Active learning-based multistage sequential decision-making model with application on common bile duct stone...

Area of Science:

• Decision Sciences
• Medical Informatics
• Machine Learning

Background:

• Multistage sequential decision-making is crucial in healthcare for diagnosis and treatment.
• Doctors face challenges in cost-effectively acquiring patient information for optimal decisions.
• Existing methods may not fully capture the sequential and active nature of clinical information gathering.

Purpose of the Study:

• To develop an active learning-based method for modeling sequential medical decision-making.
• To optimize the cost-effective collection of necessary information in clinical settings.
• To evaluate the proposed model's effectiveness in real-world and simulated scenarios.

Main Methods:

• An active learning framework was designed to guide information acquisition sequentially.
• The model was developed to mimic a doctor's decision-making process in selecting tests or data.
• A two-stage version of the model was specifically investigated for enhanced performance.

Main Results:

• The proposed active learning method demonstrated effectiveness in sequential decision-making tasks.
• The two-stage model showed particular promise in simulation studies.
• Validation on a pediatric common bile duct stone evaluation case study confirmed the model's practical applicability.

Conclusions:

• The developed active learning approach provides a robust framework for sequential medical decision support.
• Optimizing information collection through active learning can lead to more cost-effective and accurate medical decisions.
• The model shows potential for improving diagnostic processes in various pediatric and adult healthcare applications.