Imaging informatics for personalised medicine: applications and challenges

Tianming Liu, Hanchuan Peng, Xiaobo ZhouInt. J. Funct. Informatics Pers. Medicine 2009 · DOI

Introduction

Modern medicine is undergoing a paradigm shift toward a "personalized" approach, moving away from reactive treatments toward predictive, preventive, and highly individualized care. At the heart of this transformation is imaging informatics, which serves as the essential bridge between raw biomedical imaging data and clinical practice. By integrating advanced image analysis with clinical and genomic data, the field aims to translate high-resolution scans into actionable clinical phenotypes.

The central role of Imaging Informatics in Biomedicine, bridging core technical challenges with diverse clinical applications.

The Translation Gap

The sheer volume and complexity of biomedical imaging—spanning molecular, cellular, and organic scales—present a significant "translation gap." The primary challenge lies in managing massive datasets and ensuring that algorithms are robust enough for clinical use. To be truly effective, imaging findings must be integrated with other data types, such as genomics, to predict how a specific patient will respond to a particular treatment.

Methodological Pillars

This review synthesizes the state of the art across several key technological domains that underpin personalized medicine:

  • Image-Guided Therapy (IGT): Utilizing real-time tracking and data fusion to assist surgeons during complex procedures.
  • Voxel-Based Morphometry & DTI: Quantifying structural and connectivity changes in the brain, which is vital for diagnosing conditions like Alzheimer’s.
  • Molecular Imaging: Employing target-specific tracers to visualize cellular events in vivo.
  • Informatics Infrastructure: Evolving Picture Archiving and Communication Systems (PACS) from simple storage into active decision-support tools.

Big Data and Collaboration

Personalized medicine research relies on large-scale collaborative projects that provide the necessary "big data." Key initiatives include:

  • ADNI (Alzheimer’s Disease Neuroimaging Initiative): A database of ~800 patients featuring MRI, PET, and clinical metadata.
  • fBIRN (Functional Biomedical Informatics Imaging Network): A multi-site network focused on standardized fMRI studies for Schizophrenia.
  • caBIG (Cancer Biomedical Informatics Grid): An initiative to validate imaging biomarkers for cancer detection and therapy follow-up.
  • NAMIC (National Alliance of Medical Image Computing): A project providing open-source software (like ITK) for medical image analysis.

Key Insights and Results

Imaging informatics is critical for developing "imaging-based phenotypes." These can be correlated with genetic data to enable targeted drug delivery and personalized treatment regimens.

The Calibration Crisis

A significant hurdle in multi-center studies is facility calibration. Variations in MRI scanner manufacturers, head coil characteristics, and acquisition trajectories can lead to divergent results, necessitating sophisticated "rationalization" methods to ensure data consistency.

Benchmarking Progress

While the field is advancing, benchmarks show room for improvement. Current state-of-the-art algorithms for subcortical brain structures achieve Dice Coefficients of ~70% for the hippocampus/amygdala and ~80% for the thalamus and putamen. However, quantitative imaging is already proving its worth in differential diagnosis; for example, structural MRI and Diffusion Tensor Imaging (DTI) can now distinguish between Alzheimer’s Disease (AD) and Dementia with Lewy Bodies (DLB) by identifying specific regional volume losses.

Future Directions and Challenges

To reach its full potential, the field must address several limitations:

  1. Data Sharing: Social and legal hurdles, including HIPAA compliance and IRB regulations, often prevent the seamless pooling of human imaging data.
  2. Integration: There is currently no "unified field theory" for integrating imaging informatics with bioinformatics (genomics) and electronic health records.
  3. Algorithm Robustness: Many current tools are not yet reliable or interpretable enough for routine clinical use. The authors call for open-source, open-architecture systems to foster community-wide validation and collaboration.

Summary updated

BibTeX
@article{liu2009imaging,
  author = {Tianming Liu and Hanchuan Peng and Xiaobo Zhou},
  title = {{Imaging informatics for personalised medicine: applications and challenges}},
  journal = {Int. J. Funct. Informatics Pers. Medicine},
  year = {2009},
  doi = {10.1504/ijfipm.2009.027587},
}
Figure