Comparative Analysis of Neural Networks in the Diagnosis of Emerging Diseases based on COVID-19

Abstract

Dermatological diseases are frequently encountered in children and adults for various reasons. There are many factors that cause the onset of these diseases and different symptoms are generally seen in each age group. Artificial neural networks can provide expert-level accuracy in the diagnosis of dermatological findings of patients with COVID-19 disease. Therefore, the use of neural network classification methods can give the best estimation method in dermatology. In this study, the prediction of cutaneous diseases caused by COVID-19 was analyzed by Scaled Conjugate Gradient, Levenberg Marquardt, Bayesian Regularization neural networks. At some points, Bayesian Regularization and Levenberg Marquardt were almost equally effective, but Bayesian Regularization performed better than Levenberg Marquard and called Conjugate Gradient in performance. It is seen that neural network model predictions achieve the highest accuracy. For this reason, artificial neural networks are able to classify these diseases as accurately as human experts in an experimental setting.

Keywords

• Bayesian Regularization Neural Network
• COVID-19
• dermatological findings
• Levenberg–Marquardt Neural Network
• Scaled Conjugate Gradient Neural Network

References

1. [1] M.T.Hagan,M.B.Menhaj,TrainingfeedforwardnetworkswiththeMarquardtalgorithm.IEEETrans.NeuralNetw.,5(1994),989–993.
2. [2] MuratKayri,PredictiveAbilitiesofBayesianRegularizationandLevenberg–MarquardtAlgorithmsinArtificialNeuralNetworks:AComparative Empirical Study on Social Data, Math. Comput. Appl. 21, (2016), 20, DOI:10.3390/mca21020020
3. [3] R.Dybowski,V.Gant,ClinicalApplicationsofArtificialNeuralNetworks,CambridgeUniversityPress,2007.
4. [4] V. Arbabi, B. Pouran, G. Campoli, H. Weinans, A.A. Zadpoor, Determination of the mechanical and physical properties of cartilage by coupling poroelastic-based finite element models of indentation with artificial neural networks, J Biomech. 49, (2016), ,631–637.
5. [5] J.Dolz,L.Massoptier,M.Vermandel,SegmentationalgorithmsofsubcorticalbrainstructuresonMRIforradiotherapyandradiosurgery:asurvey, IRBM, 36(4), (2015), 200–212.
6. [6] O.Er,N.Yumusak,F.Temurtas,Chestdiseasediagnosisusingartificialneuralnetworks,ExpertSystemswithApplications,37,(2010),7648–7655.
7. [7] O. Er, F. Temurtas, A.C. Tanrikulu, Tuberculosis Disease Diagnosis Using Artificial Neural Networks, J Med Syst, 34, (2010), 299–302, DOI: 10.1007/s10916-008-9241-x
8. [8] J.M.Haglin,G.Jimenez,A.E.M.Eltorai,Artificialneuralnetworksinmedicine,HealthandTechnology,9,(2019),1–6,DOI:10.1007/s12553-018- 0244-4

9. [9] M. Kirisci, H. Yilmaz, M.U. Saka, An ANFIS perspective for the diagnosis of type II diabetes, Annals of Fuzzy Mathematics and Informatics, 17, (2019), 101–113, DOI: 10.30948/afmi.2019.17.2.101
10. [10] M.Kirisci,Comparisonofartificialneuralnetworkandlogisticregressionmodelforfactorsaffectingbirthweight,SNAppl.Sci.,1,(2019),378,DOI: 10.1007/s42452-019-0391-x
11. [11] M. Kirisci, H. Yilmaz, Artificial Neural Networks for Analysis of Factors Affecting Birth Weight, American Journal of Information Science and Computer Engineering, 5, (2019), 17–24.
12. [12] J.Kojuri,R.Boostani,P.Dehghani,F.Nowroozipour,N.Saki,Predictionofacutemyocardialinfarctionwithartificialneuralnetworksinpatientswith nondiagnostic electrocardiogram, Journal of Cardiovascular Disease Research, 6(2), (2015), 51–59.
13. [13] C.C.Lin,Y.K.Ou,S.H.Chen,Y.C.Liu,J.Lin,Comparisonofartificialneuralnetworkandlogisticregressionmodelsforpredictingmortalityinelderly patients with hip fracture, Injury. 41(9), (2010), 869–873.
14. [14] I.M.Nasser,S.S.Abu-Naser,PredictingTumorCategoryUsingArtificialNeuralNetworks,InternationalJournalofAcademicHealthandMedical Research, 3, (2019), 1–7.
15. [15] T.Nowikiewicz,P.Wnuk,B.Małkowski,A.Kurylcio,J.Kowalewski,W.Zegarski,Applicationofartificialneuralnetworksforpredictingpresenceof non-sentinel lymph node metastases in breast cancer patients with positive sentinel lymph node biopsies, Archives of Medical Science, 13, (2017), 1399-–1407.
16. [16] M.Shioji,T.Yamamoto,T.Ibata,T.Tsuda,K.Adachi,N.Yoshimura,Artificialneuralnetworkstopredictfuturebonemineraldensityandboneloss rate in Japanese postmenopausal women, BMC Research Notes, 10, (2017), 590.
17. [17] H. Yu, D.C. Samuels, Y.Y. Zhao, Y. Guo, Architectures and accuracy of artificial neural network for disease classification from omics data, BMC Genomics, 20, (2019), 167, DOI:10.1186/s12864-019-5546-z
18. [18] W.S.McCulloch,W.Pitts,Alogicalcalculusoftheideasimmanentinnervousactivity,BulletinofMath.Biophysics,5,(1943),115–133.
19. [19] J.j.Hopfield,Neuralnetworksandphysicalsystemswithemergentcollectivecomputationalabilities,Proc.Nat.Acad.Sci.,79,(1982),2554–2558.
20. [20] T.Kohonen,Self-organisedformationoftopologicallycorrectfeaturemaps,BiologicalCybernetics,43,(1982),59–69.