Table 1.

Models and accuracies of digital twins in research.

Year	Applications	Modeling	Database	Accuracy	Reference
2022	Cancer analyses progression and specific organ identification	Natural language processing labeling with CNNa, recurrent neural network models, term frequency-inverse document frequency ensemble model	Computed tomography scans (2009-2021)	In validation population: 93.8% for lung cancer, 92.5% for liver cancer, and 96.1% for adrenal cancer with term frequency-inverse document frequency ensemble model; 96.8% for lung, 99% for liver, and 99.7% for adrenal with augmented CNN.	[39]
2022	Personal health care improvement with emotion recognition	Automatic detection of emotion from EEGb signals with gradient boosting, k-nearest neighbor, and RFc models	EEG images	99.9% with gradient boosting, 98.6% with decision tree classifier, 99.7% with k-nearest neighbor, and 99.6% with RF	[40]
2022	Prostate cancer progression with biochemical recurrence and seminal vesicle	Machine learning methods with support vector machine, RF, NN, recurrent neural network, and long short-term memory models	Clinical data warehouse with patients with cancer	82.7% of accuracy for biochemical recurrence and 83.9% for seminal vesicle	[41]
2021	Abdominal aortic aneurysm severity detection	Inverse analysis with CNN and long short-term memory models	Digital patient’s data set	99.91% for detection and 97.79% for severity	[42]
2021	Prevention of stroke and treatment of poststroke	Support vector machine	EEG data set	76% accuracy and 0.84 for performance	[43]
2019	Fault diagnosis pattern	Deep neural network model with deep-neural network and deep transfer learning (DFDDd)	Digital patient’s data set	98% for accuracy with DFDD and 91.5% with deep-neural network	[44]
2019	Ischemic heart disease detection	Deep neural model	Pulmonary tuberculosis diagnostic electrocardiogram database	85.8% for the implemented model	[45]

^aCNN: convolutional neural network.

^bEEG: Electroencephalography.

^cRF: random forest.

^dDFDD: fault diagnosis method using deep transfer learning.