Related Experiment Video
Updated: Aug 19, 2025
07:15
Machine Learning Algorithms for Early Detection of Bone Metastases in an Experimental Rat Model
Published on: August 16, 2020
6.9K
A deep belief network-based clinical decision system for patients with osteosarcoma
Wenle Li1,2, Youzheng Dong3, Wencai Liu4
1Department of Orthopaedic Surgery, People's Hospital of Xinjiang Uygur Autonomous Region, Urumqi, Xinjiang, China.
Frontiers in Immunology
|December 5, 2022
Summary
Deep belief networks (DBN) show superior performance in predicting lung metastasis and overall survival for osteosarcoma patients compared to other machine learning algorithms. This AI approach aids in personalized prognosis and clinical decision-making for this bone cancer.
Area of Science:
- Oncology
- Biomedical Informatics
- Machine Learning
Background:
- Osteosarcoma is a primary bone cancer with a high mortality rate, particularly in pediatric and adolescent populations.
- Accurate prognosis is crucial for tailoring treatment strategies and improving patient outcomes.
- Machine learning (ML) models are increasingly utilized for predicting cancer metastasis and survival.
Purpose of the Study:
- To evaluate the efficacy of a deep belief network (DBN) algorithm in predicting lung metastasis in osteosarcoma patients.
- To compare the performance of DBN against six other ML algorithms for metastasis prediction.
- To develop and validate an ML-based model for predicting overall survival in osteosarcoma patients.
Main Methods:
- A deep belief network (DBN) model was developed and compared with Random Forest, XGBoost, Decision Tree, Gradient Boosting Machine, Logistic Regression, and Naive Bayes classifiers.
- The DBN model's performance was assessed using accuracy, precision, recall, and F1 score on training and validation datasets.
- A DBN-based lung metastasis prediction model was integrated into a Cox proportional hazards model to predict patient survival, with performance evaluated using Area Under the Curve (AUC) and calibration curves.
Main Results:
- The DBN algorithm demonstrated superior performance over other ML algorithms, achieving high accuracy (0.917/0.888), precision (0.896/0.643), recall (0.956/0.900), and F1 scores (0.925/0.750) in training/validation sets.
- The DBN survival Cox model showed good discrimination for 1-, 3-, and 5-year survival prediction (AUCs: 0.851, 0.806, 0.793 in training set) with strong calibration.
- The DBN model also exhibited promising results in the validation set, and a nomogram was created for clinical decision support.
Conclusions:
- Deep belief networks offer a highly accurate and robust approach for predicting lung metastasis in osteosarcoma patients.
- The DBN-based survival model provides reliable prognostic information, aiding in personalized treatment planning.
- The developed DBN model and nomogram can serve as valuable tools for clinicians in managing osteosarcoma patients.
