Skip to main content

Table 2 Classification performance on BreakHis dataset

From: Texture features in the Shearlet domain for histopathological image classification

Method Perf RP + Magnitude RP Magnitude
SVM DTB SVM DTB SVM DTB
(1): Baseline performance of existing Shearlet-based methods
Vo et al. ACC% \(\underline{85} .\underline{69} \% \pm \underline{1} .\underline{18} \%\) \(80.65 \% \pm 1.05\%\) \(83.63\% \pm 0.99\%\) \(78.95\% \pm 0.74\%\) \(76.10\% \pm 0.66\%\) \(76.52\% \pm 0.64\%\)
AUC \(\underline{0} .\underline{9121} \pm \underline{0} .\underline{0053}\) \(0.8687 \pm 0.0079\) \(0.8915 \pm 0.0103\) \(0.8443 \pm 0.0183\) \(0.7717 \pm 0.0123\) \(0.7926 \pm 0.0175\)
Sen \(\underline{0} .\underline{7210} \pm \underline{0} .\underline{0276}\) \(0.4855 \pm 0.0346\) \(0.6766 \pm 0.0231\) \(0.4496 \pm 0.0157\) \(0.4698 \pm 0.0213\) \(0.4056 \pm 0.0301\)
Prec \(\underline{0} .\underline{8035} \pm \underline{0} .\underline{0301}\) \(0.8268 \pm 0.0296\) \(0.7732 \pm 0.0201\) \(0.7881 \pm 0.0208\) \(0.6701 \pm 0.0192\) \(0.7247 \pm 0.0159\)
Meshkini and Ghassemian ACC% \(\textit{78.06}\% \pm \textit{0.80}\%\) \(\textit{79.43} \% \pm \textit{1.09}\%\) \(\textit{77.33}\% \pm \textit{1.23}\%\) \(\textit{79.76}\% \pm \textit{0.71}\%\) \(76.44\% \pm 0.96\%\) \(77.46\% \pm 0.85\%\)
AUC \(\textit{0.8339} \pm \textit{0.0030}\) \(\textit{0.8523} \pm \textit{0.0055}\) \(\textit{0.8111} \pm \textit{0.0086}\) \(\textit{0.8579} \pm \textit{0.0078}\) \(0.7991 \pm 0.0139\) \(0.8357 \pm 0.0112\)
Sen \(\textit{0.5177} \pm \textit{0.0194}\) \(\textit{0.4927} \pm \textit{0.0213}\) \(\textit{0.4899} \pm \textit{0.0291}\) \(\textit{0.4980} \pm \textit{0.0237}\) \(0.4722 \pm 0.0232\) \(0.4218 \pm 0.0246\)
Prec \(\textit{0.7049} \pm \textit{0.0223}\) \(\textit{0.7688} \pm \textit{0.0313}\) \(\textit{0.6973} \pm \textit{0.0284}\) \(\textit{0.7779} \pm \textit{0.0300}\) \(0.6788 \pm 0.0215\) \(0.7496 \pm 0.0181\)
Zhou et al. ACC% \(\textit{68.66}\% \pm \textit{1.23}\%\) \(\textit{70.21} \% \pm \textit{0.36}\%\) \(\textit{65.20}\% \pm \textit{1.47}\%\) \(\textit{69.64}\% \pm \textit{0.47}\%\) \(66.47\% \pm 0.88\%\) \(70.58\% \pm 0.49\%\)
AUC \(\textit{0.6426} \pm \textit{0.0245}\) \(\textit{0.6781} \pm \textit{0.0173}\) \(\textit{0.5832} \pm \textit{0.0193}\) \(\textit{0.6154} \pm \textit{0.0116}\) \(0.6022 \pm 0.0198\) \(0.6709 \pm 0.0165\)
Sen \(\textit{0.2875} \pm \textit{0.0310}\) \(\textit{0.0677} \pm \textit{0.0139}\) \(\textit{0.2290} \pm \textit{0.0273}\) \(\textit{0.0456} \pm \textit{0.0096}\) \(0.2415 \pm 0.0326\) \(0.0956 \pm 0.0123\)
