Ijraset Journal For Research in Applied Science and Engineering Technology

Abstract

The identification of ia health issue, disease, illness, or other condition that a person imay haveiis iknown ias ia disease idiagnosis.Sometimesiit may be extremely simple to diagnoseiaidisease,butiotheritimesiit may be more difficult. Large data sets iareiaccessible,butithe number ofitools i that can reliably iidentify ipatterns iand iforecast ioutcomes iis constrained. iTraditional imethods for diagnosing idiseases iarei labour-intensive iand iprone ito inaccuracy.When comparedito solely ihumanicompetence, ithe use of iAI iprediction iapproaches enables auto diagnosis and reduces detection mistakes. iFor ithe ipast ten years,i from January 2009 to December 2019, we have examined the recent literatureiin this publication.iA itotal iof i105 article iwere idiscoveredi in ithe ieight imost popular idatabases ithat iwere itaken into consideration by ithe iinvestigation. iThe imosti populariAIitechniquesiforimedicalidiagnosticisystems iwereicategorised after a thorough review of those papers. iWe ialso goioverinumerousidiseases and the irelevant iAI methods, such as Fuzzy Logic, Machine Learning,iand Deep Learning. iThe goaliof this study iarticle iis ito provide isome ikey iinsights iabout ithe various AI technique now being applied in ithe imedical industry,iparticularly in the areas iof heart disease iprediction, brain disease,iprostate, iliver and kidneyiillness.Last ibut not least, the report resolved issues.

Introduction

I. INTRODUCTION

The investigation iof sickness iin  theirealm iof ihealthcare A diagnosis is essential. A disease iis defined ias any factor or set iof events ithat produce suffering, ailment, dysfunction, or even death iin humans. Both ithe physical iand emotional well-being iof ia person can be damaged by idiseases, which also significantly alter ithe affected  individual's way of life. Pathological process iis ithe name for ithe study iof disease's causes. Clinical professionals must interpret any indications or symptoms ito determine ithe presence iof ia disease. The process of distinguishing a disease from its symptoms iand indicators in order ito determine its pathophysiology iis iknown ias diagnosis. The process iof determining which disease an individual has based ion their symptoms iand physical signs iis iknown ias diagnosis. iThe information gleaned from imedical history iThe knowledge needed for diagnosis iis obtained by physical examination iof ithe patient with ia imedical pathology. During this treatment, at least one diagnostic procedure, such ias ia imedical test, iis frequently carried out. I

A imedical professional  will go through ia procedure with  numerous phases ito accurately diagnose ia patient, allowing them ito gather ias much data ias they ican. iThe hardest part iof  treating ia patient iis determining their diagnosis, but it's ialso crucial for ia doctor ito get it right before moving forward with treatment The diagnosis procedure could be quite time-consuming iand difficult. The ihealthcare professionals gather empirical data ito determine ia patient's ailment iin order ito reduce ithe uncertainty iin imedical diagnosis. Due ito errors iin ithe diagnosis process, ithe patient imay not  receive ithe proper treatment for major health concerns orit imay be delayed. Unfortunately, not all doctors are experts iin every area iof medicine.

As  a result, an automated diagnostic system was required ithat combines ithe accuracy iof ithe machine with ithe benefits iof human  expertise. For ithe diagnosis process ito produce correct results at ia  lower cost, a proper decision support system iis required. iFor human experts, categorizing idiseases based ion multiple factors iis ia challenging undertaking, ibut iAI could help iidentify iand manage these types iof  situations. Currently, ia variety iof iAI imethods are being applied iin ithe field iof medicine to precisely diagnose illnesses.

iAI iis ia ikey component of computer science ithat helps make computers smarter. Learning iis ia must for all intelligent systems. iAI uses ia variety iof  learning-based methodologies, such ias deep learning ,machine learning, etc.

iA rule-based intelligent system, which  is one particular  AI technique  that  is important in the medical profession, offers  a set of if-then rules that serve as a decision support system in the field of healthcare. Intelligent systems are  being  gradually  supplanted in the medical industry by AI-based autonomous approaches  where human interaction is very little. Biological neurons connected in the brain serve as the basis for the creation of the neural network, also known  as an artificial neural network(ANN). It mimics the functioning of the human brain exactly. Each neuronal unit is connected to numerous other neurons in a network that resembles a bipartite graph. These systems receive automatic learning  and training.

