Table 9.

Ablation study of the proposed components in three pre-training settings: image-only, text-only, and image-text.

Settings	Methods	Reporting: Spanish					Reporting: English					Diagnosis	Prognosis
		BLEU-2	BLEU-3	BLEU-4	ROUGE-L	CIDEr	BLEU-2	BLEU-3	BLEU-4	ROUGE-L	CIDEr	Image-Text	Image-Text
Full	Med-MLLM	47.6	42.0	38.1	55.4	73.4	32.4	28.7	24.9	46.7	53.0	78.2	92.0
(a)	w/o image-only (PCL)	45.8	39.9	34.7	54.3	70.9	31.2	27.5	23.0	45.3	51.8	75.7	90.2
(b)	w/o image-only	39.5	35.4	32.2	48.7	57.8	28.1	24.0	19.3	42.1	45.9	69.4	82.3
(c)	w/o text-only (MLM)	45.7	39.6	34.2	54.7	71.1	31.3	27.2	21.8	44.9	50.7	73.3	86.7
(d)	w/o text-only (SR)	44.2	38.5	33.6	54.1	70.3	30.2	26.7	21.5	44.0	49.4	76.3	89.1
(e)	w/o text-only (FIA)	46.0	40.1	36.5	54.2	70.6	31.8	27.5	23.2	45.5	51.6	75.4	88.3
(f)	w/o image-text	42.4	38.3	34.2	53.7	66.5	29.8	25.6	20.7	44.5	48.9	74.1	85.4

We perform the analysis on COVID-19 reporting, diagnosis, and prognosis. All values are reported in percentage (%).