Table 2.
Related studies and strategies of CT-/MRI-based BCa grading during the past 20 years.
| Study | Patient | Imaging | Target | Approach or strategy | Results and findings | |
|---|---|---|---|---|---|---|
| Tuncbilek et al., 2009 (97) | 24 patients from single center | DCE | Tumor | Extracting peak time
enhancement in the first (Emax/1), second (Emax/2), third (Emax/3), fourth (Emax/4) and fifth (Emax/5) minute after contrast administration, and the steepest slope for statistical analysis with tumor grade. |
Emax/1and steepest slope had statistically significant correlation with tumor grade. | |
| Avcu et al., 2011 (98) | 63 patients from single center | DWI | Tumor | Mean ADC values were measured from the tumor mass. | The mean ADC value were significantly different between the high- and low-grade BCa. | |
| Rosenkrantz et al., 2013 (37) | 37 patients from double centers | T2WI, DWI |
Tumor | Tumor diameter, normalized T2 signal intensity and mean ADC value were extracted. | Mean ADC value was statistically significant between the high- and low-grade BCa, with an AUC of 0.804 for the classification of this two groups. | |
| Kobayashi et al., 2014 (104) | 132 patients from single center | DWI | Tumor | Mean ADC value was calculated. | Mean ADC value was significantly lower in tumors with higher Ki-67 Lis and higher grade. | |
| Sevcenco et al., 2014 (105) | 43 patients from single center | DWI | Tumor | Mean ADC value was obtained. | Mean ADC value achieved favorable performance in predicting tumor grade, with an AUC of 0.906. | |
| Sevcenco et al., 2014 (106) | 41 patients from single center | DWI | Tumor | Mean ADC value, p53 and p21 were obtained. | Mean ADC value and p21 were the independent predictors for BCa grade, with an AUC of 0.981. | |
| Wang et al., 2014 (102) | 30 patients from single center | DWI | Tumor and referenced regions like urine | Mean ADC value and normalized ADC (nADC) values were calculated. | The performance of using the nADC with urine as reference was the best, with the AUC of 0.995. | |
| Zhang et al., 2017 (107) | 128 patients from single center | *CECT | Tumor | Six texture features, including mean, SD, entropy, mean of positive pixels (MPP), skewness and kurtosis, were extracted. | Mean, entropy and MPP were significantly different between the high-grade BCa and low-grade on both unenhanced and enhanced images. MPP obtained from unenhanced images achieved the best performacne, with the AUC of 0.779. | |
| Mammen et al., 2017 (108) | 48 patients from single center | CT | Tumor | Texture features including Kurtosis, skewness and entropy, were extracted. | Only entropy showed significant inter-group differences, and it achieved an AUC of 0.83 in differentiation of low- and high-grade BCa. | |
| Zhang et al., 2017 (25) | 61 patients form single center | DWI ADC maps |
Tumor | 102 radiomics features, including the histogram and GLCM features | The model developed could achieve favorable performance for BCa grading, with the AUC of 0.861, significantly better than that of using the ADC value alone. | |
| Wang et al., 2019 (76) | 100 patients from single center | T2WI, DWI and ADC maps |
Tumor | 924 features were extracted, including morphological features and six categories of texture features like histogram features, GLCM features, *GLRLM features, *GLSZM features, *NGTDM features, and *GLDM features. | The multi-modal MRI-based radiomics approach has the potential in preoperative grading of BCa, with the AUC of 0.9276. | |
| Wang et al., 2020 (15) | 58 patients from single center | T2*-weighted imaging and DWI | Tumor | Apparent transverse relaxation rate R2* and mean ADC value were calculated. | R2* and mean ADC value were significantly different between low- and high-grade BCa, with the AUC of 0.714 and 0.779 in the classification process, respectively. | |
| Zhang et al., 2020 (109) | 145 patients from single center | CT | Tumor | 1316 radiomics features, involving the morphological features, histogram features, GLCM features, GLRLM features, GLSZM features, GLDM features, were calculated. |
The proposed radiomics model achieved a good performance, with AUC of 0.85 using the testing cohort. | |
*CECT indicates the contrast enhanced CT.
*GLRLM indicates the gray-level run length matrix; GLSZM indicates the gray-level size zone matrix; NGTDM indicates the neighborhood gray tone difference matrix; GLDM indicates the gray-level dependence matrix.