A prospective, mix-method study was carried out at the kidney transplantation clinic of Urmia University of Medical Sciences in 2017. At first, a prospective cross-sectional study detected all serious and significant potential drug-drug interactions (pDDIs) occurring in the prescriptions of all transplant recipients visited during a 2-month period. A checklist including demographic and medication data of recipients was used to collect the data. The Medscape DDI checker tool was used to determine all serious and significant interactions among the recipients’ active medications. Data were analyzed using descriptive statistical methods. Then, a comprehensive qualitative study was conducted using interviews (with 6 clinicians) and observations of the prescription workflow in the kidney transplantation clinic (for 2 months) in order to model a pDDI CDSS. Based on the common pDDIs, our clinical context, and considering internationally published guidelines, a pDDI CDSS was designed and its performance in detecting pDDIs was evaluated in comparison to the Medscape and using medication lists of 100 randomly selected patients.
conditions or events, these systems quickly examine the patient data instances which it needs the clinical personnel attention. In case of event presence, the notifications are given as alerts or reminders. One significant alert class focuses to recognize the errors in patient management potentially. Instances involve alerts which are modeled to recognize significant medication omission, inconsistency in medication prescription, significant laboratory test with patient’s care and condition. The present computer systems which recognize alerts and errors in occurrence are knowledge-based typically. The medical knowledge that is derived from medical experts are codified and presented as rules that are used to patient data to detect detrimental events and conditions [9, 10]. For a medical decision support range, rule-based alerting models has been modeled like automated dosing guidelines, finding specific adverse drug reactions , drug-allergy checking. Those models had been modeled for other medical conditions involving, treatment protocols monitoring of infectious diseases, deviations detection from those protocols , growth disorders detection [13,14] and significant event detection medically in chronic conditions management like congestive heart failure. Various advantages are comprised by rule-based systems. They depend on clinical knowledge and probably, they are helpful medically. The automation of these rules is simple and might be used towards patient data readily which are accessible through electronic format. Human expert's inputs are needed by the rule creation that takes more time and complex one. The constrained coverage of huge potential adverse events space is accessed by rules majorly that is composite adverse events particularly. Rule based alerting models are composite and inflexible to adjust acceptable performance medically in environments they are adapted. Because of unacceptable huge false alert rates, it is unusual for alert rules to be turned off . The produced alerts might be avoided because of huge false alert fatigue  even though those rules stay active.
The diagnostic design models developed by research communities have been converted into point-based risk score sheets in which the data are divided in group based on age groups of 40 – 84 years. The midpoints for these categories were noted with an reference values with several categories. The marginal category reference values were determined and observations were recorded in this perspective. A reference group (0 points) was established for each variable as the lower risk class— e.g. the 40-49 age class, women with systolic blood pressure < 120 mmHg, and so on . Palaniappan and Awang  have examined the comparison of different methods of data mining for the diagnosis of patients with heart disease. The comparison included naïve bays, decisiontree and the network of neural networks. The results showed that in the diagnosis of heart disease patients the naïve bays can achieve the highest precision . Clustering is one of the most popular clustering techniques, however the first selection of the center is an important problem which strongly affects its results. The study examined different methods of initial centric selection for K-means clustering technique is used for diagnosis of problems with range values, inlier, outlier, random attribute values and random row values. Back propagation algorithm wide range of improvements were proposed with weights for training in feed-forward neural network. Such scheme uses gradient descent technique for supervised learning in multilayer perceptron to increase the efficiency with various strategies .
Data mining tools used for classification, machine learning, artificial intelligence rather than a prior diagnosis of disease infection. Data mining tools can be used as essential tools to get results which can be used to support decisions. The clarity and accuracy of results are main reasons for choosing decisiontree algorithm in this work. First of all, tumor diagnosis paper records are collected from more than ten medical centres in Basra, Iraq. The first stage, these medical records transformed into electronic medical records (EMR). This stage makes the data entry took more than two months in the manual input of medical data. SQL Server data tool 2012 used to apply decisiontree and association rules algorithms on tumor diagnosis records.
Breast cancer  is a type of cancer originating from the breast tissue, commonly from the inner lining of the milk ducts or lobules supplying the ducts with milk. Breast cancer occurs in both men and women, although the former type is rare. It remains the number one form of cancer that woman are diagnosed with around the world. Even with enhanced treatment, the lack of early detection has put women at even higher risk of dying from this disease. Statistics reveal that there were 40,000 female deaths and 232,670 new cases recorded in the United States in 2014 .
