1) Medicine is increasingly becoming a data-intensive field due to the digitization of health records, research data, and patient self-tracking data. 2) The volume and diversity of biomedical data, known as "Big Data", provides opportunities to gain insights and improve patient outcomes but also poses challenges around data integration and analysis due to issues like heterogeneity and noise. 3) Techniques like data mining, machine learning, and knowledge discovery in databases are used to extract meaningful information and discover patterns in large and complex biomedical data to support areas like predictive analytics and personalized medicine.

1. Intereg Project: Biomedical Informatics
Ljiljana Majnarić Trtica
Session 1: Medicine as a data science

2. I. Medicine as a data science
In the last decades, the life science, biomedicine and health care are increasingly turning into a data
intensive science. This is associated with the expansion of available electronic data, including:
digitization of electronic health records (EHR), aggregation of research data into databases in
pharmaceutical industries, the release of stored patient data by the governments for research
purposes (e.g. patients health insurance claims), aggregation of research data from clinical trials,
epidemiological and biomedical research, the emergency of the high tech medicine (omics-medicine)
and the onset of patient self-tracking and remote monitoring by using mobile devices and biosensors.
The frequently cited definition of Medical Informatics is that of Shortliffe and Perrault (1990): “… the
rapidly advancing scientific field that deals with the storage, retrieval and optimal use of biomedical
information, data and knowledge for problem solving and decision making”. From this definition, it is
clear that the key role of Medical Informatics is to support medical doctors in making decisions.
More recently, the distinction has been made between Medical Informatics and Biomedical
Informatics. While the objectives of interest of Medical Informatics include populations, individuals,
organs and tissues, Biomedical Informatics, as its objectives, also has the microscopic levels of bodily
organisation, including cells and molecules.
Medical (Biomedical) Informatics is a science of data in clinical medicine (biomedicine). This data
has some specific characteristics, such as:
 Different resources of data
 Increasing size (volume)
 Increasing diversity
 Multi-dimensional (of different meaning, subclasses)
 Highly complex (an example is the microscopic structure of a yeast protein network) (Fig. 1)
 Often weakly structured (as the text in the patient records, the signals from physiological sensors)
 Noisy (missing and inconsistent)
These characteristics are the reasons that there is a growing need for this data integration and
modeling, using different computer methods for data analysis. These methods are the content of
Medical (Biomedical) Informatics.

3. Fig. 1. The computer-based visualization of the yeast protein network
As it is visible in the Fig. 1, a great challenge is how to find unknown structures (structural homologies)
in the enormously high number of uncharacterized data. By application of a special visualization
method, these structures become visible, thus enabling medical professionals to understand these
data more easily.
Increase in volume and diversity of data in biomedical practice and science, in the last decades, has
got the common term ”Big Data”. Big Data provides us the opportunity to gain insights into the
meaning of data, derive new knowledge and foster discoveries further on, that is expected to improve
patient outcomes, reduce costs and accelerate biomedical advances.
Some examples of how Big Data can be used to improve medical practice:
 To boost applicability of clinical research studies into real-world scenarios where population
heterogeneity is an obstacle, thus, changing the paradigm: from the hypothesis driven to data-driven
medicine (Fig. 2).
 To foster extraction and effective and innovative use of knowledge hidden within the huge volumes
of data,
 To enable patients identification who are at risk for unfavorable health outcomes (disease, death,
hospital (re-)admission),
 To enable effective and precision medicine through patient risk stratification ( a key task toward
personalized health care),
 To enable predicitive analytics in personalized health care

4. Fig. 2. Big Data and the paradigm changes in biomedical science: Hypothesis-driven vs data-driven medicine
(Doctoral thesis of the author)
Clinical research tasks should determine research methods. This is the opposite to what is nowadays,
where clinical projects meet the criteria of the established research methods.
The paradigm changes also means the switch from the descriptive (curable) to the predictive,
preventable and proactive, participatory (with patient active participation) medicine (P4 medicine, or
personalized medicine).
The increasing role of personalized medicine, in biomedical science in practice, evolves together
with the two major technological advances, including: 1) omics-based medicine and 2) computer-
based methods for data analysis (Medical and Biomedical Informatics).
The omics-based medicine includes the new-generation of DNA sequencing, that is combined with
new molecular biology methods: transcriptomics, proteomics and metabolomics. These new
technologies have enabled the development of the new scientific discipline, systems biology, that
means an integrative analysis of data of different levels of bodily organization. This new discipline
enables connections between phenotypes and molecular patways (Fig. 3) and identification of new
targets for personalized treatments.

