• Home
  • Print this page
  • Email this page
Home About us Editorial board Ahead of print Current issue Search Archives Submit article Instructions Subscribe Contacts Login 

 Table of Contents  
Year : 2016  |  Volume : 2  |  Issue : 2  |  Page : 61-63

Imaging-driven Digital Biomarkers

Center for Computational Science, University of Miami, Miami, FL, USA

Date of Submission26-Feb-2016
Date of Acceptance10-Apr-2016
Date of Web Publication29-Apr-2016

Correspondence Address:
Dr. Enrico Capobianco
Center for Computational Science, University of Miami, Gables One Tower 600, 1320 South Dixie Highway, Suite 600K, Loc: 2965, Coral Gables, Florida 33146
Login to access the Email id

Source of Support: None, Conflict of Interest: None

DOI: 10.4103/2395-3977.181440

Rights and Permissions

Cancer imaging is crucial for advancing diagnostics and treatment, thus directly impacting medical practice. A wealth of medical information is collected worldwide in the form of digital medical images as a main component of electronic health records. Correspondingly, the unprecedented information, qualitative or quantitative, that is becoming available, calls for integrative multiplexed approaches linking genetic molecular profiles, imaging, and experimental omics evidence, but also clinical records, social-driven, and self-quantified data. The paper aims to emphasize and justify the potential of integrative approaches centered on a systems' view of newly generated markers, namely digital ones.

Keywords: Biomarkers, cancer imaging, digital medicine, heterogeneity, networks

How to cite this article:
Capobianco E. Imaging-driven Digital Biomarkers. Cancer Transl Med 2016;2:61-3

How to cite this URL:
Capobianco E. Imaging-driven Digital Biomarkers. Cancer Transl Med [serial online] 2016 [cited 2020 Sep 25];2:61-3. Available from: http://www.cancertm.com/text.asp?2016/2/2/61/181440

  Introduction Top

Digital medicine is expected to lead major changes in multiple directions, with effects on measurement, interpretation, and insight. This can transform the importance of digital data and the extent they will affect the medical decision processes. At a structural level, instead of disparate collections of patient records, both collecting and assembling of the digital records will induce effective risk stratifications and more accurate monitoring of the quality of health care delivery. At a strictly clinical level, the deployment of observational evidence can build an unprecedented knowledge base to improve the currently constrained randomized trials.

Digital medicine is also leading to the next generation of biomarkers, newly defined entities derived as a result of merging experimental, clinical, observational, social, and personal evidence and records. [1],[2],[3] These bio-entities promise to expand measurement resolution and detection accuracy, with regard to both physiological and pathological conditions at either individual or population level, and to mitigate the curse of both dimensionality and complexity by scalable and efficient computational analysis. As biomarkers belong to several categories, digital phenotypes should induce a re-classification of biomarkers, and will depend on the novelty of information contents. It can be expected an expansion of patients' profile to multiple dimensions, which with the help of sensitive monitoring technologies combined with connected devices, can bring implementations in personalized health programs. [4],[5],[6]

When it comes to cancer, biomarkers related to molecular imaging acquire a prominent role. [7],[8] Novel and more sophisticated sources of imaging have been set, for instance, The Cancer Imaging Archive. The National Cancer Institute has supported the creation of both a big data resource and of a quantitative imaging network to build field developments and expand critical tumor response biomarkers in clinical trials. [9] At a qualitative level, distinct features of imaging biomarker scan complement other types of biomarkers and emphasize the ones delivering new predictive and early-response information by serial measurements over a period. A main goal is targeting heterogeneity by exploring the microenvironment, perturbed or not, and by elucidating the multitude of interactions occurring between signals from various cell types. [10],[11],[12] However, as multimodal cancer imaging evolves, reaching an increased predictive power requires integrative approaches. [13] In particular, the required power can be achieved by joint consideration of mechanic coupling of cell movements to tissues and multi-scale modeling.

  Heterogeneity Calling for Networks Top

A current bottleneck in image analysis is the lack of spatial resolution, which can be partially addressed by the consideration of intra-tumoral environments in light of inference about new specific phenotypes associated with genetic changes. [14] In such domain, networks are an emerging methodological approach of great potential, one dealing well with nonlinear complex dynamics. [15] On one hand, the specificity of cancer features and their classification allow to represent heterogeneity as a mixture of information generated by a milieu of cell sub-populations, each with covariates explaining differentiated phenotypes under the influence of genetic and epigenetic forces. On the other hand, networks inherently modularize or form communities subject to distinct connectivity dynamics, while cross-linking. Thus, the bridge is established by an association between the mixture components and the network modules.

