Quiz1

Neural Networks
Student’s Name
Institutional Affiliation

Abstract
The paper delineates neutral networks broadly and describes their purpose primarily in medicine. As per research, neural networks dig deeper and shape the functional aspects of computational features of the brain. The paper goes further and touches on the applications of neural networks particularly in medicine. Also, the text elucidates the notion of inference in general terms in addition to the concept of the Bayesian network model. In this specific part, it explains the theory that forms the basis of Bayesian algorithm and highlights some of its applications in clinical decision support systems. Moreover, the paper clarifies the notion of decision trees and supports it with an illustration which can help in breast cancer diagnosis. Furthermore, in the paper is an inference engine which can be utilized when diagnosing simple pyelonephritis with all symptoms mentioned. In conclusion is the definition of some unique set of related objects which describes uncomplicated pyelonephritis in a patient. The set is illustrated as per the textbook.
Keywords: Neural Networks, decision trees, inference, Bayesian network model

Neural Networks
Question 1
Neural networks are estimation models established on the mathematical paradigm that opposed to old computing have an operation and structure that is similar to a brain. They are referred to connectionist systems, adaptive systems or parallel distributed systems since they are made up of a series of interrelated processing fundamentals that work in a parallel manner.

They lack a unified control in a conventional sense, as all interrelated processing essentials alter or adapt concurrently with information flow and adaptive regulations. One major original objective of the neural network is to comprehend and structure the functional aspects of computational characteristics of the brain for performing cognitive practices, for instance, concept connotation, sensorial perception, and learning. There are numerous applications of neural networks in medicine. They are used in medical image analysis, electronic signal analysis, and in radiology. In addition to that, neural networks are used in medical data mining and pharmaco-epidemiology (Papi, Molnar, Schaefer, Dombovari, Tulassay,& Feher, 2011).
Question 2
Inference in Bayesian algorithms is the practice of computing a probability distribution of choice for example P (A/B=True). In other words, it is the process of estimating a posterior distribution of a variable when evidence is given. This term is used interchangeably with another name: queries. Examples of inference practice include identifying the most likely disease given the symptoms, the likelihood of a system to fail when the current state is known, and how stocks will behave in future given the current behavior. A Bayesian network, on the other hand, is a compacted, graphical model. A lot of research has been conducted in the last three decades to generate useful algorithms for inference in this networks. These algorithms are categorized into three: the first category is based on the concept of case analysis or conditioning. Some algorithms of this type try to reduce the network to resemble a tree structure so that it becomes easy to manage or control. Others attempt to break the network into smaller bits that are solved concurrently. The second model is centered on the concept of tree clustering and maximizes on the manageability of inference concerning three structures. The third class of algorithm is based on the idea of variable elimination. This category of algorithms take a probabilistic model over m elements and lessen them to m-1 while maintaining the capability of the model to give answers of interest. Bayesian algorithms are used in Prediction of Heart Diseases (CCDSS) by employing the Naïve Bayes data mining algorithm
Question 3
Almost every aspect of medicine such as classification or diagnosis requires decision making that should be made efficiently and reliably. Theoretical simple models of decision making having the likelihood of automatic learning are the most suitable for executing such tasks. This is where decision trees come in. This is a technique which is efficient and dependable in decision making and provides a high exactness with a simple illustration of consolidated knowledge, and it has been employed in various areas of medical decision making. While making decisions, problems are usually created as a decision tree (cite). By configuring the matter in this manner and accumulating numerical values to various branches, the issue can be investigated and the optimum selection established. Different branches of the tree, the likelihood (probability) of a particular outcome arising and the importance the decision-maker places to that result. Usually, the probabilities employed to inform decision models arise from research data, and since probabilities are used, the levels of uncertainty that come with the choices and outcomes are minimized. Thus, for a given decision problem, decision trees are ideal. The following is a tree diagram for diagnosing breast cancer.

Question 4

The following figure shows an inference engine for diagnosing simple pyelonephritis. The major symptoms associated with is disease are flank pain, fever, chills and vomiting. This helps to ascertain the likelihood of possible disease as most diseases share the same symptoms.
Question 5
By activating all four nodes, (flank pain, chills, fever, and vomiting), the probability of simple pyelonephritis is maximized. Apparently, each of the four nodes might be linked to other ailments, for instance, cystitis since its major symptoms include vomiting, fever, and chills. To make a diagnosis which has high accuracy, the posterior probability of the diseases in question are required so that they can be put into consideration. Once both possibilities are available, they can be compared in order to select the condition that is more probable given the symptoms. The occurrence of simple pyelonephritis in all population is taken to be 10%. This depicts the prior knowledge before any symptoms are observed.
P (cystitis) = 0.1
For cystitis, the posterior probability of simple pyelonephritis can be employed to obtain the posterior of simple pyelonephritis.
P (cystitis) = 0.011
P (simple pyelonephritis) = 0.988 (this is because all symptoms are put into consideration)
Patients major=flank pain, vomiting, chillsThe same applies to the minor criteria to make a set:
Patients minor= {fever}
|FINDINGS| =2
|PATIENT’S-CRITERIA-MAJOR| =3
|CLINICAL-FINDINGS ∩PATIENT’S-CRITERIA-MAJOR|≥2
The above arrangement is the set theory for the condition described. This is useful as it helps individuals understand the logic of probability and many other aspects that are helpful in solving phenomena related to diagnosis.

References
Papik, K., Molnar, B., Schaefer, R., Dombovari, Z., Tulassay, Z., & Feher, J. (2011). Application of neural networks in medicine-a review. Medical Science Monitor, 4(3), MT538-MT546.