Data Mining in Healthcare and Biomedicine: A Survey of the Literature

داده کاوی در زمینه ی بهداشت و درمان : مروری بر آثار

گروه 13

ساعت 8تا10

عاطفه محمدی ، نرگس سادات فاضلی، فاطمه زهرا پیرندستانی

Abstract

• As a new concept that emerged in the middle of 1990’s, data mining can help researchers gain both novel and deep insights and can facilitate unprecedented understanding of large biomedical datasets. Data mining can uncover new biomedical and healthcare knowledge for clinical and administrative decision making as well as generate scientific hypotheses from large experimental data, clinical databases, and/or biomedical literature.

​

Introduction to data mining

What is data mining?

While data mining mainly originated from work done in the field of statistics and machine learning as an interdisciplinary field, data mining has advanced from these beginnings to include pattern recognition, database design, artificial intelligence, visualization, etc. There are several definitions of the term data mining. One of the most widely-used definitions states that “data mining is the analysis of (often large) observational datasets to find unsuspected relationships and to summarize the data in novel ways that are both understandable and useful to the data owner”.

Data mining has matured into one way of addressing the growing availability of digital data and the gap between that data availability and the use of knowledge derived from them.

​

• It is worth discussing the relationship between data mining and knowledge discovery in databases (KDD) because of their similarity in processes and outcomes. The data mining process consists of six steps: business understanding, data understanding, data preparation, modeling, evaluation, and deployment while KDD is organized into data selection, data preprocessing, data transformation, data mining, and interpretation/evaluation. Thus, data mining is considered as part of the KDD, and data mining in the KDD process
• is a set of applications of specific algorithms for extracting patterns from preprocessed or “ready-to-data-mine” data.
• The successful application of data mining provides novel biomedical and healthcare knowledge which can be effectively used to support clinical decision making as well as administrative decision making in healthcare delivery.

​

How is data mining different from statistics?�

Under the aegis of scientific methods, the use of statistics has been the main data analysis method used in most scientific fields in recent history. According to Seifert, compared to statistics, “data mining represents a difference of kind rather than degree”. There are several significant distinctions between statistics and data mining.

• First, while statistics tends to first use conservative analysis strategies, data mining is more flexible about which methods are to be used in which order to mine data. Although data mining is fundamentally based on mathematics (as is statistics), many data mining approaches partially adopt heuristics, in addition to mathematics, to resolve real-world problems.

​

• Second, while statistics typically uses a sample of data drawn from a population, data mining typically uses data encompassing the entire population. Conversely, because data mining clusters data and discovers hidden patterns, data mining should, for the best results, use the entirety of the population data.
• Third, while statistics deals with numeric data only, data mining is able to handle multiple kinds of data (e.g. CT/MRI images, sounds, text, discrete data, etc.).
• Last, while statistics is hypothetico-deductive, data mining is inductive. In statistics, a hypothesis is built and then data is collected to test the hypothesis, but data mining, without a hypothesis, explores data that have been collected in advance, and discovers hidden patterns from data.

​

How does data mining in biomedical and healthcare applications differ from other applications?

• Medical data are primarily generated through the delivery of patient care. Therefore, mining of medical data inevitably is involved with privacy and legal issues. For this reason, data mining in the biomedical and health care fields differs considerably from that done in other fields. This key difference requires discussion of the uniqueness of data mining in the biomedical and healthcare fields.

How does data mining in biomedical and healthcare applications differ from other applications?

First, in many cases, the quality of data in the biomedical and healthcare fields is inferior to that found in other fields because of many reasons:

1. Medical data inevitably contains many missing values
2. Because hospital information systems or hospital databases are primarily designed for financial/billing purposes and not for medical/clinical purposes , it can be especially challenging to obtain high quality data for clinical data mining.
3. much of medical data (especially lab test results) are paper-based which, in turn, results in medical data that are often incomplete in terms of electronic availability

How does data mining in biomedical and healthcare applications differ from other applications?

Second, researchers in the healthcare arena must ensure patient privacy and handle patient data in accordance with HIPAA regulation.

In healthcare applications of data mining it is, however, equally important to ensure patient safety and maintain the security and confidentially of sensitive information as it is for researchers to make data sets available to other researchers

How does data mining in biomedical and healthcare applications differ from other applications?

Third, there are also legal considerations with the use of health care data.

The use of medical data mining could, for example, reveal previously unknown medical errors, which could, in turn, lead to lawsuits against healthcare providers.

For example, simple transcription errors or data entry mistakes might be found during the data preprocessing and/or mining evaluation and interpretation phrases.

New words

• Inevitably: ناچارا
• HIPAA: Acronym that stands for the Health Insurance Portability and Accountability

قانون انتقال و پاسخ گویی الکترونیک بیمه سلامت

​

• Inferior:نامرغوب،درجه دوم
• Emerge(v.):ظاهر شدن
• Novel:جدید- جذاب
• Unprecedented:چیزی که قبلا هرگز اتفاق نیافتاده یا انجام نشده- بی سابقه
• Hypotheses: فرضیه ها
• Interdisciplinary: بین رشته ای
• Observational: مشاهده ای
• Unsuspected: نامحسوس
• Mature(v.):به بلوغ رسیدن- کامل شدن
• Derive(v.):مشتق شدن
• Consists of:شامل می شود
• Deployment: استقرار

​

New words

• Interpretation:تفسیر
• Extracting:استخراج کردن
• Delivery: تحویل دادن
• Statistics:آمار
• Under the aegis of:تحت حمایت یا تحت نظارت چیزی یا کسی بودن
• Significant:قابل توجه
• Distinctions: تفاوت ها
• Conservative: محافظه کار- پیرو سنت قدیم
• Fundamentally:اساسا
• Heuristics: اکتشافی- پی برنده- یادگیرنده
• Encompassing:شامل شدن
• Conversely:به طور برعکس
• Cluster(v.): خوشه بندی کردن- دسته بندی کردن
• Discrete:گسسته
• Deductive:استنتاجی
• Inductive:استقرایی
• Lawsuits: دعوی ، دادخواهی

​