Basics of Image Processing: The Facts and Challenges of Data Harmonization to Improve Radiomics Reproducibility

ISBN-13: 9783031484452 / Angielski

?ngel Alberich-Bayarri; Fuensanta Bellv?s Bataller

Spis treści

1. Era of AI quantitative imaging

Luis Marti-Bonmati (La Fe, Valencia, Spain)

Transforming clinical care from qualitative to quantitative, using automated methods capable of quantifying image features closely related to measurable clinical endpoints. Definition of the different AI-based methods that are currently gold-standard for imaging biomarkers and in which directions they are evolving.

2. Principles of image formation in the different modalities

Mauro Iori (Reggio Emilia)

Description of the image acquisition and reconstruction, explaining the components of the equipment, the physical process underlying image formation and the different types of reconstruction algorithms. This chapter will cover all three imaging modalities: ionising radiation (CT), non-ionising radiation (MRI) and nuclear medicine (mainly PET).

3. How to extract radiomic features from the image?

Ana Jiménez, Bas Hulsken (Quibim)

Definition and extraction of image features, classification of the quantitative features (shape features, first-order statistic features, second-order statistic, higher-order statistics features) and standardization of radiomic features.

4. Facts and needs to improve Radiomics reproducibility

Peter van Ooijen (University Medical Center Groningen)

The reliability of radiomics features (RFs) is crucial for quantifying the ROIs of an image. This chapter will focus on defining the confounding factors that could lead to inaccurate and unreliable radiomics feature estimation, explaining how the inconsistencies that exist across CT scanners and MRI sequences may decrease the reliability of image-derived RFs. Over the needs that underlie the process of evaluating these factors and reducing their influence as far as possible.

5. What is harmonisation and how does it differ from standardisation?

Guang Yang (Imperial College London)

Definition of the harmonization process in the field of medical imaging, categorization of the methods under the image domain and the feature domain and to address the advantages, disadvantages, and challenges posed by these harmonization methods. In addition, the difference with standardization, a term that is often used interchangeably, although they are different concepts, will be pointed out.

6. Harmonisation in the image domain

Sotirios Bisdas (University College London)

Harmonization methods performed on the whole image (raw or reconstructed) before feature extraction and thus aim to harmonize images acquired across different centers/scanners/protocols. In this chapter, a briefly review of the methods in such a way can be applied at every stage of medical image processing from image acquisition to image analysis.

7. Harmonisation across MRI

Fabio García, Fuensanta Bellvís, Ángel Alberich (Quibim)

Because structural MRIs are acquired in arbitrary units, the intensity harmonization is crucial to enable the comparability of examinations acquired with different MR-systems, coils, and acquisition parameters. Some intensity harmonization techniques (IHTs) that can make the MRI standardized and comparable will be explained. Furthermore, given that IHT does not remove the scanner effects in the radiomic feature level, other harmonization methods will be evaluated classifying them into two classes, namely (i) harmonizing the MRI images before the feature extraction and (ii) harmonizing the extracted radiomic features (e.g., ComBat method).

8. Harmonisation in the features domain

José Lozano, Fuensanta Bellvís (Quibim)

The methods categorized under the feature domain are performed after (or within) feature extraction and aim to harmonize extracted radiomic features. In this chapter, the methods will be listed. Two different approaches will be reviewed: identification of reproducible features (a convenient approach) and normalization techniques (statistical approaches). The normalization techniques are further divided into basic statistical normalization (rescaling/standardization); intensity harmonization techniques; ComBat method and its derivatives; normalization using Deep Learning (DL).

9. Selection of the optimal harmonisation method(s) for the problem under study

Emanuele Neri (University of Pisa)

There are multiple harmonization methods that can be applied at different stages of the process: before feature extraction or after the features are extracted. The evaluation of the method or methods to be applied is an important challenge that will be evaluated in this chapter, analysing the problem to be solved according to the image modality, the objectives to be achieved and the predictive or prognostic model to be developed from the extracted radiomics features, etc. An appropriate decision tree will be defined to act as a guide for real problems.

10. Conclusions

Ángel Alberich, Fuensanta Bellvís (Quibim)

This chapter will compile the main conclusions obtained in the previous 9 chapters, showing the advantages and disadvantages of the different harmonization methods and will show the questions that have been answered from this book. Finally, the future tasks to be developed will be shown, aiming to create quantification methods as robust, validated and standardized as possible.

O autorze

Ángel Alberich-Bayarri. Telecommunications Engineer with specialisation in electronics by the Technical University of Valencia and PhD in Biomedical Engineering for his research on the application of advanced image processing techniques to magnetic resonance imaging. Founder and CEO of Quibim (Quantitative Imaging Biomarkers in Medicine), company dedicated to the advanced analysis of medical images using artificial intelligence. He is the author of more than 90 scientific articles in prestigious international journals and inventor of 5+ patents. He is also the author of more than 100 communications to international congresses, editor of international books and author of 20+ book chapters. He has participated in a high number of international research projects and clinical trials. He is an active member of several scientific societies, among which stands out his participation as a member of the Board of Directors of the European Society of Medical Imaging Informatics (EUSOMII).  

Fuensanta Bellvís Bataller, MS. Biomedical engineer specialized in Information and Communication Technologies. She has extensive experience in the field of companion diagnostics and medical imaging. She started her career in a start-up to develop a multidimensional tissue signature for brain tumors and worked as a software engineer optimizing the hospital management system tool used in her region. She started at Quibim as R&D engineer and is currently the VP of Clinical Studies, responsible of the department that manages clinical trials, observational studies and research projects in collaboration with biopharmaceutical companies and academic and research institutions, mainly in the areas of oncology/immunotherapy, rheumatology and neurology.

Blurb

This book, endorsed by EuSoMii, provides clinicians, researchers and scientists a useful handbook to navigate the intricate landscape of data harmonization, as we embark on a journey to improve the reproducibility, robustness and generalizability of multi-centric real-world data radiomic studies.   

In these pages, the authors delve into the foundational principles of radiomics and its far-reaching implications for precision medicine. They describe the different methodologies used in extracting quantitative features from medical images, the building blocks that enable the transformation of images into actionable predictions. This book sweeps from understanding the basis of harmonization to the implementation of all the knowledge acquired to date, with the aim of conveying the importance of harmonizing medical data and providing a useful guidance to enable its applicability and the future use of advanced radiomics-based models in routine clinical practice.  

As authors embark on this exploration of data harmonization in radiomics, they hope to ignite discussions, foster new ideas, and inspire researchers, clinicians, and scientists alike to embrace the challenges and opportunities that lie ahead. Together, they elevate radiomics as a reproducible technology and establish it as an indispensable and actionable tool in the quest for improved cancer diagnosis and treatment.