Date of Award
12-2021
Degree Type
Thesis
Degree Name
MS in Epidemiology & Biostatistics
First Advisor
DR. MASOOD KADIR
Second Advisor
DR. JUNAID RAZZAK
Third Advisor
DR. NADEEMULLAH KHAN
Department
Community Health Sciences
Abstract
Background: Pneumothorax can be fatal and needs early diagnosis and prompt management on arrival to Emergency room. The purpose of this study is to validate a Pneumothorax Machine learning model designed on an online and in-hospital dataset and validate the accuracy of this ML tool in comparison to radiologist and emergency physician.
Method: A cross sectional study modified on an online available open access tool was used. Hospital dataset from 1st Jan 2010 to 31st Dec 2020 was obtained. 4788 DICOM X-ray images were extracted and those from hospital record was manually labelled by a ML team. The internal validation was evaluated by a supervised learning Machine model with CNN was utilized on Python and Medcalc; kappa statistics were assessed on STATA v 14.2. The model was trained, an AUC and ROC were designed with sensitivity, specificity, Positive & negative predictive values and accuracy.
Results: Initial training of the model showed a validation accuracy of 96.4%, followed by pre[1]trained model with 98% accuracy & a fine tuned model having 97.9% accuracy. The sensitivity was found to be 93.99%, specificity 91.34, PPV 92.88, NPV 92.67 with 92.79% accuracy. The ML Tool was highly accurate while there was a moderate agreement between the Radiologist and Emergency Physician in presence of Pneumothorax.
Conclusion: The diagnostic investigation discovered that developing neural networks and advanced machine learning models may be used to diagnose pneumothorax using machine learning models. Integrating such AI systems into physician workflows for preliminary interpretations has the potential to provide physicians with early diagnostics and profound alerts that can help to better diagnose occult pneumothorax and reduce human errors, particularly in resource-constrained settings, while also improving overall accuracy and lowering healthcare costs.
First Page
1
Last Page
114
Recommended Citation
Abbasi, A.
(2021). Radiographic detection of pneumothorax in trauma victims by machine learning tool and validation of machine learning tool with radiologist & emergency physician, A cross sectional study.. , 1-114.
Available at:
https://ecommons.aku.edu/etd_pk_mc_mseb/15