The Future of Healthcare

Doctor-Patient Relationship in the Internet Era
Prof. Jonathan Halevy, Director General Shaare Zedek Medical

Expectations of patients from their physicians -- both family physicians and specialists -- have changed significantly in the last three decades. Outstanding among the characteristics of current medical practice is the "educated patient" who presents his problems to the treating physician with a background of information derived unselectively from the Internet.

A prominent paradox is manifested by the fact that while modern medicine is able to provide ultra-sophisticated evidence- based diagnostic and therapeutic modalities, a large segment of the public prefers to turn to non- evidence-based holistic complementary and alternative medicine modalities. Are the two phenomena connected? How should the medical community deal with these trends and what is their impact on the doctor-patient relationship and on possible cure rates?

In my presentation I will try to shed light on the doctor-patient relationship in our era in view of the scientific and technological changes currently taking place in the practice of medicine.

The Digital Transformation of American Radiology
Dieter Enzmann MD, Chair of UCLA Radiology

The era of "big data" promises to transform radiology. The current radiology business model is a professional service that is being digitally deconstructed into components, with each component being transformed by computational bioinformatics. Bioinformatics involves the use of computational techniques in pattern recognition, data mining, machine learning, visualization and other techniques to unravel biologic mechanisms. The data mining subset of bioinformatics is already affecting radiology and will transform the shared standards on which the profession is based. As the standards change, the profession will transition into an information business. This transition will accelerate as image interpretation begins to tap into both national and international databases to make use of "population images."

The data mining version of bioinformatics will have several manifestations, including "population images," "radiogenomics," "computational radiology" and ultimately Watson-like diagnostic computers. The summation of one or more image features from a global data set can create phenotypic "population images." That image feature depicted may be an anatomic structure, a disease manifestation, a tumor characteristic or a hemodynamic abnormality. Such "population images" can be used to classify an individual disease/health state, or they can compare different treatment groups in clinical trials.

Bioinformatic techniques in non-imaging databases can be mined in conjunction with imaging databases to better understand biologic mechanisms. Image phenotypes are the end product of complex, combined cellular and tissue network behavior and may provide a means for teasing out such biologic network behavior at multiple levels of organization between the genomic and organ levels. Radiogenomics, therefore, consists of correlating image phenotypes captured in population images to different scales of molecular phenotypes to illuminate underlying biologic mechanisms.

Since diagnostic imaging and molecular diagnosis are both individually valuable services, why not integrate them? Integration provides incremental value if the information being linked is normally used at the same point in time and space. Clinical imaging phenotypes and histologic/molecular phenotypes provide different tiers of complementary information about the state of biologic networks in cells and tissue. Integration of this information into an information business "product" can save time in diagnosing and monitoring diseases. Such integrated products can use serial imaging as an organizing principle for managing chronic diseases.

Adopting the principles of systems biology, we can create "systems radiology," which uses a limited number of image features to define cell and tissue network states. "Systems radiology" will require "computational radiology," which refers to the ability to extract on a large scale meaningful image patterns using search, segmentation and statistical algorithms. "Computational radiology" generates "systems radiology," which can be correlated with "systems biology." The search for validated imaging biomarkers will depend on linking "systems radiology" and "systems biology" via innovative unstructured or structured data mining. Unstructured mining purposely bypasses the filter of human vision to identify correlations with potential biologic significance. The structured approach uses human identified imaging features and subjects them to hypothesis testing not for creating new prospective experiments, but rather to extensively searching data derived from already performed, variegated experiments.

Radiology will need to adapt techniques and algorithms developed in bioinformatics and systems biology to clinical imaging. Powerful computational techniques will create humanized diagnostic computers, which can become fast, cost effective tools in a radiology information business. Radiologists will play a key role in providing the necessary intuition to productively integrate computational personal medical reality. It is early, but radiology will be transformed from a transactional, professional service to an information business composed of professionals.

From Biological Discovery and Personalized Medicine: The Role of Computation in Human Genetics
Itsik Pe'er PhD, Columbia University

Investigation of the statistical association between variation in inherited DNA sequences and clinically relevant traits is predicated by the high heritability of such characteristics. Such genetic research is dually aimed. On one hand, the goal is scientific, to identify molecular underpinnings of disease, understanding mechanisms and thus potentially leading to development of treatment. The other goal is diagnostic, allowing care that is tailored to each patient based on their genetic makeup. Recent technological breakthroughs have exponentially sped up production of genetic data, producing a deluge of information. While these trends usher in the implementation of personalized medicine, they also highlight analytical challenges in such large scale data.
The talk will provide a survey of the field, with examples from our own work on methodology for detecting hidden relatedness and gene-gene interaction with application to direct-to-consumer genetics, pharmacogenetics and isolated populations.