Artificial Neural Network (ANN)
|Topics:||Artificial Intelligence, Innovation|
Technology advancements have made it easier for researchers to carry out their studies with aims of solving daily life problems. Artificial Neural Network (ANN) tools are computational models used computer science and other research disciplines and are connected to neural units to observe biological brain’s axons (Krogh, 2008). Neurophysiologist McCulloch Warren first developed the artificial neuron in 1943 and the technology has been enhanced with the technological advancements through research. There are many models used in medical research to identify and solve problems. Artificial Neural Network (ANN) is a computerized model based on structures and function of neural network systems. The Artificial Neural Network (ANN) is a data processing tool that computerizes information to biological nervous system like processing information from the human brain. Artificial Neural have many inputs connected to one output. Learning Vector Quantization (LVQ) is a prototype-based supervised classification algorithm that uses a competitive network in supervision learning through classification(Hammer & Villmann, 2002).
The Artificial Neural Network (ANN) model is used to process information and is interconnected to neurons to solve certain problems. An ANN operates through a learning process and critical in deriving useful information from complex data. The models are nonlinear statistical modeling data tools that distinguish between a complex relationship in inputs and outputs. Neural network is used to observe different sets of data before computation and helps researchers to arrive at solutions(Mao, 2011). The tools are cost-effective and ideal for computing functions and analyzing the data. Neural networks are used to provide algorithms to create, visualize, train and stimulate the neural networks. Researchers can perform regression, clustering, classification and dimensionality.
The interconnected neural units are computed using a summation function and self-learning. The interconnected neural networks are made up of numerous layers that are cubic to allow signal transverse from the initial input to the last output. The first layers consist of input neurons that send data via synapses to the next layer of neurons until it reaches the output neurons through the synapses(Krogh, 2008). The most advanced ANN is the dynamic neural network. Neural networks are used to solve daily problems just like human beings. Likely, the advances dynamic is more efficient and can operate with millions of neural used in brain research to stimulate new patterns in the neural networks (Mao, 2011).
The technology is used in many medical studies such as recognizing speech and computerizing visions. ANN consist of interconnection patterns arranged between the neuron’s layers, interconnection weights that are used in the learning process and lastly the activation functions which are responsible for the conversion of neuron’s input to activation output. Artificial Neural Networks (ANN) is used in data extraction and detection of complex trends beyond human ability (Mao, 2011). ANN has adaptive learning ability and can accomplish different tasks based on the computed data. Besides, ANN tools have self-organization that enables them to represent the data obtained during the learning process. Neural networks are more efficient in problem-solving as compared to conventional computers. Conventional computers are instructed by human beings while neural networks process information just like people (Tiruan, 2011).
Learning Vector Quantization (LVQ) is one of the artificial neural networks represented by prototypes defined in future space in the observed data. LVQ creates prototypes that can be easily interpreted by experts. Learning Vector Quantization (LVQ) classifies input data through assigning them to the same class with the output. Vector quantization is used to establish encoding schemes through algorithms. LVQ has both competitive and linear layers which have the different task. The competitive layer is used in classifying input vectors while the linear layers are used to transforms competitive layers to target classification. Learning Vector Quantization is used in various ways including speech recognition, optimization of problems, and image compression among others(Hammer & Villmann, 2002). Artificial Neural Network permits a nonlinear relationship between the variables and important in pattern recognition(Tiruan, 2011).
The use of Artificial Neural Network (ANN) in the Medical Diagnosis
The advancement in medical research through the use of ANN has made it possible for medics to diagnose various diseases through computer-aided diagnostic system. Computer aided medical diagnosis (CAMD) has been used for many some decades to solve health problems among patients (Lin, Vasilakos, Tang, & Yao, 2016). The system is used to enhance decision-making process, extracting and visualizing complex clinical diagnosis. Computer aided medical diagnosis technology has led to more research with attempts to improves its efficiency. CAMD is used to detect diseases, classification, and testing drug compatibility among other medical uses(Lin et al., 2016).
Computer aided medical diagnosis are integrated with artificial neural networks. The science of machine learning led to advancement in artificial neural networks system that includes models that enhance the functionality and structure of biological ANN. The ANN depends on multiple inputs that are interconnected to neurons to promote the exchange of information between the neurons(Lin et al., 2016).
The ANN interconnections are integrated with numerical weights to enable adapt to various inputs. Computer aided medical diagnosis uses ANN to approximate any arbitrary function through learning the observed data(Lin et al., 2016). The effectiveness of ANN depends on the type of models and complexity, algorithms learning and robustness of ANNs in the engineering systems. ANNs applications in medicine through proper algorithms have prolonged lives of many patients. Artificial neural networks are widely embraced in medical diagnosis and enhancement of decision making. The technology is used in the diagnosis of different types of cancer. Cancer detection machines such as HLND use ANNs to give an accuracy of cancer diagnosis and management. The systems have patient’s data, and the installed ANNs can detect and predict colorectal cancer by providing efficiency.
