Table 3.

Deep learning method for predicting diabetic foot

Author	Deep learning method	Purpose	Prediction Result
Amith Khandakar, 2021 [11]	Convolutional Neural Network	Used in the automatic identification and classification of diabetic foot ulcer (DFU) images, it predicts risks by analyzing wound images	Predicting risks through wound image analysis
V Sathya Preiya, (2 023) [36]	Recurrent Neural Network	It can capture long-term dependencies in the data, enabling more accurate prediction of the risk of diabetic foot	By inputting data, it outputs a predicted probability value, where a higher probability indicates a higher risk
Yufan He, (2021) [37]	Autoencoder	Used for feature extraction, aiding in the discovery of potential variables related to risk	By inputting data, a prediction probability is obtained, and the higher the probability, the greater the risk
Manu Goyal, (2020) [38]	Deep Learning Ensemble Methods	Improving overall prediction performance by combining the prediction results of different models	Through processes such as data. preprocessing, model training, and ensemble strategies, a prediction probability value is obtained, and the higher the probability, the greater the risk
Jing Zhao, (2022) [39]	Generative Adversarial Networks, GANs	Used for generating high quality simulated data to make the model’s predictions of diabetic foot more accurate	Through data generation and feature enhancement, a prediction probability value is obtained, which categorizes patients into high-risk and low-risk groups
Maheswari D, (2024) [40]	Attention Mechanisms	It can improve prediction accuracy, enhance model interpretability, and promote the enhancement of diabetic foot prediction model performance	Different risk characteristics are assigned weights, and a risk probability is obtained based on these weights. The higher the probability, the greater the risk