Inci Zaim Gokbay graduated from the Department of Electronics Engineering, Isik University, in 2002, the M.S. degree with high honor from the Department of Electri- cal and Electronics Engineering of Bahcesehir University, in 2007 and Ph.D. degree in Biomedical Engineering from the Institute of Science, Istanbul University, in 2013. She was a Research and Teaching Assistant in the Department of Electrical and Electron- ics Engineering from 2004 to 2009, Lecturer in Vocational School from 2009 to 2014 in Bahcesehir University. She has been an Assistant Professor in the Department of Infor- matics, Istanbul University since 2014. She is the general coordinator of the “A Child for Life, Life for a Child” project founded by Istanbul Development Agency (ISTKA). Her research interest covers decision support systems, clinical decision support systems, machine learning, biological image processing.
With the continuous expansion of the scale of power grids, substation automation and unattended attention has been widely concerned . One of the key aspects of grid automation management is the ability to diagnose faults in the grid and adjust the corresponding protection devices . At present, many artificial intelligence algorithms are applied to the field of power network fault diagnosis, and achieved good application effect .
The inaccuracy in the results was due to images where the cysts appeared along (i.e. too flat) and not “bubble-like”. These were mistaken for the normal structures in healthy liver and not flagged by the NN classifier as cysts. As for fatty liverdiagnosis, the inaccuracy was due to the features derived from FOS MRI images, intensities of the image blocks selected may contain non-liver tissues. Thedrawback of the classification method is the requirement of varying parameter settings for different liver classes. In order to apply this method extensively in clinical usage, standardization of the parameters is strongly recommended.
ANA is usually assessed by the indirect immunofluo- rescent (IIF) method using HEp-2 cells. The target antigens of ANA in type 1 AIH contain a heterogeneous group of structures, such as nuclear DNA, nuclear structural and functional proteins or centromeres : different immu- nofluorescent staining types including homogeneous, speckled, nucleolar and discrete speckled patterns are shown on HEp-2 cells . We previously revealed that the most common immunofluorescent staining type in patients with AIH type 1 was a homogeneous pattern . Notably, heterogeneous nuclear ribonucleoprotein (hnRNP) A2/B1, which belongs to RNA-binding protein, was recently identified by Ballot’s group as one of the liver-specific nuclear antigens in type 1 AIH . hnRNP A2/B1 is Table 1 Clinical significance of autoantibodies in liverdisease
In order to compare the performance of the proposed LWA classifier, we refer to Figure 5 in which we plot the AUC- ROC of our proposed scheme with and without taking examples that the model has abstained from classifying. The AUC-ROC of NN and SVM are also plotted as a reference. It can be noticed that the AUC-ROC of LWA on accepted (not abstained) examples is always better and its overall AUC- ROC is comparable to the AUC-ROC of conventional SVM. As discussed earlier, the increase in abstention penalty decreases the fraction of abstentions. For the low value of = 0.1, the LWA classifier rejects all examples whereas for high value of = 0.5, no abstentions take place. Furthermore, As expected, when the fraction of absetention drops to zero for large values of , the performance of LWA becomes comparable to a conventional SVM. However, for = 0.12, the fraction of abstention is equal to 53% with an AUC-ROC of 95. For = 0.17, the fraction of abstention is equal to 21% with an AUC-ROC of 93 and when = 0.3, the fraction of abstention is equal to 4% with an AUC-ROC of 91. This shows that the LWA classifier achieves near perfect classification AUC-ROC if it is permitted to abstain from producing labels for 53% test examples. LWA has automatically detected that its confidence for correctly predicting these examples is low and thus abstained from these misclassifications. This shows the effectiveness of the proposed approach in comparison to conventional classification techniques. The python implementation of the LWA classifier runs in under 5-6 minutes on a laptop with an Intel core i5-3317U 1.70 GHz processor and 4 GB RAM. B. Re-Evaluation of rejected examples by medical expert The 7 test cases from 53% of data for which the LWA classifier abstained from generating labels were given to an experienced radiologist (DS) for re-evaluation. The radiologist was not provided the original labels for these cases and was asked to diagnose these cases. It is interesting to notice that, for 3 out of these 7 cases, the radiologist generated labels were different from the original labels. These cases are shown in Figure 6. This shows that the abstentions produced by the proposed LWA method were indeed difficult to classify even for trained medical experts. These cases can refer to further testing through elastography, CT or biopsy. These results clearly indicate the effectiveness of the proposed approach. C. Comparison with medical expert’s analysis
The porphyrias are a group of rare metabolic disorders that result from defects in heme biosynthesis. Erythropoietic protoporphyria (EPP) is the most common inherited porphyria in children and is diagnosed in most individuals after the onset of cutaneous manifestations. Hepatobiliary disease affects the minority of individuals with EPP and usually manifests in patients with an established diagnosis of EPP. We report on a classic but rare case of EPP that masqueraded as cholestasis. An 8-year-old boy was referred to the Hepatology Clinic after an abrupt onset of jaundice with a longstanding history of dermatitis. The diagnosis of EPP was established with liver biopsy, which revealed dense, dark-brown pigment in hepatocytes and Kupffer cells that, on polarization, displayed bright-red birefringence and centrally located Maltese crosses. Plasma total porphyrins and erythrocyte protoporphyrin were elevated and confirmed a diagnosis of EPP. We hope to raise awareness of this diagnosis among pediatricians, hepatologists, and pathologists and increase the consideration of EPP in patients with cholestatic liverdisease and chronic dermatitis.
