Bio-potential Electrode Sensing Processing and Bio-signal Transmission

138 Figure 4.22 Continuation c Electromyography EMG; d graphic hypnograms FONG et al 2011 139

4.3 Data Mining Applications and Patients’ Medical Records

Keeping patients’ medical records is long been practiced. This can be observed from the archives of visiting cards kept in clin- ics that contain the medical history and diagnosis of patients. The reason of keeping the medical history is to enable the track- ing of patients’ health and to serve as a warning to doctors about the existing medical issues of patients such as allergies. In addition, medical records show the pattern of patients’ health and may alert doctors if there is something suspicious that re- quires medical attention. This enormous amount of data is the reason many physicians refuse to transfer them into digital forms as it requires a lot of manpower and resources, not to mention the change required in the whole operations in dealing with electronic systems. The use of manual data entry can be observed from a visit to the clinic; first the administrator would retrieve the visiting card of a patient, pass it to the doctor, and new records will be made to the card after the diagnosis. Even though this procedure appears unproblematic, it will result in long-term problems the first being the preservation of enormous amount of records that consume a lot of space. Another com- mon problem is the intelligible hand writings that cannot be un- derstood by another physician and the loss of valuable medical records crucial to the descendants of patients or new physicians due to migration of passing of patients. A small clinic operates in suburban may serve up to hundreds of people, while this fig- ure can easily exceed hundred thousand in major health institu- tions. This results in a large volume of medical records when diagnosis details of each patient are combined with their per- sonal information. The medical data of every patient could easi- ly require megabytes or even gigabytes of data storage. It is easy to imagine the database of one hospital moreover the col- lective data bank of all citizens. This leads to the question of using a reliable database that allows the keeping of enormous 140 amount of information and at the same time enables them to be extracted easily. Those are the basic enquires made on data storage. Even though it is not comparable to the collective data of the Internet, the database is still comparatively large. For ex- ample, the epidemic condition of swine in fluenza in the Mexico City in 2009 resulted in the influx of more than 10 000 patients in health institutions in a day. This coupled with other cases such as suspected and confirmed A H1N1 infections has re- sulted in the built-up of vast amount of data in as short time. This paves the way for mining technology which enables clus- tered extraction from the database for the related information to study virus mutation such as in the case of influenza in a highly populated city. Similarly to search engines and Internet browsers, statistics are employed to rapidly retrieve specific information from an enormous data bank. A browser links to the pool of data on the Internet and to retrieve related information, a keyword is en- tered which subsequently extracts the data that contains the keyword at a very high speed. To aid and simplify the study, a keyword is used. For the computer, phrases are long words with spaces in between strings of words which form specific mean- ings. These words are comprehended by computers as ASCII codes abbreviation for American Standard Code for Infor- mation Interchange where characters are constituted by 7 bit codes. From this code, computers understand A as 1000001 which equals to the numerical figure 65. Therefore, words or phrases entered as keywords are recognized as a sequence of 7- bit codes or ASCII codes. In data mining, data is extracted from an enormous database through pattern or strings of code recog- nition. At the same time with advancement in increasing the ca- pacity for data storage, efficient statistic tools are developed for retrieving data with an incredibly fast speed. In current Internet browsers, by keying in the keyword “data mining” as an exam- ple, a staggering 21 million founds are obtained in just 0.18 se- 141 conds. The data retrieval mechanism is based on the enquiry of users that forms the search patterns. In general, 4 different pro- cesses are involved in data mining as shown in the charts of Figure 4.23. Even though every digital search applies the simi- lar technology, the retrieval of medical data may involve more complicated codes, especially in diagnosis images which are highly distinctive among patients. We have already known this when searching for images on the Internet where results are of- ten not related to the search keywords. Basically, four associa- tions are involved in the process of data mining: ¾ Associations: Connection is identified from data extrac- tion. For instance, diabetes is highly associated with obesity even not in every single case as diabetic patients are mostly overweight. ¾ Classes: Grouping of data based on categories, such as the group of diabetic patients. ¾ Clusters: Grouping of data based on logics, such as groups of patients based on locations or demographics. This type of association is especially helpful in analys- ing patterns of diseases. ¾ Sequential patterns: Predictions and patterns are made through the extraction of data. For instance, obese pa- tients can be predicted to suffer from other chronic dis- eases than non-obese patients. Next, a case study involving the digital records of a diabetic pa- tient will be examined. The record contains not only the diagno- sis history but other personal information such as demographic information. From the age of the patient, doctors are able to make precise assumption whether this patient suffers from Type 1 or Type 2 diabetes. This illustrates the usefulness of the data as a whole, even though most of the record is comprised of in- depth diagnosis details such as the amounts of glucose pre- scribed in its units. 142 Figure 4.23 the information retrieval process FONG et al 2011 The names of various nations, its respective measurement units and corresponding remedies are shown in Table 4.1. Besides patients’ in-depth diagnosis details as mentioned above, other forms of data such as images and audio files from X-rays and ECG diagnosis may be included. To extract data that involves ambiguous natures and patterns, specific procedures is required Elmaghraby, 2006, kantardzic 2011.