A clustering method based on rough sets and its application to knowledge discovery in the medical database.

This paper proposes a clustering method for nominal and numerical data based on Rough Sets and its application to knowledge discovery in the medical database. Classification is performed according to the indiscernibility relations defined on the basis of relative similarity between objects. The similarity is defined as a combination of two types of similarity measures: the Hamming distance for nominal attributes and the Mahalanobis distance for numerical attributes. Excessive generation of small category is suppressed by modifying similar equivalence relations into the same equivalence relation. An analysis of the meningoencephalitis diagnosis database was performed to validate this method. The result showed that this method could deal well with both types of attributes and discover the primary factors for diagnosis.

Automated discovery of positive and negative knowledge in clinical databases.

In this article, the characteristics of two measures, classification accuracy and coverage, were discussed. We showed that both measures are dual, and that accuracy and coverage are measures of both positive and negative rules, respectively. Then, an algorithm for induction of positive and negative rules was introduced. The proposed method was evaluated on medical databases, and the experimental results show that induced rules correctly represented expert knowledge. Several interesting patterns were also discovered.

Computer-assisted instructions for patients with bronchial asthma.

We produced computer-assisted instruction (CAI) software for bronchial asthma patients (asthma educational system with computer-assisted instruction; ASTCAI) to assist in self-management and avoid asthmatic attacks and death. ASTCAI is a question-and-answer program operating in a multimedia environment, and was evaluated from questionnaires which 33 patients were asked. Thirty-two patients could perform ASTCAI without any assistance. The responses of 31 patients (94%) indicated that they had no difficulty with manipulation, and 29 patients (88%) stated that the program was beneficial to control of their asthma. Elderly patients (over 65) required more time than younger adults. Emergency visits or admissions of at least 1 year after the first CAI trial decreased in eight out of 26 patients, while only two patients deteriorated compared to the previous year. Our results show that CAI is feasible for most patients, and through active self-learning CAI can improve motivation for self-management as well as supplement the physician's instructions.

Automated knowledge acquisition from clinical databases based on rough sets and attribute-oriented generalization.

Rule induction methods have been proposed in order to acquire knowledge automatically from databases. However, conventional approaches do not focus on the implementation of induced results into an expert system. In this paper, the author focuses not only on rule induction but also on its evaluation and presents a systematic approach from the former to the latter as follows. First, a rule induction system based on rough sets and attribute-oriented generalization is introduced and was applied to a database of congenital malformation to extract diagnostic rules. Then, by the use of the induced knowledge, an expert system which makes a differential diagnosis on congenital disorders is developed. Finally, this expert system was evaluated in an outpatient clinic, the results of which show that the system performs as well as a medical expert.

Automated acquisition of rules from clinical databases and its evaluation.

This paper presents an approach to induction of rules from databases using rough set model. The system was evaluated on three clinical databases, and induced results were compared with other conventional rule induction methods and medical experts' rules. The results show that the introduced results outperforms other methods, but that the description length of induced rules is a little short, compared with that of experts' rules, which suggests that experts' rules are combination of different kinds of reasoning, rather than simple classification.

Incremental learning of probabilistic rules from clinical databases based on rough set theory.

Several rule induction methods have been introduced in order to discover meaningful knowledge from databases, including medical domain. However, most of the approaches induce rules from all the data in databases and cannot induce incrementally when new samples are derived. In this paper, a new approach to knowledge acquisition, which induce probabilistic rules incrementally by using rough set technique, is introduced and was evaluated on two clinical databases. The results show that this method induces the same rules as those induced by ordinary non-incremental learning methods, which extract rules from all the datasets, but that the former method requires more computational resources than the latter approach.

Unaltered size of right ventricular infarct in dogs with right ventricular hypertrophy induced by pressure overload.

We investigated whether right ventricular infarct size is increased in the hypertrophied right ventricle. In dogs, we produced chronic pressure overload-induced right ventricular hypertrophy (RVH) and acute pulmonary stenosis (APS) with a similar right ventricular systolic pressure to RVH. Then, in an open-chest model, right coronary artery occlusion for 4 h followed by 1 h of reperfusion was performed in animals with RVH or APS or in controls. Regional myocardial blood flow was measured with nonradioactive colored microspheres. Compared with the control group, regional myocardial oxygen consumption in the right ventricle was increased in the APS but not in the RVH group. The risk area and collateral blood flow were comparable in the three groups. However, the infarct size relative to risk area was increased in the APS group (42.3 +/- 4.9%), but that in the RVH group (21.1 +/- 4.4%) was not different from controls (26.1 +/- 4.2%). Thus right ventricular infarction in the heart with RVH was not increased compared with the normal heart.

Knowledge discovery in clinical databases based on variable precision rough set model.