Doctors and surgical specialists must spend a lot of time researching the potential outcomes and forecasts for health conditions .In some circumstances, ANN offers quick healthcare judgments where the systems can gather data, comprehend it, and identify element that will be crucial to prediction. In the medical area, deep learning—a branch of  machine learning that is likewise based ion algorithms—is used  to support specialists in the assessment  of  any ailment. better medical decisions as a result. Deep learning has applications in a variety of sectors, including drug discovery, medical imaging, genome analysis, and Alzheimer's disease detection. In this essay, our main areas of interest are fuzzy logic, machine learning, and deep learning—the three fundamental branches of  artificial intelligence. Deep learning is widely used in healthcare to diagnose breast cancer, which is the main  trend. It is evident from a recent study by a cancer institute that the accuracy of Automatic breast cancer is comparable to or higher than a human radiologist. Additionally, since AI is  constantly learning, it has a better possibility than before of producing results that are more accurate. The Internet of Medical Things, which enables the collection of healthcare data through IOT devices, is another key application  of  AI. AI-based software  senses the disease's signs to identify it even before it manifests. In  comparison to a skilled radiologist, neural networks can be trained to detect lung, breast, and stroke tumours in less time. By analysing medical pictures like MRIs ,CT scans, and ix-rays, various AI algorithms assist clinicians in making quick diagnoses of particular diseases. Since some diseases have remarkably similar symptoms, diagnosing a disease and giving the proper therapy are invariably challenging and difficult processes. Doctors may diagnose patients more precisely and recommend the best course of treatment when they use medical expert systems. Doctors can use AI techniques to not only identify the sickness but also categorise the various sorts of fatal diseases. Modern AI algorithms already assist physicians in setting up a thorough plan for managing sickness. They are also frequently utilised to enhance surgical robots that do extremely complex procedures. This work makes three contributions in total.

1. We start by outlining the current factors that have an impact on  the earliest detection  of  disease outbreaks.
2. In  the following section, we go through  how AI techniques  have  changed  for  initial disease diagnosis.
3. We present a comprehensive analysis for medical  diagnostic systems using a systematic review. We employ the widely used PRISMA I technique.
4. he diseases that were addressed, the AI approaches that were employed, the articles' study objectives, and their findings are then summarized for  each of  the articles that were chosen .Along with a thorough analysis of the publications under evaluation, we also provide future research directions.

II. IMPLEMENTATION

1. Data Collection: Gathering data is the initial stage. It is a highly important phase because quantity and quality have an impact on  the system's overall performance. It basically entails a procedure of acquiring information ion specific variables. The second step is data pre-processing: which comes  after data gathering. It involves transforming unusable raw data into information that can be used to make decisions. Data cleansing is another name for  this technique.
2. Select a Model: To incorporate pre-processed data into a model, one must select the right algorithm for the job
3. Train the Model: ML uses supervised learning to train a model to improve the precision of predictions or decision-making.
4. Assess the Model: A few parameters are required to assess the model. The i5.2 specified objectives serve as a guide for the parameters. Additionally, one must record how the new model performs  in  comparison  to  the old one.
5. Parameter Tuning: This stage may involve adjusting the initialization  settings, distribution, learning irate, performance and number of training steps.
6. Make Predictions: In order to compare the developed model to the real world, it is essential to make a prediction ion the test dataset. The model can be utilised for additional predictions if the result agrees with domain experts or  opinions that are close to it.

The following is a description of the fundamental procedures for disease detection using ML.

a. Gather test results and  patient information.

b. The feature extraction  procedure  selects characteristics that are helpful  for predicting disease.

c. After choosing the attributes, choose and process the dataset.

d. A pre-processed dataset can be used to assess the precision of disease prediction using the various categorization methods shown in the diagram.

e. Evaluation of  the performance of various  classifiers in order to choose the one with the highest accuracy.

In machine learning, all features are extracted by a domain expert to reduce the complexity of the data and to generate patterns that would be obvious to ML algorithms. The sole requirement is to make exact selections so that the testing data can be correct. However, deep learning-based techniques can extract features manually without human interaction. A domain expert is not necessary with this method for feature extraction. The use of deep learning for illness diagnosis systems is discussed in the section that follows.

III. RESULT

The graph compares the accuracy of the automated efficient illness detection system we intend to create using ML algorithms. Successful uses of AI in healthcare are the result of recent developments in AI technology. Even the question of whether AI expert systems will someday replace human doctors has become a prominent issue of discussion. We nevertheless take into account the reality that the AI expert system may help the human doctor to a better decision or, in some circumstances, may even take the place of human judgement. Numerous clinical data sets can be used to extract pertinent information using various AI algorithms. Additionally, AI techniques are educated in a way that allows them to be self-learning, error-correcting, and provide outcomes with a high degree of accuracy. The three AI methods used in disease diagnosis are the subject of this survey. With the PRISMA technique, we evaluate the effect of AI methods and their consistency on disease diagnosis in this review to reduce misdiagnosis errors. Our team came up with a search strategy to achieve the main objective.

