Table 2.

Diagnostic evaluation indicators.

Dataset	model	TPa	FNb	FPc	TNd	Sensitivity	Specificity	PPVe	NPVf	PLRg	NLRh	Accuracy
Training set	Gpt-4 Model 1	10	34	5	73	0.2273	0.9359	0.6667	0.6822	3.5455	0.8257	0.6803
Training set	GPT-4 Model 2	20	24	6	72	0.4545	0.9231	0.7692	0.7500	5.9091	0.5909	0.7541
Training set	GPT-4 Model 3	26	18	12	66	0.5909	0.8462	0.6842	0.7857	3.8409	0.4835	0.7541
Training set	GPT-4 Model 4	16	28	13	65	0.3636	0.8333	0.5517	0.6989	2.1818	0.7636	0.6639
Training set	GPT-4 Model 5	38	6	4	74	0.8636	0.9487	0.9048	0.9250	16.8409	0.1437	0.9180
Training set	GPT-3.5 Model 1	13	31	14	64	0.2955	0.8205	0.4815	0.6737	1.6461	0.8587	0.6311
Training set	GPT-3.5 Model 2	22	22	24	54	0.5000	0.6923	0.4783	0.7105	1.6250	0.7222	0.6230
Training set	GPT-3.5 Model 3	26	18	28	50	0.5909	0.6410	0.4815	0.7353	1.6461	0.6382	0.6230
Training set	GPT-3.5 Model 4	22	22	31	47	0.5000	0.6026	0.4151	0.6812	1.2581	0.8298	0.5656
Training set	GPT-3.5 Model 5	25	19	21	57	0.5682	0.7308	0.5435	0.7500	2.1104	0.5909	0.6721
Test set	GPT-4 Model 5	17	5	5	25	0.7727	0.8333	0.7727	0.8333	4.6364	0.2727	0.8077
Test set	GPT-3.5 Model 5	12	10	9	21	0.5455	0.7000	0.5714	0.6774	1.8182	0.6494	0.6346

^aTP: true positive.

^bFN: false negative.

^cFP: false positive.

^dTN: true negative.

^ePPV: positive predictive value.

^fNPV: negative predictive value.

^gPLR: positive likelihood ratio.

^hNLR: negative likelihood ratio.