Table 2.

The classification accuracy results (%) using the DL approach on the BreaKHis dataset (Bi means the fine-tuned blocks).

Classification Scheme	Magnification Factor	Accuracy Based on Block-Wise Fine-Tuning Strategy
		B6	B6-B5	B6-B4	B6-B3	B6-B2	B6-B1
Binary	40×	53.23	88.27	94.56	98.13	97.96	97.45
	100×	74.27	88.76	94.79	94.95	97.39	96.09
	200×	73.74	85.52	93.10	94.95	96.63	94.95
	400×	68.03	83.27	89.03	94.05	92.38	91.64
Eight-class	40×	31.80	74.83	80.95	88.95	88.27	85.88
	100×	44.79	65.96	79.15	79.80	83.22	83.71
	200×	49.49	60.44	76.94	79.63	80.64	79.12
	400×	47.03	56.69	67.29	76.77	72.49	76.39