Radiomics: the process and the challenges - PubMed (original) (raw)

Review

. 2012 Nov;30(9):1234-48.

doi: 10.1016/j.mri.2012.06.010. Epub 2012 Aug 13.

Yuhua Gu, Satrajit Basu, Anders Berglund, Steven A Eschrich, Matthew B Schabath, Kenneth Forster, Hugo J W L Aerts, Andre Dekker, David Fenstermacher, Dmitry B Goldgof, Lawrence O Hall, Philippe Lambin, Yoganand Balagurunathan, Robert A Gatenby, Robert J Gillies

Affiliations

Review

Radiomics: the process and the challenges

Virendra Kumar et al. Magn Reson Imaging. 2012 Nov.

Abstract

"Radiomics" refers to the extraction and analysis of large amounts of advanced quantitative imaging features with high throughput from medical images obtained with computed tomography, positron emission tomography or magnetic resonance imaging. Importantly, these data are designed to be extracted from standard-of-care images, leading to a very large potential subject pool. Radiomics data are in a mineable form that can be used to build descriptive and predictive models relating image features to phenotypes or gene-protein signatures. The core hypothesis of radiomics is that these models, which can include biological or medical data, can provide valuable diagnostic, prognostic or predictive information. The radiomics enterprise can be divided into distinct processes, each with its own challenges that need to be overcome: (a) image acquisition and reconstruction, (b) image segmentation and rendering, (c) feature extraction and feature qualification and (d) databases and data sharing for eventual (e) ad hoc informatics analyses. Each of these individual processes poses unique challenges. For example, optimum protocols for image acquisition and reconstruction have to be identified and harmonized. Also, segmentations have to be robust and involve minimal operator input. Features have to be generated that robustly reflect the complexity of the individual volumes, but cannot be overly complex or redundant. Furthermore, informatics databases that allow incorporation of image features and image annotations, along with medical and genetic data, have to be generated. Finally, the statistical approaches to analyze these data have to be optimized, as radiomics is not a mature field of study. Each of these processes will be discussed in turn, as well as some of their unique challenges and proposed approaches to solve them. The focus of this article will be on images of non-small-cell lung cancer.

Copyright © 2012 Elsevier Inc. All rights reserved.

PubMed Disclaimer

Figures

Figure 1

Figure 1

The process and challenges in radiomics.

Figure 2

Figure 2

The computed tomography (CT) phantom. This phantom has several regions to test image quality such as low contrast detectability and spatial resolution.

Figure 3

Figure 3

Effect of two different reconstruction algorithms on same raw CT data (A and B) where A shows a “standard smooth image” and B shows the same raw data reconstructed using a higher contrast algorithm. To appreciate the effect of these reconstruction algorithms the profiles (in Hounsfield Units) along the vertical lines are shown (C and D, respectively). Even the average Hounsfield Units in the tumor are different for the different algorithms.

Figure 4

Figure 4

Metabolic volume calibration; PET phantom with differently sized sphere sources filled with FDG activity within a background activity. By varying the source to background activity ratio the capability of the PET scanner to reconstruct the correct sphere volume can be quantified.

Figure 5

Figure 5

The variation in slice thickness (A) and pixel size (B) for a dataset of 74 patients.

Figure 6

Figure 6

Representative examples of lung tumors attached to anatomical structures like pleural wall, mediastinum or heart that are difficult to segment automatically.

Figure 7

Figure 7

Architecture of the proposed Radiomics database (RDB). High-level database schema capturing the following data types: image types (orange), image features (purple), patient/clinical (blue), and molecular (green) data. Each box represents a set of normalized tables. This schema supports multiple tumors for one patient, with multiple images series, using multiple segmentations generating different image features.

Figure 8

Figure 8

Unsupervised hierarchical clustering of lung tumor image features extracted from CT images from 276 non-small cell lung cancer patients. Tumor segmentation for each CT image was performed in a semi-automated fashion. Quantitative imaging features were calculated using Definiens (Munchen, Germany) and represent many 2-dimensional and 3-dimensional characteristics of the tumor. Aspects such as tumor volume, shape and texture were represented. Each of the numerical imaging features was median-centered and all features were clustered using complete linkage, with correlation used as the similarity measure. The resulting heatmap is visualized using red to represent higher than median feature values, and green to represent lower than median feature values. Each row of the heatmap represents a specific imaging feature across patients and each column represents all features for a patient’s lung tumor from CT.

Similar articles

Cited by

References

    1. Lambin P, Rios-Velazquez E, Leijenaar R, Carvalho S, Stiphout RV, Granton P, Zeghers K, J. GR, Boellard R, Dekker A, Hugo JWLA. Radiomics: extracting more information from medical images using advance feature analysis. The European Journal of Cancer. 2011 - PMC - PubMed
    1. Jaffe CC. Measures of response: RECIST, WHO, and new alternatives. Journal of clinical oncology : official journal of the American Society of Clinical Oncology. 2006;24(20):3245–3251. - PubMed
    1. Burton A. RECIST: right time to renovate? The Lancet Oncology. 2007;8(6):464–465.
    1. Rubin DL. Creating and curating a terminology for radiology: ontology modeling and analysis. Journal of Digital Imaging. 2008;21(4):355–362. - PMC - PubMed
    1. Opulencia P, Channin DS, Raicu DS, Furst JD. Mapping LIDC, RadLex™, and Lung Nodule Image Features. Journal of Digital Imaging. 2011;24(2):256–270. - PMC - PubMed

Publication types

MeSH terms

LinkOut - more resources