Algorithms of Artificial Neural Network (ANN) in the Medical Diagnosis
Integration of ANN into Computer aided medical diagnosis (CAMD) helps in health monitoring, and doctors can monitor the conditions of the patient’s from any location. The system access patient’s and process the data by using the algorithm. Artificial neural networks are used to solve medical problems by diagnosing diseases such as cancer, cardiovascular diseases, diabetes and urinary diseases among others. Artificial neural networks are used to diagnose through learning the patterns in the symptoms. The learning algorithms enable medical practitioners to analyze the disease through the computer-aided medical diagnosis(Lin et al., 2016). The input data in ANNs are used to analyze blood and urine samples to detect bacterial and viral infection among other pathogens. ANN’s learning algorithms diagnosis diabetic patients, cancer, leukemia and tuberculosis among other diseases(Lin et al., 2016).
an A-level paper for you.
Artificial neural network is used to diagnose many diseases. In cardiovascular and cancer diseases, doctors have been able to diagnose millions of patients using ANN integrated into computer-aided medical diagnosis (CAMD) (lin et al., 2016). Terminal diseases such as cancer and cardiovascular diseases need an accurate diagnosis, and ANN is used to different cancer types at their early stages(Al-Shayea, 2011). Neural networks have been successful in the diagnosis of cancer through ANN oncology. The process relies on the algorithms and input patterns as shown. The medical diagnoses include both training and diagnosis process.
The ANN medical diagnoses start with training or learning process which starts from selection of target disease that is related to particular classification (Al-Shayea, 2011). The disease parameters and symptoms are determined and laboratory results examined through proper analysis. There is a database in the neural networks to validate values and verify the data. The database should be fed with the accurate information to avoid inaccuracy. The data should have the patient’s historical health record, biochemical health analysis, symptoms and previous treatment(Lin et al., 2016). Positive results obtained from the neural networks leads to the next step that involves medical diagnosis. During the diagnostic process, neural networks process the patient’s data by determining the probable diagnosis. The final process in diagnoses is the physician’s decision. The physician should use the ANN to make a decision, treatment, and management of the diseases(Al-Shayea, 2011).
Learning algorithms in ANNs are imperative for training through a learning process via adjusting the weights of the inputs(Arbib, Ballard, Bower, & Orban, 2014). The database should have a reliable pattern, and all the information will be regarded as one input to the neural network. The algorithms are used to recognize the patterns and input values classified (Al-Shayea, 2011). The weights are first initialized with random values before matrix computation which also requires learning rates and momentum. The momentum value should be lower than the learning rate for effective performance. The detection, diagnoses, and prognoses start from data entry. The patient’s data must be computerized before the evaluation process. According to Amato et al (2014), the weight is controlled by the learning rate while momentum acts as stabilizers during the diagnosis.
The patient’s data are assessed using different algorithms (Arbib et al., 2014). There most algorithms include;
- K-nearest neighbors based algorithms
- Fuzzy logic based algorithms
- Deep learning based algorithms
- Artificial neural network based algorithms
- Decision tree based algorithms
K-nearest neighbors based algorithms are used for classifying different cases and referring to the nearest learning case in the future space. The cases are mixed before constituting a specific case set for example in cancer diagnosis (Al-Shayea, 2011). The data in the space are further classified into k disjoint subsets, and the cases are represented by one particular case. The Artificial neural network based algorithms train the historical data of all patients in the systems(Arbib et al., 2014). The algorithm automatically partitions brain-computer interface into various groups. Self-organizing map algorithm has been used to detect various diseases including interstitial lung diseases. SVM and Severity prediction algorithms are also used in medical diagnosis to identify and predict various diseases(Arbib et al., 2014).
In conclusion, artificial neural networks have been successful in the medical field, and there is the easy management of various diseases. Doctors can use the model in conditions where their knowledge and brain are unable. ANN is more accurate than conventional computers, and diagnoses of diseases have been made easy and effective. ANNs are used to diagnose complex diseases by identifying their symptoms and treatment suitable methods. Computer aided diagnosis and ANN prevents misdiagnosis.
- Amato F, González-Hernández J, Havel J. (2011) Artificial neural networks combined with experimental design: a “soft” approach for chemical kinetics. Talanta. 93: 72–78, 2012.
- Al-Shayea, Q. (2011). Artificial Neural Networks in Medical Diagnosis. … Journal of Computer Science Issues (IJCSI), 8(2), 150–154. https://doi.org/1694-0814
- Arbib, M., Ballard, D., Bower, J., & Orban, G. (2014). Neural Networks Algorithms , Applications. Neural Networks (Vol. 7). https://doi.org/10.1016/0893-6080(94)90073-6
- Hammer, B., & Villmann, T. (2002). Generalized relevance learning vector quantization. In Neural Networks (Vol. 15, pp. 1059–1068). https://doi.org/10.1016/S0893-6080(02)00079-5
- Krogh, A. (2008). What are artificial neural networks? Nat Biotechnol, 26(2), 195–197. https://doi.org/10.1038/nbt1386
- Lin, D., Vasilakos, A. V., Tang, Y., & Yao, Y. (2016). Neural networks for computer-aided diagnosis in medicine: A review. Neurocomputing, 216, 700–708. https://doi.org/10.1016/j.neucom.2016.08.039
- Mao, J. (2011). Why artificial neural networks? Artificial Neural Networks in Medical field, 29, 31–44. https://doi.org/10.1109/2.485891
- Tiruan, J. S. (2011). Artificial Neural Network. Neuron, 36, 1–27. Retrieved from http://leenissen.dk/fann/wp/
Offered for reference purposes only.