Journal of Modern Medical

 Introduction
Thyroid diseases are common disorders worldwide. Various studies have shown their high prevalence in different countries. The timely diagnosis of these diseases and their control can prevent their progress and reduce their consequences. The use of appropriate methods based on artificial intelligence and learning machine algorithms can be useful to make a correct diagnosis in medical cases. The use of smart techniques can provide results that show the factors of various diseases, and based on these results, physicians can act to prevent the disease. Extracting knowledge from massive amounts of data using learning models can lead to the identification of laws governing diseases and provide valuable information to health professionals to identify the causes of diseases, diagnose, predict, and treat patients according to environmental factors. In this research, we aim to use learning models based on artificial neural network (ANN) to predict thyroid diseases.

Methods
 This is an applied survey study that was conducted in 2022. The data of more than 400 patients referred to Imam Reza Hospital in Lar County, Iran from 2021 to 2022 were examined. The proposed method is a hybrid algorithm resulting from the combination of particle swarm optimization (PSO) algorithm and ANN. The process includes three main phases. First, the dataset of thyroid patients is used for pre-processing. In this phase, the input dataset are normalized, and incomplete records are removed from the data set. Then, the normalized dataset enters the second phase for clustering and data labeling. In this phase, the K-means algorithm is used to create clusters in the PSO algorithm. Then, the clusters enter the PSO algorithm to find the optimal clusters. After performing the optimization operation by the PSO algorithm, the optimal clusters containing two clusters of healthy people and patients (labeled datasets) are entered into the next phase as output. In the third phase, the labeled input dataset enters the neural network model designed for the learning process of the intelligent model. The neural network used in this phase was the feed-forward neural network with one input layer, 20 hidden layers, and one output layer. After the training of the intelligent model, this model was used to distinguish people with thyroid diseases from healthy people.

Results
The proposed model was trained on a dataset with 400 samples, including 300 people with thyroid diseases and 100 healthy people. The regression coefficient (R) value of the proposed model in three modes of training, validation, and test was 0.98, 0.97 and 0.95, respectively. The area under the ROC curve was 0.98, and the error rate was 0.004. The results of the proposed method were compared with previous methods based on accuracy. According to Table 1, the proposed model was able to predict thyroid disease with 96% accuracy, compared to previous methods.



Conclusion
In this study, an intelligent method based on PSO algorithm and ANN was presented for distinguishing patients with thyroid diseases from healthy people. The proposed model’s goal was to improve the process of thyroid disease detection. The main part this model was ANN which tried to find the hidden patterns of patients compared to healthy people by using the labeled datasets. Since the way the ANN works depends on how its structure is defined, we used the PSO algorithm is proposed.

 Ethical Considerations
Compliance with ethical guidelines
This study has ethical approval from the Islamic Azad University of Estahban Branch (Code: IR.IAUESTAHBAN.REC.1401.023).

Funding
This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Conflicts of interest
The authors declared no conflict of interest. 

