Table 7.

Results for lung adenocarcinoma baseline and KD approaches as percentages with corresponding 95% CIs. Baseline models were trained from scratch until convergence on the corresponding magnification. Boldface text indicates the best-performing model for each magnification and metric

	Lung Adenocarcinoma
	Baseline	KD
mag = 0.625x (16 μm/pixel)
Accuracy	88.00 (86.07–89.95)	89.32 (87.37–91.26)
F1-Score	54.38 (46.12–64.67)	57.75 (49.74–68.32)
Precision	57.75 (45.53–74.19)	60.95 (48.76–77.30)
Recall	55.98 (48.72–64.62)	58.29 (51.06–67.19)
mag = 1.25x (8 μm/pixel)
Accuracy	90.45 (88.52–92.40)	93.29 (91.44–95.09)
F1-Score	67.57 (56.49–77.07)	73.17 (63.03–82.70)
Precision	69.85 (55.79–80.25)	76.28 (62.54–87.58)
Recall	68.32 (58.07–78.98)	73.02 (63.99–83.26)
mag = 2.5x (4 μm/pixel)
Accuracy	93.14 (91.25–94.94)	93.74 (91.94–95.49)
F1-Score	72.84 (64.11–81.28)	71.88 (64.13–81.07)
Precision	72.03 (63.53–81.02)	73.56 (64.22–87.48)
Recall	75.51 (65.39–86.16)	72.69 (65.27–82.38)
mag = 5x (2 μm/pixel)
Accuracy	94.18 (92.40–95.85)	94.18 (92.45–95.85)
F1-Score	75.33 (66.30–84.23)	79.63 (69.80–87.41)
Precision	76.85 (66.27–88.75)	79.75 (69.62–88.88)
Recall	75.45 (66.74–85.65)	82.00 (71.36–90.62)