Since a large amount of clinical data are being stored electronically, discovery of knowledge from such clinical databases is one of the important growing research area in medical informatics. For this purpose, we develop KDD-R (a system for Knowledge Discovery in Databases using Rough sets), an experimental system for knowledge discovery and machine learning research using variable precision rough sets (VPRS) model, which is an extension of original rough set model. This system works in the following steps. First, it preprocesses databases and translates continuous data into discretized ones. Second, KDD-R checks dependencies between attributes and reduces spurious data. Third, the system computes rules from reduced databases. Finally, fourth, it evaluates decision making. For evaluation, this system is applied to a clinical database of meningoencephalitis, whose computational results show that several new findings are obtained.

Induction of expert system rules based on rough sets and resampling methods.

Automated knowledge acquisition is an important research issue in developing medical expert systems. While several methods of symbolic inductive learning have been proposed, most of the approaches focus on inducing some rules to classify cases correctly. On the contrary, medical experts also learn other information important for medical diagnostic procedures from clinical cases. In order to acquire both kinds of knowledge, we developed a program that extracts not only classification rules for differential diagnosis, but also other medical knowledge needed for diagnosis. This system is based on a diagnosing model of a medical expert system RHINOS, which diagnoses causes of headache and facial pain. We apply this program to the same domain and compared the induced results with expert rules. The results show that the combination of a rule induction method with resampling methods is effective to estimate the performance of induced results, especially when only small training samples are available without domain knowledge.

Impaired endothelium-dependent vasodilatory responses in hindlimb blood flow in dogs with congestive heart failure.

The vasodilatory response is known to be diminished in chronic heart failure. In the present study, we used dogs with chronic heart failure to investigate whether endothelium-dependent vasodilation in the resistance vessels in the hindlimb is decreased in chronic heart failure, and if so, to determine if endothelium-derived relaxing factor (EDRF) is involved in the mechanism. We induced heart failure in dogs by continuous rapid ventricular pacing. Under anesthesia, an internal carotid artery-femoral artery bypass system was established and perfusion pressure to the femoral artery was kept constant. Changes in hindlimb blood flow in response to various concentrations of acetylcholine (ACh), adenosine 5'-diphosphate disodium salt (ADP) and nitroglycerin (NTG) were examined in dogs with and without heart failure. The vasoconstrictive responses to L-NG-monomethyl L-arginine monoacetate (L-NMMA) were also studied. In dogs with heart failure, vasodilatory responses to both of the endothelium-dependent vasodilators, ACh and ADP, were decreased, while there were no differences in the responses to the endothelium-independent vasodilator NTG. The vasoconstrictive response to L-NMMA was diminished in heart failure. Thus, endothelium-dependent vasodilation in hindlimb resistance vessels is impaired in dogs with congestive heart failure. This impairment may be related to diminished EDRF/NO production or release in these vessels.

Induction of medical expert system rules based on rough sets and resampling methods.

Automated knowledge acquisition is an important research issue in improving the efficiency of medical expert systems. Rules for medical expert systems consists of two parts: one is a proposition part, which represent a if-then rule, and the other is probabilistic measures, which represents reliability of that rule. Therefore, acquisition of both knowledge is very important for application of machine learning methods to medical domains. Extending concepts of rough set theory to probabilistic domain, we introduce a new approach to knowledge acquisition, which induces probabilistic rules based on rough set theory (PRIMEROSE) and develop a program that extracts rules for an expert system from clinical database, using this method. The results show that the derived rules almost correspond to those of medical experts.

Coronary pressure-flow, pressure-function, and function-myocardial oxygen consumption relations in postischaemic myocardium.

STUDY OBJECTIVE: The aim was to study coronary pressure-flow, pressure-myocardial wall motion and myocardial wall motion-myocardial oxygen consumption relations in postischaemic myocardium with prolonged myocardial dysfunction (stunned myocardium) in comparison with normal myocardium. DESIGN: Regional myocardial wall thickening was measured with ultrasonic crystals, and postischaemic myocardial dysfunction was achieved by occlusion of a carotid-left anterior descending coronary artery bypass (15 min) and reperfusion (30 min). Coronary perfusion pressure was decreased in 10 mm Hg steps by constricting the bypass before and after producing postischaemic myocardial dysfunction. At each constriction step, coronary flow, regional wall thickening, and regional myocardial oxygen consumption in the area perfused by the left anterior descending artery were measured. EXPERIMENTAL MATERIAL: 12 anaesthetised open chest mongrel dogs, weighing 12-16 kg, were studied. MEASUREMENTS AND MAIN RESULTS: At the basal level, myocardial wall thickening of postischaemic myocardium was depressed compared to normal myocardium, at 18.5(SD 8.9)% v 1.3(7.1)%, p less than 0.01. With coronary stenosis, wall thickening gradually decreased at a coronary pressure below 60 mm Hg in normal myocardium, but remained unchanged until mean coronary pressure was reduced to 50 mm Hg in postischaemic myocardium. Myocardial wall thickening of postischaemic myocardium was always more depressed than normal myocardium. At any level of coronary pressure, coronary flow in postischaemic myocardium was not different from normal myocardium. There was no difference in regional myocardial oxygen consumption between normal and postischaemic myocardium at any level of coronary pressure. However, regional myocardial oxygen consumption in postischaemic myocardium was higher than in normal myocardium performing similar levels of myocardial wall thickening. CONCLUSION: The coronary pressure-function relation but not the pressure-flow relation changed in postischaemic myocardium after a 15 min coronary occlusion. Regional myocardial oxygen consumption was relatively increased in postischaemic myocardium.

