Skip to main content

Table 4 Classification performance on Warwick-QU 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% \(\underline{93} .\underline{35} \% \pm \underline{5} .\underline{23} \%\) \(86.14\% \pm 6.10\%\) \(92.13\% \pm 8.53\%\) \(85.55\% \pm 6.83\%\) \(73.31\% \pm 11.27\%\) \(73.97 \% \pm 12.26\%\)
AUC \(\underline{0} .\underline{9802} \pm \underline{0} .\underline{0351}\) \(0.9362 \pm 0.0347\) \(0.9829 \pm 0.0345\) \(0.9338 \pm 0.0437\) \(0.8157 \pm 0.0998\) \(0.8163 \pm 0.1047\)
Sen \(\underline{0} .\underline{8929} \pm \underline{0} .\underline{1349}\) \(0.8804 \pm 0.1107\) \(0.8661 \pm 0.1661\) \(0.8018 \pm 0.1201\) \(0.6750 \pm 0.1976\) \(0.6625 \pm 0.1872\)
Prec \(\underline{0} .\underline{9639} \pm \underline{0} .\underline{0583}\) \(0.8294 \pm 0.0786\) \(0.9579 \pm 0.0690\) \(0.8696 \pm 0.0766\) \(0.7256 \pm 0.1569\) \(0.7324 \pm 0.1391\)
Meshkini and Ghassemian ACC% \(\textit{82.54}\% \pm \textit{8.52}\%\) \(\textit{87.90}\% \pm \textit{6.25}\%\) \(\textit{86.14}\% \pm \textit{7.25}\%\) \(\textit{84.23}\% \pm \textit{9.54}\%\) \(81.95\% \pm 8.73\%\) \(84.38 \% \pm 8.86\%\)
AUC \(\textit{0.9462} \pm \textit{0.0326}\) \(\textit{0.9331} \pm \textit{0.0592}\) \(\textit{0.9293} \pm \textit{0.0512}\) \(\textit{0.8800} \pm \textit{0.0676}\) \(0.9013 \pm 0.0455\) \(0.9329 \pm 0.0546\)
Sen \(\textit{0.7214} \pm \textit{0.1475}\) \(\textit{0.7821} \pm \textit{0.1380}\) \(\textit{0.7589} \pm \textit{0.1892}\) \(\textit{0.7286} \pm \textit{0.1828}\) \(0.7196 \pm 0.1409\) \(0.8196 \pm 0.1896\)
Prec \(\textit{0.8707} \pm \textit{0.0937}\) \(\textit{0.9389} \pm \textit{0.0811}\) \(\textit{0.9324} \pm \textit{0.0921}\) \(\textit{0.8973} \pm \textit{0.1164}\) \(0.8574 \pm 0.1125\) \(0.8519 \pm 0.1161\)
Zhou et al. ACC% \(\textit{63.60}\% \pm \textit{7.56}\%\) \(\textit{75.85}\% \pm \textit{8.72}\%\) \(\textit{55.15}\% \pm \textit{8.52}\%\) \(\textit{67.76}\% \pm \textit{10.21}\%\) \(58.79\% \pm 6.61\%\) \(72.21 \% \pm 9.57\%\)
AUC \(\textit{0.6752} \pm \textit{0.1375}\) \(\textit{0.8544} \pm \textit{0.1017}\) \(\textit{0.5771} \pm \textit{0.1091}\) \(\textit{0.6769} \pm \textit{0.1280}\) \(0.6006 \pm 0.1252\) \(0.7699 \pm 0.1103\)
Sen \(\textit{0.4571} \pm \textit{0.1358}\) \(\textit{0.5964} \pm \textit{0.1550}\) \(\textit{0.4054} \pm \textit{0.1296}\) \(\textit{0.4804} \pm \textit{0.1944}\) \(0.3375 \pm 0.1814\) \(0.5821 \pm 0.1092\)
Prec \(\textit{0.6442} \pm \textit{0.1660}\) \(\textit{0.8349} \pm \textit{0.1336}\) \(\textit{0.5157} \pm \textit{0.1430}\) \(\textit{0.6812} \pm \textit{0.1553}\) \(0.5764 \pm 0.1998\) \(0.7665 \pm 0.1601\)