5. Fig. 3. The challlenge of systems biology in creating molecular pathways and networks
P4 medicine is: Personalized, Predictive, Preventive and Participatory medicine. The key benefits of
P4 medicine include the ability to:
- detect disease at an earlier stage, when it is easier and less expensive to be treated effectively
- stratify patients into groups that enable selection of optimal therapy (Fig. 4)
- reduce adverse drug reactions by more effective early assessment of individual drug responses
- improve selection of new biochemical targets for drug discovery
- reduce the time, cost and failure rate of clinical trials for new therapies
- shift the emphasis in medicine from reaction to prevention and from disease to wellness
Fig. 4. Stratification of patients into groups to enable selection of optimal therapy
There are also some obstacles to effectively use Big Data for practical purposes. These obstacles deal
with the following problems:

6.  The problem of heterogeneous data (biomedical data are used from various sources and show
different structural dimensions, varying from microscopic (omics-data) to the macroscopic world (e.g.
data on disease prevalence in the population statistics)
 The problem of data sharing and distribution among different providers and departments
 Often noisy, missing, inconsistent and non-standardized data
There is also a gap between the available data and data that are applicable for practical purposes.
This is the reason why data processing is an important step in the process of knowledge discovery
from data.
Knowledge discovery in databases (KDD) is the process that includes several steps: data selection,
data pre-processing, data transformation, data mining (considered as the proces of data analysis)
and results interpretation (Fig. 5).
Fig. 5. The steps in the KDD process
The challenge of KDD from Big Data is to: extract meaningful information from data, gain new
knowledge, discover previously unknown insights, find patterns and make sense of data. Many
different approaches have been developed of KDD from Big Data, including: new mathematical and
graphical methods, Data Mining (DM)and Machine Learning (ML) methods (mostly used methods in
the past).
Data mining is the term that has a dual meaning. It can be considered as a key step in the KDD process
(the term has been used: Knowledge Discovery and Data Mining, KDD, and as the computational
process of discovering previously unknown, valid patterns and relationships in large data sets, that
can be used for prediction, classification and clustering purposes.

7. Data mining, when it is considered as a computer-based method, consists of a combination of
sophisticated methods, including: statistical models, mathematical algorithms and ML methods (algorithms
that improve their performance automatically through experience).
Application of Data Mining techniques and methods in Health Care domain has led to:
 the developmnet of intelligent systems and decision support systems (rule-based expert systems)
 improvement of the prediction of unfavorable health outcomes and diagnosis
 improved disease classification
 the discovery of relationships between pathological data and clinical data and between patients
characteristics and medications efficiency
 candidate selection process for medical tests and procedures
One new concept has been developed, in association with KDD. This is the concept of the Human-
Computer Interaction (HCI). Interaction is the key topic in this concept (Fig. 6). In this context:
 The KDD is a process ranging from the physical side of data to the human side of knowledge (defined
as the cognitive process).
 The challenge is in making knowledge to be usable by end users (by making sense of data).
 The process added to KDD is INTERACTION (COMMUNICATION) with the human end user (medical
expert).
 It is the human end user (not machine) who posses the problem solving intelligence, hence, the
ability to ask intelligent questions about the data.
 The human (medical expert) is able to solve complex problems sometimes intuitively (that is, without
the need to describe the exact rules or processes used during the problem analysis).
Fig. 6. Visual presentation of the HCI concept (the origin: Medical University of Graz, the group for HCI)

8. Or, according to the words of Albert Einstein (USA/German-born physicist, 1879 - 1955): Computers
are incredibly fast, accurate, and stupid. Human beings are incredibly slow, inaccurate, and
brilliant. Together they are powerful beyond imagination.