Artificial intelligence applications in allergic rhinitis diagnosis: Focus on ensemble learning.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Medicine
• Computational Biology

Background:

• Allergic rhinitis (AR) diagnosis relies on symptoms and lab tests, often limited in outpatient settings.
• Challenges in AR diagnosis include physician expertise variations and examination method limitations.
• Nasal provocation tests and smear examinations are not routinely performed in standard clinical practice.

Purpose of the Study:

• To develop an intelligent diagnosis and detection method for allergic rhinitis (AR) using ensemble learning.
• To leverage artificial intelligence for improved accuracy in AR diagnosis.
• To address diagnostic variability through advanced computational approaches.

Main Methods:

• Collected clinical data for AR and seven other diseases with similar symptoms.
• Employed ensemble learning algorithms, developing an adaptive random forest-out-of-bag-easy ensemble (ARF-OOBEE) classifier.
• Compared ARF-OOBEE and GC Forest against five common machine learning algorithms using metrics like G-mean and AUC.

Main Results:

• Ensemble classification algorithms, ARF-OOBEE and GC Forest, demonstrated superior performance over other models.
• Nearly 2% improvement in G-mean and AUC parameters was observed with ensemble classifiers.
• Ensemble models showed excellent capability in handling large-scale and unbalanced datasets.

Conclusions:

• The ARF-OOBEE ensemble learning model offers strong generalization and comprehensive classification abilities.
• This AI-driven approach is suitable for effective auxiliary diagnosis of allergic rhinitis.
• Ensemble learning presents a promising avenue for enhancing diagnostic accuracy in AR.