Fig. 3 shows decisiontree for output of algorithm J48 and case of signal processing through Db3 wavelet transform. Tracing a branch from main node to leaf leads to a state of compressor and decryption of information in each tree branch as if-then sentences develops essential regulation for fuzzy classification of failures of compressor. Evidently, for each tree the top node is best node for classification. Other features in nodes of decisiontree are arranged based on decreasing significance. In decision trees, a category of spectrum features is placed which is significant for correct classification. Those features that lack essential significance for classifying failures of compressor are excluded from the model. Data distribution level for set of features that compose a class is represented by in-parenthesis numbers within decisiontree and before condition of compressor at that class. In this case, the first number in parenthesis suggests number of data which could be classified properly through a set of features. The second number refers to number of data that are incorrectly placed in that class. If first number in parenthesis is much smaller than all data included in the model to suggest a failure, that class could be excluded from final model.
Today, the diagnosis of Alzheimer’s disease (AD) or mild cognitive impairment (MCI) has attracted the attention of researchers in this field owing to the increase in the occurrence of the diseases and the need for early diagnosis. Unfortunately, the nature of high dimension of neural data and few available samples led to the creation of a precise computer diagnostic system. Machine learning techniques, especially deep learning, have been considered as a useful tool in this field. Inspired by the concept of unsupervised feature learning that uses artificial intelligence to learn features from raw data, a two-stage method was presented for an intelligentdiagnosis of Alzhei- mer’s disease. At the first stage of learning, scattered filtering, an uncontrolled two- layer neural network was used to directly learn features from raw data. At the second stage, SoftMax regression was used to categorize health statuses based on the learned features. The proposed method was validated by the data sets of Alzheimer’s Brain Images. The results showed that the proposed method achieved very good diagnos- tic accuracy and was better than the existing methods for brain image data sets. The proposed method reduces the need for human work and makes it easy to intelligently diagnose for big data processing, because the learning features are adaptive. In our experiments with the Alzheimer’s Disease Neuroimaging Initiative (ADNI) data, a dual and multi-class classification was conducted for AD/MCI diagnosis and the superiority of the proposed method in comparison with the advanced methods was shown. Keywords: Alzheimer’s disease, Sparse filtering, Unsupervised feature learning, Intelligentdiagnosis, SoftMax regression
The eye is one of the main human organs which links to the inner body and is continuously exposed to a harsh outside environment where it is continually in contact with pathogenic airborne organisms. Although the eyelid may help to protect the eye, the warm, moist, enclosed environment between the conjunctiva and the eyelid also enables contaminating bacteria to establish an infection. The number of organisms responsible for infection of eye is relatively small, but they can proliferate rapidly and cause serious and irreversible damage to the eyes, which makes rapid diagnose essential (J. W. Gardner, Boilot, & Hines, 2005). Usually this kind of diagnosis is based on the study of symptoms, such as changes in bodily appearance, feel or functions etc (Dutta, Hines, Gardner, & Boilot, 2002). Since different diseases produce distinctive smells sometimes specific characteristic odors, smelling the bacteria becomes a significant part of diagnosis.
