Construction of a Diagnostic Algorithm for Diagnosis of Adult Asthma Using Machine Learning with Random Forest and XGBoost
An evidence-based diagnostic algorithm for adult asthma is necessary for effective treatment and management. We present a diagnostic algorithm that utilizes a random forest (RF) and an optimized eXtreme Gradient Boosting (XGBoost) classifier to diagnose adult asthma as an auxiliary tool. Data were g...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2023-09-01
|
| Series: | Diagnostics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2075-4418/13/19/3069 |
| _version_ | 1797576061195649024 |
|---|---|
| author | Katsuyuki Tomita Akira Yamasaki Ryohei Katou Tomoyuki Ikeuchi Hirokazu Touge Hiroyuki Sano Yuji Tohda |
| author_facet | Katsuyuki Tomita Akira Yamasaki Ryohei Katou Tomoyuki Ikeuchi Hirokazu Touge Hiroyuki Sano Yuji Tohda |
| author_sort | Katsuyuki Tomita |
| collection | DOAJ |
| description | An evidence-based diagnostic algorithm for adult asthma is necessary for effective treatment and management. We present a diagnostic algorithm that utilizes a random forest (RF) and an optimized eXtreme Gradient Boosting (XGBoost) classifier to diagnose adult asthma as an auxiliary tool. Data were gathered from the medical records of 566 adult outpatients who visited Kindai University Hospital with complaints of nonspecific respiratory symptoms. Specialists made a thorough diagnosis of asthma based on symptoms, physical indicators, and objective testing, including airway hyperresponsiveness. We used two decision-tree classifiers to identify the diagnostic algorithms: RF and XGBoost. Bayesian optimization was used to optimize the hyperparameters of RF and XGBoost. Accuracy and area under the curve (AUC) were used as evaluation metrics. The XGBoost classifier outperformed the RF classifier with an accuracy of 81% and an AUC of 85%. A combination of symptom–physical signs and lung function tests was successfully used to construct a diagnostic algorithm on importance features for diagnosing adult asthma. These results indicate that the proposed model can be reliably used to construct diagnostic algorithms with selected features from objective tests in different settings. |
| first_indexed | 2024-03-10T21:47:09Z |
| format | Article |
| id | doaj.art-e87267e079164d45814dada0b4dab90c |
| institution | Directory Open Access Journal |
| issn | 2075-4418 |
| language | English |
| last_indexed | 2024-03-10T21:47:09Z |
| publishDate | 2023-09-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Diagnostics |
| spelling | doaj.art-e87267e079164d45814dada0b4dab90c2023-11-19T14:14:18ZengMDPI AGDiagnostics2075-44182023-09-011319306910.3390/diagnostics13193069Construction of a Diagnostic Algorithm for Diagnosis of Adult Asthma Using Machine Learning with Random Forest and XGBoostKatsuyuki Tomita0Akira Yamasaki1Ryohei Katou2Tomoyuki Ikeuchi3Hirokazu Touge4Hiroyuki Sano5Yuji Tohda6Department of Respiratory Medicine, Yonago Medical Center, National Hospital Organization, Yonago 683-0006, JapanDivision of Respiratory Medicine and Rheumatology, Department of Multidisciplinary Internal Medicine, School of Medicine, Tottori University, Yonago 683-8503, JapanDepartment of Respiratory Medicine, Yonago Medical Center, National Hospital Organization, Yonago 683-0006, JapanDepartment of Respiratory Medicine, Yonago Medical Center, National Hospital Organization, Yonago 683-0006, JapanDepartment of Respiratory Medicine, Yonago Medical Center, National Hospital Organization, Yonago 683-0006, JapanAllergy Center, Kindai University Hospital, Osakasayama 589-8511, JapanDepartment of Respiratory and Allergorogy, Kindai University, Osakasayama 589-8511, JapanAn evidence-based diagnostic algorithm for adult asthma is necessary for effective treatment and management. We present a diagnostic algorithm that utilizes a random forest (RF) and an optimized eXtreme Gradient Boosting (XGBoost) classifier to diagnose adult asthma as an auxiliary tool. Data were gathered from the medical records of 566 adult outpatients who visited Kindai University Hospital with complaints of nonspecific respiratory symptoms. Specialists made a thorough diagnosis of asthma based on symptoms, physical indicators, and objective testing, including airway hyperresponsiveness. We used two decision-tree classifiers to identify the diagnostic algorithms: RF and XGBoost. Bayesian optimization was used to optimize the hyperparameters of RF and XGBoost. Accuracy and area under the curve (AUC) were used as evaluation metrics. The XGBoost classifier outperformed the RF classifier with an accuracy of 81% and an AUC of 85%. A combination of symptom–physical signs and lung function tests was successfully used to construct a diagnostic algorithm on importance features for diagnosing adult asthma. These results indicate that the proposed model can be reliably used to construct diagnostic algorithms with selected features from objective tests in different settings.https://www.mdpi.com/2075-4418/13/19/3069adult asthmaartificial intelligencediagnostic assistantmachine learningrandom forestXGBoost |
| spellingShingle | Katsuyuki Tomita Akira Yamasaki Ryohei Katou Tomoyuki Ikeuchi Hirokazu Touge Hiroyuki Sano Yuji Tohda Construction of a Diagnostic Algorithm for Diagnosis of Adult Asthma Using Machine Learning with Random Forest and XGBoost Diagnostics adult asthma artificial intelligence diagnostic assistant machine learning random forest XGBoost |
| title | Construction of a Diagnostic Algorithm for Diagnosis of Adult Asthma Using Machine Learning with Random Forest and XGBoost |
| title_full | Construction of a Diagnostic Algorithm for Diagnosis of Adult Asthma Using Machine Learning with Random Forest and XGBoost |
| title_fullStr | Construction of a Diagnostic Algorithm for Diagnosis of Adult Asthma Using Machine Learning with Random Forest and XGBoost |
| title_full_unstemmed | Construction of a Diagnostic Algorithm for Diagnosis of Adult Asthma Using Machine Learning with Random Forest and XGBoost |
| title_short | Construction of a Diagnostic Algorithm for Diagnosis of Adult Asthma Using Machine Learning with Random Forest and XGBoost |
| title_sort | construction of a diagnostic algorithm for diagnosis of adult asthma using machine learning with random forest and xgboost |
| topic | adult asthma artificial intelligence diagnostic assistant machine learning random forest XGBoost |
| url | https://www.mdpi.com/2075-4418/13/19/3069 |
| work_keys_str_mv | AT katsuyukitomita constructionofadiagnosticalgorithmfordiagnosisofadultasthmausingmachinelearningwithrandomforestandxgboost AT akirayamasaki constructionofadiagnosticalgorithmfordiagnosisofadultasthmausingmachinelearningwithrandomforestandxgboost AT ryoheikatou constructionofadiagnosticalgorithmfordiagnosisofadultasthmausingmachinelearningwithrandomforestandxgboost AT tomoyukiikeuchi constructionofadiagnosticalgorithmfordiagnosisofadultasthmausingmachinelearningwithrandomforestandxgboost AT hirokazutouge constructionofadiagnosticalgorithmfordiagnosisofadultasthmausingmachinelearningwithrandomforestandxgboost AT hiroyukisano constructionofadiagnosticalgorithmfordiagnosisofadultasthmausingmachinelearningwithrandomforestandxgboost AT yujitohda constructionofadiagnosticalgorithmfordiagnosisofadultasthmausingmachinelearningwithrandomforestandxgboost |