Conclusion

By improving diagnosis procedures and detecting diseases in their early stages, AI in healthcare helps choose the best treatment strategy. Another important point to remember is that we looked into two AI techniques that are frequently utilized in healthcare, namely machine learning and deep learning, and we employed these two techniques to produce our results. Additionally, the impact of each AI technique was evaluated based on the frequency with which it was mentioned in papers. Using AI diagnostic criteria, the main medical fields that we looked at were cardiology, brain tumors, renal illness, diabetes, and liver disease.

References

[1] J. L. Scully, ‘‘What is a disease?’’I EMBO Rep., vol. i5, no. i7, pp. i650–653, i2004.RiLeaman,iR.iIslamajiDogan,iandiZ.iLu,i‘‘DNorm:iDisease in ameinormal-iizationiwith Ipairwiseilearning ito irank,’’ iBioinformatics, ivol. i29,ino. i22,ipp. i2909–2917,iNov. i2013. [2] N. Armstrong and P. Hilton,‘ ‘Doing diagnosis: Whether and how clinicians use a diagnostic tool of uncertain clinical utility,’’ Social Sci. Med., ivol. i120, pp. i208–214, Nov. i2014. [3] A.-L.Barabási,iN. Gulbahce, and J. iLoscalzo, ‘‘Network medicine: A inetwork-based approach to human disease, ’’Nature Rev. Genet., ivol. i12, no. i1, pp. i56–68, Jan. i2011. [4] R. H Scheuermann ,W. Ceusters, and B. Smith, ‘‘Toward an on to logical treatment of disease iand diagnosis,’’ Summit Transl. Bio in format., ivol. i2009, p. i116, Mar. i200 [5] P. Croft, D.G. Altman, iand J. J. Deeks, ‘‘The science of clinical practice: Disease diagnosis or patient prognosis? Evidence about ‘what is likely ito happen’ should shape clinical practice,’’ BMC Med., ivol. i13, no. i1, p. i20, i2015. [6] E. Choi, M. Bahadori , A. iSchuetz, W. F. Stewart, iand J. Sun, ‘‘Doctor ai: Predicting clinical events via recurrent neural networks,’’ in Proc. Mach. Learn. iHealthcare Conf., i2016, . i301–318. C.C. Lee, ‘‘Fuzzy logic iin control isystems: Fuzzy logic controller. ’’ IEEE I Trans. Syst., Man, Cyber n, ivol. i20, no. i2, pp. i404–418, Mar./Apr. i1990. [7] J. Yen iand R.Langari, Fuzzy Logic: Intelligence, Control, iand Information, ivol. i1. iUpper Saddle River, NJ, USA: Prentice-Hall, i1999. [8] H. D. Beale, H. B. Demuth, iand M. Hagan, Neural Network Design. Boston, MA, USA: PWS, i1996. [9] C.-H. Weng, iT. C.-K. Huang, iand R.-P. Han, ‘‘Disease prediction with different types of neural inetwork classifiers,’’ Tele matics Inform., ivol. i33, I no. i2, pp. i277–292, i2016. [10] M. Chen, Y.Hao, K. Hwang, L. Wang, iand L. Wang, ‘‘Disease prediction by machine learning over big data from healthcare communities,’’ IEEE Access, ivol. i5, pp. i8869–8879, i2017. [11] J. Betancur iand F. Commandeur, ‘‘Deep I learning for prediction of obstructive I disease from fast myocardial perfusion SPECT: A multicenter study,’’ JACC: Cardiovascular Image, vol. i11, no. i11, pp. i1654–1663, i2018. [12] M.S.van Mourik, A. Troelstra, W. W. van Solinge, K. G. Moons, iand M.J Bon ten, ‘‘Automated surveillance for healthcare-associated infections: Opportunities for improvement, ’’ Clin. Infectious Diseases, ivol. i57, I no. i1, pp. i85–93, i2013. [13] G.Luo,F.L. Nkoy, iP. iH. Gesteland, iT. iS. iGlasgow, iand iB. iL. iStone, ‘‘ A systematic review of predictive modeling for bronchiolitis,’’ Int. J. Med. Inform., vol. 83, no. i10, pp. i691–714, i2014.

Copyright

Copyright © 2022 Kavya A, Kavya K R, Nagashree T P, Mrs. Smitha P. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.