this research proposes an intelligentdecision support system for acute lymphoblastic leukaemia diagnosis from microscopic blood images. A novel clustering algorithm with stimulating discriminant measures (sDM) of both within- and between-cluster scatter variances is proposed to produce robust segmentation of nucleus and cytoplasm of lymphocytes/lymphoblasts. Speciically, the proposed between-cluster evaluation is formulated based on the trade-of of several between-cluster measures of well-known feature extraction methods. the sDM measures are used in conjuction with Genetic Algorithm for clustering nucleus, cytoplasm, and background regions. subsequently, a total of eighty features consisting of shape, texture, and colour information of the nucleus and cytoplasm sub- images are extracted. A number of classiiers (multi-layer perceptron, Support Vector Machine (SVM) and Dempster-Shafer ensemble) are employed for lymphocyte/lymphoblast classiication. Evaluated with the ALL-IDB2 database, the proposed sDM-based clustering overcomes the shortcomings of Fuzzy C-means which focuses purely on within-cluster scatter variance. It also outperforms Linear Discriminant Analysis and Fuzzy Compactness and separation for nucleus-cytoplasm separation. the overall system achieves superior recognition rates of 96.72% and 96.67% accuracies using bootstrapping and 10-fold cross validation with Dempster-Shafer and SVM, respectively. The results also compare favourably with those reported in the literature, indicating the usefulness of the proposed sDM-based clustering method.
classification tasks, thus we have selected sets of which the class values are nominal , . Selection of the sets further depended on their size, larger data sets generally means higher confidence. Here, we choose six algorithms namely, Bagging algorithm, logistic model trees algorithm, REP tree algorithm, NaiveBayes algorithm, Dagging algorithm, KStar algorithm are used for comparison. A comparison is based on sensitivity, specificity and accuracy by true positive and false positive in confusion matrix. B. Naive Bayes Algorithm:
Data mining is the extraction of hidden predictive information from large databases . It is a powerful new technology with great potential to help companies focus on the most important information in their data warehouses. Data mining is actually part of the knowledge discovery process. The Knowledge Discovery in Databases  process comprises of a few steps leading from raw data collections to some form of new knowledge. Classification in data mining find a model for class attribute as a function of the values of other attributes. Classifier generates meaningful description for each class that are used to classify instances of the given dataset . To Classifying credit card transactions as legitimate or fraudulent is an example of classification. There are several approaches for the classification like neural network, SVM, decisiontree etc. In this paper decisiontree is illustrated as classifier. Decisiontree is flow like structure . Decisiontree induction is the top down process. At the top the root is selected using some attribute selection measures like information gain, gain ratio, gini index etc. During induction of the tree attribute is selected by this attribute selection measures . Although decisiontree is induced by various algorithms, but sometimes it happens that it generates unwanted & meaningless rules as it grows deeper, it is called overfitting. Pruning is needed to avoid large tree or problem of overfitting . Pruning means reducing size of the tree that are too larger and deeper. The problem of noise and overfitting reduces the efficiency and accuracy of data.
Finite tree automata are generalizations of words au- tomata structured in trees and they keep their essen- tial properties: they have good logical and ensembles’ properties, along with effective algorithmic. These characteristics allow for a large number of implemen- tations, especially as far as typing and programs anal- ysis are concerned. Also, they allow defining regular trees languages with an operational semantic, applying a general frame for XML languages and supplying a validation execution environment while serving as ba- sic tool for the static analysis (proofs, decision-making procedures in logic).
In cases where two or more articulatory attribute questions result in subsets with equal homogeneity, a prioritised reference list is used. This list ranks articulatory attributes in order of importance, based on how common their occurrence is in X-SAMPA. This was done to prevent the decisiontree from having excessively long or thin branches. For articulatory attributes which have the same frequency of occurrence, those with the greatest eﬀect on visual speech were prioritised. Section 2.2 and Appendix B assisted in gauging the magnitudes and number of features involved in the sound’s articulation, which were used to identify it ranking of visual importance. This was intended to help better the decisiontree splits by grouping attributes which are more visually pronounced. A holistic view of the decision tree’s algorithm is shown in Figure 6.2. This framework is used for each of the algorithms discussed in this chapter. Figure 6.2 starts with the root node of the tree, which represents all the dynamic visemes for one of the seven TBCLRWr features. A recursive procedure is then entered, which runs one of the three node split- ting algorithms, discussed later in this Chapter (the three algorithms are abbreviated to MD/KM/SS-CART in Figure 6.2). The framework for the decisiontree also includes con- trol of meta parameters related to stopping criteria. These include node occupancy and a minimum improvement in deviation between parent and child nodes. These parameters are important in preventing over fitting during training.