Prec \(\textit{0.4996} \pm \textit{0.0357}\) \(\textit{0.7915} \pm \textit{0.0314}\) \(\textit{0.4035} \pm \textit{0.0391}\) \(\textit{0.7728} \pm \textit{0.1106}\) \(0.4361 \pm 0.0291\) \(0.7436 \pm 0.0687\)
Dong et al. ACC% \(\textit{79.49}\% \pm \textit{1.39}\%\) \(\textit{78.40} \% \pm \textit{0.54}\%\) \(\textit{73.54}\% \pm \textit{1.61}\%\) \(\textit{75.58}\% \pm \textit{0.78}\%\) \(78.40\ \pm 1.20\%\) \(77.56\% \pm 0.97\%\)
AUC \(\textit{0.8390} \pm \textit{0.0163}\) \(\textit{0.8359} \pm \textit{0.0110}\) \(\textit{0.7641} \pm \textit{0.0172}\) \(\textit{0.8018} \pm \textit{0.0069}\) \(0.8090 \pm 0.0114\) \(0.8082 \pm 0.0156\)
Sen \(\textit{0.5980} \pm \textit{0.0342}\) \(\textit{0.4360} \pm \textit{0.0139}\) \(\textit{0.4860} \pm \textit{0.0312}\) \(\textit{0.3328} \pm \textit{0.0327}\) \(0.5176 \pm 0.0270\) \(0.4380 \pm 0.0129\)
Prec \(\textit{0.7077} \pm \textit{0.0226}\) \(\textit{0.7855} \pm \textit{0.0169}\) \(\textit{0.6011} \pm \textit{0.0324}\) \(\textit{0.7600} \pm \textit{0.0155}\) \(0.7204 \pm 0.0243\) \(0.7480 \pm 0.0307\)
(2): Proposed Shearlet-based methods for textured bio-medical image classification
CM ACC% \(87.26\% \pm 1.18\%\) \(79.33 \% \pm 1.17\%\) \(86.93\% \pm 0.75\%\) \(79.95\% \pm 0.39\%\) \(78.85\% \pm 1.42\%\) \(75.08\% \pm 0.80\%\)
AUC \(0.9365 \pm 0.0104\) \(0.8679 \pm 0.0134\) \(0.9283 \pm 0.0052\) \(0.8687 \pm 0.0085\) \(0.8350 \pm 0.0143\) \(0.7823 \pm 0.0148\)
Sen \(0.7391 \pm 0.0286\) \(0.4137 \pm 0.0351\) \(0.7520 \pm 0.0242\) \(0.4677 \pm 0.0188\) \(0.5649 \pm 0.0371\) \(0.2911 \pm 0.0208\)
Prec \(0.8358 \pm 0.0230\) \(0.8500 \pm 0.0270\) \(0.8166 \pm 0.0108\) \(0.8140 \pm 0.0164\) \(0.7021 \pm 0.0253\) \(0.7717 \pm 0.0274\)
LBP ACC% \(89.51\% \pm 0.69\%\) \(81.63 \% \pm 0.96\%\) \(87.15\% \pm 0.51\%\) \(79.58\% \pm 1.05\%\) \(86.07\% \pm 1.04\%\) \(80.01\% \pm 1.11\%\)
AUC \(0.9477 \pm 0.0043\) \(0.8905 \pm 0.0101\) \(0.9296 \pm 0.0054\) \(0.8665 \pm 0.0131\) \(0.9153 \pm 0.0080\) \(0.8630 \pm 0.0075\)
Sen \(0.7964 \pm 0.0233\) \(0.5081 \pm 0.0180\) \(0.7508 \pm 0.0173\) \(0.4484 \pm 0.0305\) \(0.7177 \pm 0.0355\) \(0.4641 \pm 0.0296\)
Prec \(0.8590 \pm 0.0149\) \(0.8444 \pm 0.0299\) \(0.8241 \pm 0.0126\) \(0.8182 \pm 0.0212\) \(0.8158 \pm 0.0132\) \(0.8200 \pm 0.0219\)
LOSIB ACC% \(87.81\% \pm 0.63\%\) \(80.63\% \pm 1.03\%\) \(85.78 \% \pm 0.52\%\) \(78.94\% \pm 1.47\%\) \(80.31\% \pm 1.09\%\) \(78.49\% \pm 0.90\%\)
AUC \(0.9295 \pm 0.0056\) \(0.8771 \pm 0.0073\) \(0.9092 \pm 0.0042\) \(0.8516 \pm 0.0174\) \(0.8486 \pm 0.0085\) \(0.8206 \pm 0.0119\)
Sen \(0.7585 \pm 0.0252\) \(0.4972 \pm 0.0225\) \(0.7234 \pm 0.0177\) \(0.4315 \pm 0.0284\) \(0.5665 \pm 0.0274\) \(0.4617 \pm 0.0170\)
Prec \(0.8380 \pm 0.0163\) \(0.8121 \pm 0.0220\) \(0.8036 \pm 0.0116\) \(0.8071 \pm 0.0436\) \(0.7447 \pm 0.0220\) \(0.7580 \pm 0.0247\)
