Corporate news
Bo Hu: "Semantic Web technologies can help save lives."
03. November 2008, by Tassilo Pellegrini 
Bo, you mainly deal with the application of semantic technologies in medicine and life sciences. What is your motivation to do this? What fascinates you about this field of research?
Looking back from this point in time, I cannot help to ask myself: while information and communication technologies (ICT) have achieved incredible advances in the last decade or so and have been successfully applied in numerous fields to improve the quality of our everyday life, why do we still find old fashioned services dominant in public health care? Have we, as computer scientists or IT professionals, done enough in the so-called semantic web (SW) era to make patients suffer less and to make medical doctors’ lives easier?
I probably can tell others "yes, I have tried and failed", but I could not lie to myself and neither could my colleagues. We believe, in many cases, simple technologies can make a lot of difference if they are applied to the right problems. This is evident in electronics health record management and sharing where Semantic Web technologies increase data interoperability. New technologies can also grant patient data well defined "meanings" to help in data interpretation, classification, annotating and reusing available information from the Internet, etc.
Till now, no other domain compares to the medicine and life science domain in terms of the amount of data that need to be processed, the diversity of data formats, and the time-critical and safety-critical requirements. Potential semantic web and agent technologies will be put to the test regarding their robustness, scalability, flexibility, and usability. Of course, there is also the ultimate prize - my system/tool can save human life!
Life sciences, especially cancer research, are one of the major adopters of Semantic Web standards. What is the prospect of using computational semantics for improving research and development in this field?
Skeptics always say "Yeah! Semantic web technologies are fantastic, but what can they do for me that a database cannot?" I am always tempted to reply, think of the entire Internet as a gigantic distributed database with hidden data held in individual hospitals, research centres, and institutes. Is access to these data useful? Yes. Ground-breaking discoveries can be made if researchers in medicine are exposed to data rich in diversity. Is access to these data easy? Yes and no. The sheer size of the data available is more than any one can handle. Semantic web technologies serve to provide us access to such data and make good use of such data.
A combination of semantic web technologies and cancer research presents a twofold challenge, or, if we put it in a positive tone, opportunity. Semantic web technology can be of great help in data management, such as data sharing, data classification, etc. So, instead of isolated individual data/information islands, the archipelagic data landscape is rendered into a holistic representation. Knowledge that is not available to individual data/information sources can be drawn and passed on to an integrated system.
On the other hand, semantic web (and agent) technologies can help in formalising and modelling medical and biological systems. Unlike other walks of life, many sub-domains of medicine are subject to controlled nomenclatures providing solid ground upon which semantically enriched applications can be built. In particular, cancer research has been reaching out to seek help in domain ontologies to unify local vocabularies, ontology-based data annotation and retrieval, evidence-based internet search engines for knowledge reuse, data integration, user friendly interface to increase acceptance, etc.
Although many of the applications are “icing on the cake”, as skeptics put it, they help us to have a better understanding of the numerous data that we have gathered. Useful and meaningful interpretation and use of such data will slowly emerge therefrom.
One of your fields of expertise are agent technologies. What role will they play in a future healthcare infrastructure? Can you cite examples to illustrate their purpose?
I wouldn’t say Agent technology is one of my fields of expertise. What we are really interested in is how to apply agent and semantic web technologies to offer better services to patients as well as relieve doctors of the routine, repetitive tasks.
It is becoming increasingly clear that significant improvements can be achieved in clinical data management, patient management, data classification, etc., if such tasks can be made -- at least partially -- distributed and designated to those who specialise in parts of that task.
Moreover, in the situation of comorbidity (e.g. heart disease, AIDS, cancer, diabetes, or mental health), it is not a surprise to find that a patient is examined in one hospital; his/her case is reviewed by clinicians from another hospital; and he/she is treated in a third hospital by yet another group of clinicians due to speciality and availability.
Along with the opportunities come the challenges. On one hand, in distributed environments, it is difficult to exploit the available data from different sources, especially data that is normally projected onto the body of a patient to reach diagnostic and prognostic decisions. Many of the available data are interrelated calling for paradigms that facilitate knowledge discovery by intelligently integrating data sources.
An Agent-based framework is particularly useful in this case where individual agents are equipped with “memory” and “reasoning/thinking” capabilities to constantly acquire new knowledge and solve allocated tasks. Communication among agents, prescribed by a common vocabulary/ontology, ensures the entire community works towards a common goal.
On the other hand, frameworks with agents encapsulating special functions deliver better customised and personalised healthcare. As a result, we will be witnessing more patient power and better adherence to treatment regimens.
What are the biggest obstacles for the success of semantic technologies (incl. agents) in medicine and life sciences? Is it just about technological feasibility or are other factors concerned too?
The challenges are present in many multidisciplinary fields. People from different communities may need to acquire knowledge from one another. Such knowledge might be superficial or it might require deep understanding. Effective communication should be maintained to ensure that any research, design, and implementation satisfy needs from both sides.
There is also the usability concern. Many people with computing background tend to narrow down our research focus on issues that are interesting and/or challenging from the computer science perspective. In many cases, we have developed a fantastic concept proving prototype that is hardly acceptable from a clinician/healthcare practitioner’s point of view.
There are also cases when a prototype system works fine with respect to toy examples or purposely built/tailored evaluation environments, but fails to function as expected when deployed. Real life problems never quite fit what we could imagine in the lab.
An easy solution to both of the above issues is to involve medical experts as early as possible, even at the proposal stage. Gaining trust from domain experts is easier done with basic but working improvements instead of ambitious but unfeasible promises.
At a more technical level, challenges sometimes come from false expectations in probably both communities. Personally, I believe many fantastic concepts are ‘sold’ to others too early, even before an agreement is reached in the computer science community.
In many occasions, SW technology failed to deliver what it once promised. SW technologies cannot guarantee semantic and data interoperability. Nor can it bring instant fundamental changes to the current clinical system and paradigm. Any concrete improvement on this front requires international efforts from multiple disciplines such as medicine, sociology, psychology, computer science, IT, just to name a few.
So are the Agent technologies that are still in the “infant” stage and too early for more sophisticated systems than the current generation of application mainly in data management with limited reasoning and mediating capabilities.
About Bo Hu
Dr Bo Hu is a researcher at SAP Research CEC Belfast. He received his PhD in Computer Science from the Robert Gordon University, Aberdeen in 2004. Between 2002 and 2008, he worked as a Research Fellow in the Intelligence, Agent, Multimedia Group (IAM), School of Electronics and Computer Science, University of Southampton. During his days in Southampton, he was actively involved in UK EPSRC Advanced Knowledge Technology IRC and EU FP6 projects. His main research interest is in Knowledge Management (KM), KM in pervasive computing environments, Semantic Web, Web 2.0 and their applications in e-learning and e-healthcare.
Links of interest
Past and ongoing projects in the area of life sciences and bio informatics:
Comments
Add new comment