAN-CIRCLE). At 2x SR, the proposed model achieved better perceptual quality with less blurring than the other considered state-of-the-art SR methods, while producing comparable results at 4x SR. We also evaluated the robustness of our mode's radiomic feature in terms of quantization on a diferent lung cancer CT dataset using Principal Component Analysis (PCA). Intriguingly, the most important radiomic features in our PCA-based analysis were the most robust features extracted on the GAN-super-resolved images. These achievements pave the way for the application of GAN-based image Super-Resolution techniques for studies of radiomics for robust biomarker discovery.

• Language: English
• Work type: Article
• Typology: 1.01 - Original Scientific Article
• Organization: EF - School of Economics and Business
• Publication status: Published
• Publication version: Version of Record
• Year: 2021
• Number of pages: Str. 1-12
• Numbering: Vol. 11, art. 21361
• PID: 20.500.12556/RUL-134022
• UDC: 659.2:004
• ISSN on article: 2045-2322
• DOI: 10.1038/s41598-021-00898-z
• COBISS.SI-ID: 83224579
• Publication date in RUL: 22.12.2021

Record is a part of a journal

• Title: Scientific reports
• Shortened title: Sci. rep.
• Publisher: Nature Publishing Group
• ISSN: 2045-2322
• COBISS.SI-ID: 18727432

Licences

• License: CC BY 4.0, Creative Commons Attribution 4.0 International
• Link: http://creativecommons.org/licenses/by/4.0/
• Description: This is the standard Creative Commons license that gives others maximum freedom to do what they want with the work as long as they credit the author.
• Licensing start date: 01.11.2021