[Cardiopulmonary exercise test in patients with DDD pacemaker: report of two cases with increased exercise capacity by decreased ventricular tracking limit rate setting].

We reported 2 patients with complete A-V block with a DDD pacemaker whose exercise capacity was increased by decreased ventricular tracking limit rate setting (VTL) of their pacemakers. Cardiopulmonary exercise test was used for estimating exercise capacity. Case 1: A 15-year-old girl complained of fainting. Her electrocardiogram (ECG) revealed complete A-V block (atrial rates 100/min, ventricular rates 39/min). After implantation of a DDD pacemaker and the VTL setting at 152/min, her bradycardia disappeared, however, she complained of dyspnea after a few minutes' walk. We performed symptom-limited cardiopulmonary exercise test with a motor-driven treadmill. When the pacing rate reached VTL (152/min), ECG suddenly changed to approximately 2:1 pacing (80/min) and the patient complained of dyspnea. Concomitant rapid increases in VE, VCO2 and RQ suggested that dyspnea was caused by the marked change in pacing rates on VTL. With the lowered VTL (110/min), there was no rapid increase in VE, VCO2 and RQ, and dyspnea subsided when the pacing rate reached VTL. At the same time, the peak VO2 and exercise time were increased by 15% and 8%, respectively. Case 2: A 47-year-old man complained of syncope. His ECG revealed complete A-V block (atrial rates 100/min, ventricular rates 33/min). After a DDD pacemaker implantation (VTL: 150/min), he experienced dyspnea while walking up the stairs in his office. Like in Case 1, when the VTL was lowered from 150/min to 110/min, both the peak VO2 and exercise time were increased by 11%.(ABSTRACT TRUNCATED AT 250 WORDS)

[Exercise tolerance in patients after successful percutaneous transluminal coronary angioplasty].

In 32 patients with successful percutaneous transluminal coronary angioplasty (PTCA), we performed treadmill exercise tests (TMET) before and about one month after PTCA to assess the correlation between the improvement in coronary artery lesions and exercise tolerance. Either the Bruce protocol (B: n = 12) or the modified Bruce protocol (MB: n = 20) was used; with the latter being applied to patients whose cardiac function seemed depressed. In 15 patients, oxygen consumption (VO2) was measured by analyzing the expired gases, 13 patients underwent exercise thallium-201 myocardial perfusion scintigraphy before and after PTCA, whose results were compared with those of TMET. In both B and MB protocols, the treadmill walking time was significantly prolonged after PTCA, compared to that before PTCA (B: 7.4 +/- 1.3 vs 9.5 +/- 1.9, MB: 11.4 +/- 3.5 vs 12.7 +/- 3.5 min). Heart rates (HR) and rate pressure products (RPP) were significantly increased after PTCA in both protocols (HR B: 139 +/- 18 vs 154 +/- 17, MB: 121 +/- 20 vs 137 +/- 19 bpm, RPP B: 26,500 +/- 5,600 vs 30,300 +/- 6,700, MB: 19,400 +/- 6,200 vs 22,700 +/- 6,600 mmHg.bpm), however, systolic blood pressure did not change significantly after PTCA in either protocol. While there was a significant improvement in VO2 after PTCA (21.6 +/- 6.3 vs 25.7 +/- 4.2 ml/kg/min), the O2-pulse remained unchanged. Thallium-201 myocardial scintigraphy revealed improvement of myocardial perfusion in 8 of the 13 cases examined.(ABSTRACT TRUNCATED AT 250 WORDS)

Immunochemical detection of human blood in feces.

We have developed a new immunochemical test for fecal occult blood utilizing enzyme-linked immunosorbent assay (ELISA) of human hemoglobin (HbAo) and transferrin (Tf) simultaneously. The ELISA had a sensitivity of about 15 ng/ml Hb, and the measurable range was 1.5-750 micrograms Hb per g feces. The stability of Tf in feces was greater than that of Hb. In 17 out of 18 patients with colon cancer, 8 out of 15 patients with colon polyps, and 11 out of 20 patients with upper-gastrointestinal disorders. The Hb and Tf values were more than 10 micrograms/g feces, in terms of Hb concentration. The ELISA for human fecal HbAo and Tf might be useful for the diagnosis of gastrointestinal disorders.