Fig. 2: Performance of FABP4, GPT-4 and integrated LRTI diagnostic classifiers in the derivation and validation cohorts. | Nature Communications

Fig. 2: Performance of FABP4, GPT-4 and integrated LRTI diagnostic classifiers in the derivation and validation cohorts.

From: Integrating a host biomarker with a large language model for diagnosis of lower respiratory tract infection

Fig. 2

A Confusion matrices for initial ICU diagnosis and the integrated FABP4/GPT-4 classifier in the derivation cohort. B Receiver operating characteristic curves from GPT-4 classifier, FABP4 classifier, and integrated FABP4/GPT-4 classifier in the derivation cohort. C Confusion matrices for initial ICU diagnosis and the integrated FABP4/GPT-4 classifier in the validation cohort. D Receiver operating characteristic curves from GPT-4 classifier, FABP4 classifier, and integrated FABP4/GPT-4 classifier in the validation cohort. In (A and C), the classifiers output an LRTI diagnosis if the patients had a predicted out-of-fold LRTI probability of 50% or higher, Intensity of color in confusion matrices reflects percentage of patients in each quadrant; red indicates the initial ICU diagnosis and green is the integrated FABP4/GPT-4 classifier. In (B and D), the area under the curves (AUCs) are presented as mean ± standard deviation.

Back to article page