Table 2.

Overview of recent deep-learning development for prediction.

Study	Data Set	Method/Model	Proposed Solution	Languages/Libraries/ Frameworks/ Tools/Software’s Used for Implementation and Simulation	Evaluation
Xu Chen et al. [81]	Custom Developed at Shengjing Hospital of China Medical University (Samples)	Convolutional Neural Network (CNN)	Depth Neural Network, Local Binary Patterns (LBP) features and Glutamate Cysteine Ligase Modifier subunit (GCLM)	Tensor flow	Average Absolute Error = 0.455
Chao Tong et al. [80]	Public Database Digital Hand-Atlas (Samples)	Convolutional Neural Networks (CNNs)	Convolutional Neural Networks (CNNs) and Support Vector Regression (SVR) based Model	Matlab, Keras framework with Tensor Flow	Mean Absolute Error (MAE) = 0.547
Jang Hyung Lee et al. [84]	Radiological Society of North America (RSNA) challenge dataset	CNN	CNN and CaffeNet based Model	Caffe, Tensorflow, Keras, Theano and Torch, Linux Ubuntu OS, NVIDIA GTX 1060 GPU, CUDA library, and CUDNN library.	Concordance Correlation Coefficient = 0.78
Tom Van Steenkiste et al. [83]	Radiological Society of North America challenge dataset (Samples)	Visual Geometry Group (VGG16)echanical Competence and Bone Quality Deve	Visual Geometry Group (VGG16) and Gaussian Process Regression (GPR) based Model	Not Mention	Mean Absolute Difference = 6.80 (−0.94)
Hyunkwang Lee et al. [79]	Custom-developed using open-source software OsiriX and DICOM images (Samples)	CNN	ImageNet pre-trained, fine-tuned convolutional neural network (CNN)	GoogLeNet and Caffe Zoo	Accuracy = 98.56%
Jeong Rye Kim et al. [82]	Custom Developed (Samples)	Deep Neural Network	Greulich-Pyle and Deep Neural Network-Based Model	Not Mention	Root Mean Square Error = 0.42