Dong et al. ACC% \(\textit{91.58}\% \pm \textit{6.33}\%\) \(\textit{86.07}\% \pm \textit{11.08}\%\) \(\textit{79.96}\% \pm \textit{6.46}\%\) \(\textit{83.57}\% \pm \textit{12.54}\%\) \(83.13\% \pm 5.94\%\) \(75.74\% \pm 5.06\%\)
AUC \(\textit{0.9777} \pm \textit{0.0372}\) \(\textit{0.9345} \pm \textit{0.0743}\) \(\textit{0.9234} \pm \textit{0.0547}\) \(\textit{0.9297} \pm \textit{0.0829}\) \(0.9342 \pm 0.0507\) \(0.8575 \pm 0.0826\)
Sen \(\textit{0.9589} \pm \textit{0.0663}\) \(\textit{0.8125} \pm \textit{0.1831}\) \(\textit{0.8089} \pm \textit{0.1007}\) \(\textit{0.7536} \pm \textit{0.2074}\) \(0.8143 \pm 0.1381\) \(0.6625 \pm 0.1112\)
Prec \(\textit{0.8844} \pm \textit{0.1172}\) \(\textit{0.8661} \pm \textit{0.1076}\) \(0.7673 \pm 0.0976\) \(\textit{0.8843} \pm \textit{0.1575}\) \(0.8252 \pm 0.0942\) \(0.7731 \pm 0.0968\)
(2): Proposed Shearlet-based methods for textured bio-medical image classification
CM ACC% \(\underline{95} .\underline{70} \% \pm \underline{7} .\underline{80} \%\) \(86.76\% \pm 5.35\%\) \(92.61\% \pm 8.25\%\) \(84.82\% \pm 8.92\%\) \(85.40\% \pm 6.56\%\) \(78.20\% \pm 13.74\%\)
AUC \(\underline{0} .\underline{9860} \pm \underline{0} .\underline{0353}\) \(0.9427 \pm 0.0347\) \(0.9709 \pm 0.0457\) \(0.9372 \pm 0.0451\) \(0.9440 \pm 0.0462\) \(0.8992 \pm 0.0653\)
Sen \(\underline{0} .\underline{9446} \pm \underline{0} .\underline{0983}\) \(0.8679 \pm 0.1226\) \(0.9304 \pm 0.0736\) \(0.7946 \pm 0.1404\) \(0.7946 \pm 0.0782\) \(0.7214 \pm 0.2090\)
Prec \(\underline{0} .\underline{9589} \pm \underline{0} .\underline{0945}\) \(0.8483\pm 0.0682\) \(0.9181 \pm 0.1222\) \(0.8691 \pm 0.1277\) \(0.8708 \pm 0.1060\) \(0.7756 \pm 0.1485\)
LBP ACC% \(88.49\% \pm 5.25\%\) \(86.69\% \pm 7.24\%\) \(84.12\% \pm 7.49\%\) \(80.00\% \pm 11.99\%\) \(89.71\% \pm 9.45\%\) \(82.43\% \pm 8.30\%\)
AUC \(0.9592 \pm 0.0544\) \(0.9418 \pm 0.0416\) \(0.9500 \pm 0.0670\) \(0.8944 \pm 0.0836\) \(0.9419 \pm 0.0538\) \(0.9248 \pm 0.0388\)
Sen \(0.8821 \pm 0.1152\) \(0.7982 \pm 0.1223\) \(0.8518 \pm 0.1021\) \(0.7625 \pm 0.1762\) \(0.8946 \pm 0.1305\) \(0.7304 \pm 0.1685\)
Prec \(0.8770 \pm 0.0955\) \(0.9000 \pm 0.0905\) \(0.8292 \pm 0.1426\) \(0.7921 \pm 0.1434\) \(0.8825 \pm 0.1080\) \(0.8649 \pm 0.1017\)
LOSIB ACC% \(92.57\% \pm 9.24\%\) \(87.21\% \pm 5.39\%\) \(89.52\% \pm 7.85\%\) \(87.43\% \pm 10.53\%\) \(79.34\% \pm 5.33\%\) \(76.80\% \pm 10.66\%\)
AUC \(0.9798 \pm 0.0395\) \(0.9346 \pm 0.0432\) \(0.9671 \pm 0.0554\) \(0.9344 \pm 0.0570\) \(0.9153 \pm 0.0601\) \(0.8679 \pm 0.0907\)