Modern imaging techniques, such as magnetic resonance imaging, computed tomography, positron emission tomography allow measurements of anatomical details with high resolution and maps of the dynamics generated by several physiological parameters. [16],[17],[18] Only the integration of all of the information arising from different techniques can create a representation of the tumor status. Such integration remains unfeasible in the image space, but can be obtained in a transformed space. Nodes in Networks represent the voxels and links corresponding to communication between them, reflecting a spatial metric defined on the basis of similarity/dissimilarity measures computed over the image features. The resulting connectome entangles all the interdependencies between nodes, and hosts modules identified according to the significance of intra-connected module (or community) patterns in comparison to the inter-connected module or even random patterns.

  Computational Challenges and Networks Paradigm Top

Digital biomarkers can translate information from a multitude of data through a variety of methods, and finally organize what is available into systems. Therefore, the building blocks of digital biomarkers are data (measurement challenge), methods (quantification problem), and systems (assessment of relationships and significance). Specifically, each of the three components can be summarized by a few keywords described in [Figure 1]. These are: (1) compatibility, such that data can be analyzed, cross-referenced, and cross-linked; (2) transferability, implying portability across platforms and algorithms, and reproducibility of the results; (3) generalizability, thus allowing for predictive modeling.
Figure 1. Cascade representation from data to systems through methods. Data are considered structured or not. Methods are the quantitative components. Systems are meant to consider ensemble measures and large-scale interactions between a variety of entities with different characteristics over a hierarchy of complexity levels

Click here to view

An initial assessment of compatibility, transferability, and generalizability identifies a major strength and two weaknesses. The strength pertains to data compatibility where data standards are destined to generate harmonization thus improving cross-referenced analyses and longitudinal studies while spatially expanding at both local and global scale. The weaknesses refer to method transferability, depending on how data are modeled and integrated, something still lacking universally accepted solutions, and to generalizability, which is a property reflecting the complexity embedded in each particular system. [19],[20]

Note the various characterizations reported in [Figure 1], in particular at the top layer with big versus small data and at the bottom layer where disease-specific and network-centric inferences are recalled, both implying systems views (i.e., disease gene or protein maps, or topological configurations). The translation from image units (voxels) to network units is represented in [Figure 2]. [Figure 3] exemplifies the mapping of heterogeneous features through selected regions of the network interactome, then benchmarked against suitably chosen null model to prove significance of observed patterns. This framework is suitable for the examination of cancer microenvironment, in which the presence of a mix of cell sub-populations is hard to disambiguate. [21],[22] Their reproduction at a network scale allows the isolation of sub-interactomes associated to dissected (possibly differentiated) microenvironment space. [23]
Figure 2. (a) Voxel-node relationships. Matrix representation of voxels, observed effects, and interactions of network nodes. Briefly, it is a concept of distance applies when establishing the presence of an interaction or not, and this originates from the communication degree observed at voxel level. (b) Network configurations and corresponding patterned adjacency matrices. A corresponding adjacency matrix, whose structure allows straightforward computations, can summarize each network

Click here to view
Figure 3. Dealing with interactomes, from global to sampled, and random. Biclusterized sub-interactomes refer to blocks of interacting nodes extracted with reference to a subsample and to a subset of genes or proteins. Bicluster closure indicates the recovery of the corresponding region of the network adjacency matrix. The null model is a random matrix against which to test significance of the extracted components. The circularity of these structures allows to build a methodological approach aimed to achieve significant network patterns and consistency across data

Click here to view

  Conclusion Top

We are facing an era of change destined to positively affect people, society, culture, and science. Digital medicine and digital biomarkers are an adaptive response to what is happening and promise to lead us toward a new medical culture, particularly with regard to doctor-patient relationships. Behind the scenes, huge cross-disciplinary machinery is working to connect the dots, and solidify the described scenario.

Financial support and sponsorship


Conflicts of interest

There are no conflicts of interest.