1. Rajam K, Jemina Priyadarsini R. A survey on diagnosis of thyroid disease using data mining techniques. Int J Comput Sci Mob Comput. 2016; 51(5):354-8. [Link] 
2. Hosseinpoor M, Parvin H, Nejatian S, Rezaee V, Bagherifard K. [Presenting a meta-heuristic algorithm to detect regulatory elements in the genome of breast cancer patients (Persian)]. J Adv Biomed Sci. 2020; 10(1):2070-80. [Link]
3. Dabiri MR, Farshidfar A, Najaflo Sh, Saadatdoost R. [Diagnosing thyroid disease using meta-heuristic hybrid algorithms (Persian)]. Paper presented at: Third International Conference on Electrical, Computer and Mechanical Engineering. 20 October 2020; Tehran, Iran. [Link]
4. Hosseinpoor M, Parvin H, Nejatian S, Rezaei V. [Detection and extraction of potential promoter/enhancer interactions in genome of cancer patients using an evolutionary multi-objective algorithm (Persian)]. J Health Biomed Inform. 2018; 5(2):304-13. [Link]
5. Jalili S, Faraji Daneshgar F. [Diagnosis of thyroid diseases using RFC fuzzy classification method (Persian)]. Paper presented at: The 14^th Iran Medical Engineering Conference. 13 February 2008; Tehran, Iran. [Link]
6. Hosseinpoor M. [Predicting gestational diabetes using an intelligent algorithm based on artificial neural network (Persian)]. J Mod Med Inf Sci. 2022; 8(2):126-39. [Link]
7. Zabbah I, Yasrebi Naeini SE, Ramazanpoor Z, Sahragard K. [The diagnosis of thyroid diseases using combinati on of neural networks through hierarchical method (Persian)]. J Health Biomed Inform. 2017; 4(1):21-31. [Link]
8. Aversano L, Bernardi ML, Cimitile M, Iammarino M, Macchia PE, Nettore IC, et al. Thyroid disease treatment prediction with machine learning approaches. Procedia Comput Sci. 2021; 192:1031-40. [DOI:10.1016/j.procs.2021.08.106]
9. Upadhayay A, Shukla S, Kumar S. Empirical comparison by data mining Classification algorithms (C 4.5 & C 5.0) for thyroid cancer data set. Int J Comput Sci Commun Netw. 2013;3(1):64. [Link]
10. Gharehchopogh FS, Molany M, Mokri FD. Using artificial neural network in diagnosis of thyroid diasise: A case study. International Journal on Computational Sciences & Applications. 2013; 3(4):49-61. [Link]  
11. Dhaygude P, Handore SM. A review of thyroid disorder detection using medical images. Int J Recent Innov Trends Comput Commun. 2014; 2(12):4194-7. [Link]
12. Dogantekin E, Dogantekin A,  Avci D. An expert system based on Generalized Discriminant Analysis and Wavelet Support Vector Machine for diagnosis of thyroid diseases. Expert Syst Appl. 2011; 38(1):146-50. [DOI:10.1016/j.eswa.2010.06.029]
13. Zhang X, Lee VCS, Rong J, Liu F, Kong H. Multi-channel convolutional neural network architectures for thyroid cancer detection. PLoS One. 2022; 17(1):e0262128. [DOI:10.1371/journal.pone.0262128] [PMID]
14. Abdolali F, Kapur J, Jaremko JL, Noga M, Hareendranathan AR, Punithakumar K. Automated thyroid nodule detection from ultrasound imaging using deep convolutional neural networks. Comput Biol Med. 2020; 122:103871.  [DOI:10.1016/j.compbiomed.2020.103871] [PMID]
15. Asgari N, Khosravi Danesh F. [Diagnosis of thyroid disease using neural fuzzy classification algorithm (Persian)]. Paper presented at: Second National Conference of Computer Engineering and Information Technology Management. 14 Feruary, 2014; Tehran, Bo Ali Research Group, Iran. [Link]
16. Yazdan SA, Ahmad R, Iqbal N, Rizwan A, Khan AN, Kim DH. An efficient multi-scale convolutional neural network based multi-class brain MRI classification for SaMD. Tomography. 2022; 8(4):1905-27. [DOI:10.3390/tomography8040161] [PMID]
17. Moazezi Z, Hedayati M, shirkhani Z, Azizi F. [Glucose intolerance in subclinical hyperthyroid patients (Persian)]. Iran J Endocrinol Metab. 2012; 14(2):127-34. [Link]
18. Saraswathi V, Santhakumaran A. Towards artificial neural network model to diagnose thyroid problems. Glob J Comput Sci Technol. 2011; 11(5):53-5. [Link]
19. Prerana PS, Taneja K. Predictive data mining for diagnosis of thyroid disease using neural network. Int J Res Manag Sci Technol. 2015; 3(2):75-80. [Link]
20. Zhang X, Lee VC, Rong J, Lee JC, Liu F. Deep convolutional neural networks in thyroid disease detection: A multi-classification comparison by ultrasonography and computed tomography. Comput Methods Programs Biomed. 2022; 220:106823. [PMID]