Assessing the Accuracy of ChatGPT-4 in Interpreting Arterial Blood Gas Results

Model Training:

The collected data will be used to train the artificial intelligence model. Utilizing the deep learning infrastructure provided by ChatGPT Plus, our model will be optimized to produce highly accurate results in interpreting blood gases.

During the training process, our model will be taught to interpret various blood gas samples, including assessments of acidosis-alkalosis, hypoxemia-hyperoxemia, hemoglobin, electrolytes, and metabolic parameters.

Model Testing and Validation:

The trained model will be tested on previously unseen test datasets to evaluate its performance. This step is crucial for understanding how the model will perform in real-world scenarios.

The accuracy of the model will be assessed by comparing its interpretations with feedback provided by expert anesthesiologists.

Furthermore, to comprehensively evaluate ChatGPT's effectiveness in this domain, the daily arterial blood gas results obtained in the intensive care unit will be submitted to ChatGPT for interpretation. The same questions will be posed, and the responses will be evaluated by an anesthesiology and reanimation specialist. These questions will be asked to the model:

Assessment of Acidosis and Alkalosis, Evaluation of Hypoxemia and Hyperoxemia, Evaluation of Hemoglobin Parameters, Assessment of Electrolytes, Evaluation of Metabolic Parameters (Glucose, Lactate, Bilirubin)

Inclusion criteria

• Patients aged 18 and above will be included in the study.
• The study will evaluate arterial blood gas results.

Exclusion criteria

• Venous blood gas results,
• Blood gas results with calibration errors,
• Blood gas results with incomplete data