ViSERA: Visualized & Standardized Environment for Radiomics Analysis

Introduction: Objectives: Software-driven innovation can significantly inform and guide novel research directions. This is particularly true in medical imaging where several imaging technology trends meet, including image visualization, image-format conversion, registration, fusion, filtration, segmentation, and information extraction, all converging on the need for versatile and effective user interfaces. Since medical images (e.g. CT, MRI, PET, SPECT) are stored in different formats such as DICOM, NII.GZ, and NRRD, it is imperative to enable simplified imaging format conversions and consistent coordinate registration to collaborate and provide reproducible results. By combining images using different techniques and applying various convolutional filters, researchers can generate powerful qualitative and quantitative results. Meanwhile, radiomics frameworks – involving high-throughput extraction and analysis of quantitative image features – have shown significant potential in quantifying image-region phenotypes towards improved task performances (e.g., diagnosis, prognosis, etc.). However, the lack of standardized and user-friendly environments has hindered the wider usage and deployment of radiomics applications. This software includes Visualized & Standardized Environment for Radiomics Analysis (ViSERA), an open infrastructure modular software that flexibly supports reliable and easy-to-use image visualization, processing, and standardized radiomics analyses.

Methods: ViSERA is a major, entirely-revamped upgrade to the original SERA. SERA was a Matlab-based open-source package that enabled standardized and reproducible radiomic feature extraction in compliance with the Image Biomarker Standardization Initiative (IBSI 1.0) guidelines. Contrariwise, ViSERA is Python-based software that constitutes a graphical user interface (GUI) that allows users to add different image processing algorithms and to simultaneously perform multiple algorithms in the workspace. ViSERA introduces several new components, including image visualization, segmentation, image-format conversion, registration, fusion (e.g. weighted, wavelet, and PCA fusion techniques), and convolutional filters, in addition to radiomic features extraction. Its radiomic features calculations are meticulously consistent with SERA and IBSI 1. Moreover, ViSERA has been standardized against IBSI 2.0 by implementing and validating several filter options such as mean, Laplacian of Gaussian (LoG), Laws, Gabor, and Wavelet. In the image visualization module, ViSERA employs various new practical tools such as crop, manual and automated segmentation (including subtraction, merge, and edit), contrast adjustment, and transformation (including rotation, flip, etc.). Furthermore, each module can be run independently on its own, or modularly connected to other components (i.e. the output of one module is fed as input to another). Module sequences can contain 2 or more steps. The user can easily view the results of the executions directly in the visualization window.

Results: One of the highlighted features of ViSERA is the flexibility of handling multiple different image formats regardless of their resolution, orientation, or imaging modality. This key feature in addition to its user-friendliness makes it accessible for all researchers at different skill levels, including radiation oncologists, radiologists, physicists, and computer and data scientists. Moreover, user friend initialization and connection of different algorithms enable researchers at all levels to more readily pursue individual or collaborative research on medical images.

Conclusions: Conclusion: We introduce ViSERA, a Python-based, entirely-revamped medical imaging software for standardized radiomics analysis that can help researchers and practitioners of varying expertise to more readily perform reliable and reproducible radiomics research. In our future releases, we aim to introduce more modules and algorithms within ViSERA.

• Download figure
• Open in new tab
• Download powerpoint