SFTA ACC% \(\underline{89} .\underline{72} \% \pm \underline{0} .\underline{63} \%\) \(81.83\% \pm 0.79\%\) \(87.95\% \pm 0.57\%\) \(79.90\% \pm 1.20\%\) \(83.53\% \pm 1.04\%\) \(80.39\% \pm 1.14\%\)
AUC \(\underline{0} .\underline{9527} \pm \underline{0} .\underline{0055}\) \(0.8880 \pm 0.0071\) \(0.9393 \pm 0.0065\) \(0.8769 \pm 0.0070\) \(0.8943 \pm 0.0090\) \(0.8424 \pm 0.0158\)
Sen \(\underline{0} .\underline{8040} \pm \underline{0} .\underline{0243}\) \(0.4992 \pm 0.0265\) \(0.7701 \pm 0.0239\) \(0.4601 \pm 0.0251\) \(0.6645 \pm 0.0332\) \(0.4665 \pm 0.0220\)
Prec \(\underline{0} .\underline{8593} \pm \underline{0} .\underline{0112}\) \(0.8641 \pm 0.0137\) \(0.8332 \pm 0.0112\) \(0.8197 \pm 0.0324\) \(0.7787 \pm 0.0264\) \(0.8355 \pm 0.0323\)
(3): Integrating Shearlet-based existing techniques with our proposed methods
Fusion #1 ACC% \(\underline{91} .\underline{28} \% \pm \underline{0} .\underline{51} \%\) \(81.78\% \pm 0.73\%\) \(89.58\% \pm 0.83\%\) \(80.98\% \pm 0.71\%\) \(87.29\% \pm 0.37\%\) \(80.30\% \pm 0.62\%\)
AUC \(\underline{0} .\underline{9650} \pm \underline{0} .\underline{0031}\) \(0.8981 \pm 0.0053\) \(0.9515 \pm 0.0031\) \(0.8853 \pm 0.0102\) \(0.9346 \pm 0.0051\) \(0.8719 \pm 0.0084\)
Sen \(\underline{0} .\underline{8391} \pm \underline{0} .\underline{0147}\) \(0.5085 \pm 0.0087\) \(0.8121 \pm 0.0119\) \(0.4855 \pm 0.0252\) \(0.7561 \pm 0.0214\) \(0.4504 \pm 0.0163\)
Prec \(\underline{0} .\underline{8775} \pm \underline{0} .\underline{0120}\) \(0.8508 \pm 0.0259\) \(0.8495 \pm 0.0219\) \(0.8412 \pm 0.0185\) \(0.8244 \pm 0.0094\) \(0.8516 \pm 0.0198\)
Fusion #2 ACC% \(89.33\% \pm 0.61\%\) \(80.33\% \pm 0.89\%\) \(87.82\% \pm 0.36\%\) \(80.24\% \pm 0.84\%\) \(83.23\% \pm 0.93\%\) \(77.80\% \pm 0.51\%\)
AUC \(0.9503 \pm 0.0053\) \(0.8763 \pm 0.0138\) \(0.9354 \pm 0.0037\) \(0.8742 \pm 0.0072\) \(0.8872 \pm 0.0098\) \(0.8446 \pm 0.0071\)
Sen \(0.7964 \pm 0.0245\) \(0.4544 \pm 0.0202\) \(0.7690 \pm 0.0132\) \(0.4726 \pm 0.0171\) \(0.6536 \pm 0.0155\) \(0.3694 \pm 0.0173\)
Prec \(0.8537 \pm 0.0105\) \(0.8471 \pm 0.0213\) \(0.8304 \pm 0.0098\) \(0.8216 \pm 0.0252\) \(0.7770 \pm 0.0256\) \(0.8270 \pm 0.0170\)
Fusion #3 ACC% \(90.10\% \pm 0.87\%\) \(85.08\% \pm 0.80\%\) - - - -
AUC \(0.9560 \pm 0.0053\) \(0.9256 \pm 0.0065\) - - - -
Sen \(0.8028 \pm 0.0113\) \(0.5992 \pm 0.0175\) - - - -
Prec \(0.8723 \pm 0.0281\) \(0.8887 \pm 0.0161\) - - - -
  1. The results shown in italic are of experiments that are not explored in the original research papers
  2. The underlined classification results represent the highest results for each corresponding section