Evaluation of the models generated from clinical features and deep learning-based segmentations: Can thoracic CT on admission help us to predict hospitalized COVID-19 patients who will require intensive care?

Table 3 Volumetric data, comparison between the groups and results of ROC analysis

Parameter	Group 1	Group 2	p	Cut-off	Sensitivity %	Specificity %	AUC^c
Right lung
Upper Third (%)	5 (16)	11.5 (20)	0.011^a	> 9	59.5	65.8	0.608
Middle Third (%)	14 (17)	21.5 (27)	0.015^a	> 24	48.7	77.8	0.604
Lower Third (%)	12 (18)	18.5 (30)	0.012^a	> 35	26.4	93.2	0.608
Infected Total (%)	12 (12)	21 (21)	0.002^a	> 19	54.1	76.1	0.635
Upper Third (mL)	25 (71)	51 (105)	0.032^a	> 39	60.1	60.1	0.592
Middle Third (mL)	136 (152)	174.5 (220)	0.120^a
Lower Third (mL)	41 (69)	63.5 (94)	0.049^a	> 77	46.0	73.0	0.584
Infected Total (mL)	222 (235)	307.5 (416)	0.010^a	> 354	47.3	76.9	0.611
Whole Right Lung Volume(mL)	1849 ± 574	1628 ± 515	0.008^b	≤ 1516	46.0	69.2	0.592
Left lung
Upper Third (%)	1 (7)	4 (16)	0.020^a	> 7	40.5	77.8	0.597
Middle Third (%)	10 (20)	19 (30)	0.005^a	> 12	66.2	58.1	0.622
Lower Third (%)	11 (22)	21 (31)	0.024^a	> 26	43.2	76.1	0.597
Infected Total (%)	10 (16)	15.5 (19)	0.001^a	> 11	63.5	59.8	0.640
Upper Third (mL)	9 (30)	13.5 (79)	0.012^a	> 71	27.0	92.3	0.575
Middle Third (mL)	73 (136)	125 (174)	0.080^a
Lower Third (mL)	38 (70)	67 (88)	0.060^a
Infected Total (mL)	139 (209)	232 (281)	0.092^a
Whole Left Lung Volume(mL)	1583 ± 525	1419 ± 491	0.033^b	≤ 1819	83.8	31.6	0.581
Both lungs
Total infected lung (%)	12.26 (12.55)	18.60 (19.24)	< 0.001^a	> 16.61	59.5%	70.1%	0.674
Total infected lung (mL)	377 (441)	583.5 (612)	0.001^a	> 435	68.9%	58.1%	0.638
Total Lung Volume	3433 ± 1078	3049 ± 956	0.011^b	≤ 3804	83.8%	33.3%	0.589

Group results were given as Median (IQR) or Mean ± SD according to the parameter’s distribution
Group 1: Patients discharged from inpatient floor; Group 2: Patients transferred to ICU
^aMann–Whitney test result
^bT-test result
^cROC analysis result

ISSN: 1471-2342