Cell Phenotype Classification Based on Joint of Texture Information and Multilayer Feature Extraction in DenseNet

Table 5

Comparison results of state-of-the-art methods on the 2D-HeLa dataset in terms of accuracy, precision, and F1 score (%).

Approach	Type of features	Accuracy (%)	Precision (%)	F1 score (%)
SAHLBP (BoW (VQ) + SPM + SVM) [17]	Handcrafted	84.49 ± 2.2	—	83.75
SIFT + SAHLBP (BoW (VQ) + SPM + SVM) [17]	Handcrafted	86.20 ± 2.5	—	85.35
MLBP8,1 + MLBP16,2 [18]	Handcrafted	84.33	—	—
MT-ULTP8,1 + MT-ULTP16,2 [18]	Handcrafted	86.77	—	—
Haralick-based SVM [21]	Handcrafted	84.10	—	—
Random subspace of LMC [32]	Handcrafted	90.24	—	—
AdaBoost ERC [32]	Handcrafted	91.53 ± 0.02	—	—
SIFT (BoW (LLC) + SPM + softmax) [33]	Handcrafted	89.37 ± 1.5	—	—
LTP16,2 [34]	Handcrafted	87.00	—	—
LBP-rotation invariant uniformLBP16,2 [34]	Handcrafted	82.70	—	—
Orthogonal locality preserving projection (OLPP) [34]	Handcrafted	89.30	—	—
Neighborhood preserving embedding (NPE) [34]	Handcrafted	93.20	—	—
Discriminative LBP [34]	Handcrafted	84.5	—	—
LBP-rotation invariant [34]	Handcrafted	75.01	—	—
Completed LBP [34]	Handcrafted	88.8	—	—
Random subspace ensemble of Levenberg–Marquardt neural network (RSNN) [16]	Classic neural nets	85.00	—	—
VGG-16 [1]	Deep	72.10 ± 3.98	—	—
VGG-16 + transfer learning [1]	Deep	87.07 ± 2.86	—	—
Inspection-v3 [1]	Deep	83.18 ± 2.88	—	—
Ensemble inception-v3 and Resnet152 and inception Resnet-v2 [20]	Deep	93.51 ± 2.29	—	—
Ensemble inception-v3 and Resnet152 [20]	Deep	92.56 ± 1.92	—	—
CapsNet [21]	Deep	93.08	—	—
Inspection-v3 + transfer learning [22]	Deep	90.72 ± 1.85	—	—
Inspection ResNet-v2 [22]	Deep	92.00 ± 1.97	—	—
DenseNet121 (baseline)	Deep	91.87 ± 2.05	91.75	91.04
DenseNet169 (baseline)	Deep	92.63 ± 1.63	92.37	92.03
DenseNet201 (baseline)	Deep	91.72 ± 1.48	91.78	91.59
DenseNet264 (baseline)	Deep	90.98 ± 1.96	90.36	90.68
ILQP + ML-DenseNet (proposed approach)	Handcrafted + deep	93.36 ± 1.85	93.39	93.13