  References Top

Shaffer DW, Kigin CM, Kaput JJ, Gazelle GS. What is digital medicine? Stud Health Technol Inform 2002; 80: 195-204.  Back to cited text no. 1
Elenko E, Underwood L, Zohar D. Defining digital medicine. Nat Biotechnol 2015; 33 (5): 456-61.  Back to cited text no. 2
Topol EJ. Transforming medicine via digital innovation. Sci Transl Med 2010; 2 (16): 16cm4.  Back to cited text no. 3
Ziegler A, Koch A, Krockenberger K, Grosshennig A. Personalized medicine using DNA biomarkers: a review. Hum Genet 2012; 131 (10): 1627-38.  Back to cited text no. 4
Hayes DF, Markus HS, Leslie RD, Topol EJ. Personalized medicine: risk prediction, targeted therapies and mobile health technology. BMC Med 2014; 12: 37.  Back to cited text no. 5
Godman B, Finlayson AE, Cheema PK, Zebedin-Brandl E, Gutiérrez-Ibarluzea I, Jones J, Malmström RE, Asola E, Baumgärtel C, Bennie M, Bishop I, Bucsics A, Campbell S, Diogene E, Ferrario A, Fürst J, Garuoliene K, Gomes M, Harris K, Haycox A, Herholz H, Hviding K, Jan S, Kalaba M, Kvalheim C, Laius O, Lööv SA, Malinowska K, Martin A, McCullagh L, Nilsson F, Paterson K, Schwabe U, Selke G, Sermet C, Simoens S, Tomek D, Vlahovic-Palcevski V, Voncina L, Wladysiuk M, van Woerkom M, Wong-Rieger D, Zara C, Ali R, Gustafsson LL. Personalizing health care: feasibility and future implications. BMC Med 2013; 11: 179.  Back to cited text no. 6
Mankoff DA, Pryma DA, Clark AS. Molecular imaging biomarkers for oncology clinical trials. J Nucl Med 2014; 55 (4): 525-8.  Back to cited text no. 7
Kircher MF, Hricak H, Larson SM. Molecular imaging for personalized cancer care. Mol Oncol 2012; 6 (2): 182-95.  Back to cited text no. 8
Hainaut P, Plymoth A. Targeting the hallmarks of cancer: towards a rational approach to next-generation cancer therapy. Curr Opin Oncol 2013; 25 (1): 50-1.  Back to cited text no. 9
Yankeelov TE, Atuegwu N, Hormuth D, Weis JA, Barnes SL, Miga MI, Rericha EC, Quaranta V. Clinically relevant modeling of tumor growth and treatment response. Sci Transl Med 2013; 5 (187): 187ps9.  Back to cited text no. 10
Quail DF, Joyce JA. Microenvironmental regulation of tumor progression and metastasis. Nat Med 2013; 19 (11): 1423-37.  Back to cited text no. 11
Chen F, Zhuang X, Lin L, Yu P, Wang Y, Shi Y, Hu G, Sun Y. New horizons in tumor microenvironment biology: challenges and opportunities. BMC Med 2015; 13: 45.  Back to cited text no. 12
Gatenby RA, Grove O, Gillies RJ. Quantitative imaging in cancer evolution and ecology. Radiology 2013; 269 (1): 8-15.  Back to cited text no. 13
Kalpathy-Cramer J, Freymann JB, Kirby JS, Kinahan PE, Prior FW. Quantitative imaging network: data sharing and competitive algorithm validation leveraging the cancer imaging archive. Transl Oncol 2014; 7 (1): 147-52.  Back to cited text no. 14
Dominietto M, Tsinoremas N, Capobianco E. Integrative analysis of cancer imaging readouts by networks. Mol Oncol 2015; 9 (1): 1-16.  Back to cited text no. 15
Kwee TC, Basu S, Torigian DA, Saboury B, Alavi A. Defining the role of modern imaging techniques in assessing lymph nodes for metastasis in cancer: evolving contribution of PET in this setting. Eur J Nucl Med Mol Imaging 2011; 38 (7): 1353-66.  Back to cited text no. 16
Dance A. Eyes on Cancer. Techniques for Watching Tumors do Their Thing. The Scientist. Available from: http://www.the-scientist.com/?articles.view/articleNo/31889/title/Eyes-on-Cancer. [Cited 2012 Apr 01].  Back to cited text no. 17
Feigenberg S, Campassi C, Sharma N, Kesmodel SB, Tkaczuk K, Yu JQ. Integration of modern imaging into the multidisciplinary setting: the radiation oncology perspective. Appl Radiol 2012; 19: 25.  Back to cited text no. 18
Davies PC, Demetrius L, Tuszynski JA. Cancer as a dynamical phase transition. Theor Biol Med Model 2011; 8: 30.  Back to cited text no. 19
Brock A, Krause S, Ingber DE. Control of cancer formation by intrinsic genetic noise and microenvironmental cues. Nat Rev Cancer 2015; 15 (8): 499-509.  Back to cited text no. 20
Ciriello G, Miller ML, Aksoy BA, Senbabaoglu Y, Schultz N, Sander C. Emerging landscape of oncogenic signatures across human cancers. Nat Genet 2013; 45 (10): 1127-33.  Back to cited text no. 21
Whipple CA. Tumor talk: understanding the conversation between the tumor and its microenvironment. Cancer Cell Microenviron 2015; 2 (2): e773.  Back to cited text no. 22
Knaack SA, Siahpirani AF, Roy S. A pan-cancer modular regulatory network analysis to identify common and cancer-specific network components. Cancer Inform 2014; 13 (Suppl 5): 69-84.  Back to cited text no. 23


  [Figure 1], [Figure 2], [Figure 3]

This article has been cited by
1 Precision Oncology: The Promise of Big Data and the Legacy of Small Data
Enrico Capobianco
Frontiers in ICT. 2017; 4
[Pubmed] | [DOI]


Similar in PUBMED
   Search Pubmed for
   Search in Google Scholar for
 Related articles
Access Statistics
Email Alert *
Add to My List *
* Registration required (free)

  In this article
Heterogeneity Ca...
Computational Ch...
Article Figures

 Article Access Statistics
    PDF Downloaded277    
    Comments [Add]    
    Cited by others 1    

Recommend this journal