Sen \(0.9446 \pm 0.0983\) \(0.7964 \pm 0.1155\) \(0.8464 \pm 0.1258\) \(0.8411 \pm 0.1442\) \(0.7732 \pm 0.1465\) \(0.7125 \pm 0.1685\)
Prec \(0.9042 \pm 0.1255\) \(0.9177 \pm 0.0902\) \(0.9264 \pm 0.1119\) \(0.8792 \pm 0.1265\) \(0.7983 \pm 0.1328\) \(0.7579 \pm 0.1123\)
SFTA ACC% \(94.52\% \pm 7.47\%\) \(87.76\% \pm 5.95\%\) \(92.68\% \pm 7.06\%\) \(84.23\% \pm 10.17\%\) \(85.99\% \pm 7.09\%\) \(82.39\% \pm 6.18\%\)
AUC \(0.9846 \pm 0.0350\) \(0.9353 \pm 0.0469\) \(0.9739 \pm 0.0416\) \(0.9192 \pm 0.0937\) \(0.9515 \pm 0.0410\) \(0.9119 \pm 0.0692\)
Sen \(0.9196 \pm 0.0968\) \(0.8196 \pm 0.1576\) \(0.9179 \pm 0.0710\) \(0.7821 \pm 0.1301\) \(0.7946 \pm 0.0979\) \(0.8286 \pm 0.1203\)
Prec \(0.9589 \pm 0.0945\) \(0.9107 \pm 0.0869\) \(0.9246 \pm 0.1059\) \(0.8767 \pm 0.1447\) \(0.8798 \pm 0.0881\) \(0.8186 \pm 0.1372\)
(3): Integrating Shearlet-based existing techniques with our proposed methods
Fusion #1 ACC% \(95.11\% \pm 7.67\%\) \(88.42\% \pm 7.84\%\) \(92.65\% \pm 8.26\%\) \(92.06\% \pm 7.23\%\) \(87.83\% \pm 5.07\%\) \(79.45\% \pm 10.88\%\)
AUC \(0.9860 \pm 0.0353\) \(0.9527 \pm 0.0475\) \(0.9721 \pm 0.0511\) \(0.9645 \pm 0.0603\) \(0.9651 \pm 0.0302\) \(0.8867 \pm 0.0704\)
Sen \(0.9321 \pm 0.0985\) \(0.8750 \pm 0.0807\) \(0.9321 \pm 0.0718\) \(0.9196 \pm 0.0968\) \(0.8661 \pm 0.0597\) \(0.7429 \pm 0.1543\)
Prec \(0.9589 \pm 0.0945\) \(0.8833 \pm 0.1277\) \(0.9181 \pm 0.1222\) \(0.9125 \pm 0.1006\) \(0.8724 \pm 0.0991\) \(0.8111 \pm 0.1621\)
Fusion #2 ACC% \(\underline{96} .\underline{29} \% \pm \underline{7} .\underline{89} \%\) \(90.99\% \pm 7.13\%\) \(92.02\% \pm 7.86\%\) \(90.81\% \pm 8.42\%\) \(87.79\% \pm 5.90\%\) \(79.26\% \pm 8.62\%\)
AUC \(\underline{0} .\underline{9860} \pm \underline{0} .\underline{0353}\) \(0.9580 \pm 0.0446\) \(0.9734 \pm 0.0525\) \(0.9696 \pm 0.0369\) \(0.9601 \pm 0.0322\) \(0.8843 \pm 0.0919\)
Sen \(\underline{0} .\underline{9571} \pm \underline{0} .\underline{0964}\) \(0.9232 \pm 0.0887\) \(0.9304 \pm 0.0736\) \(0.9446 \pm 0.0983\) \(0.8786 \pm 0.0436\) \(0.7411 \pm 0.1399\)
Prec \(\underline{0} .\underline{9589} \pm \underline{0} .\underline{0945}\) \(0.8913 \pm 0.1026\) \(0.9069 \pm 0.1189\) \(0.8783 \pm 0.1274\) \(0.8682 \pm 0.1247\) \(0.7994 \pm 0.1370\)
Fusion #3 ACC% \(94.49\% \pm 7.47\%\) \(85.48 \% \pm 9.86\%\) - - - -
AUC \(0.9860 \pm 0.0353\) \(0.9523 \pm 0.0570\) - - - -
Sen \(0.9304 \pm 0.0998\) \(0.8161 \pm 0.1473\) - - - -
Prec \(0.9478 \pm 0.0957\) \(0.8718 \pm 0.1454\) - - - -
  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