1 Introduction
2 Materials and methods
2.1 Patients
2.2 Research method
2.2.1 Clinicopathological data and follow-up contents
2.2.2 CT image acquisition
2.2.3 Region of interest (ROI) delineation and radiomics feature extraction
-
1. Image preprocessing: The plug in Pyradiomics (version 3.0.1) was used to resample all images, the parameters binWidth were set to 25, the parameters resampledPixelSpacing were set to 1.0 mm, Then, the image nonlinear intensity transformation and wavelet transform are performed on the original CT image to reduce computational interference and improve feature recognition ability.
-
2. Feature extraction and calculation: Based on the original CT image and processing, 1316 radiomics features were extracted from each ROI: firstorder features that describe single pixels of the image; shape features that describe the geometric characteristics of the ROI; and texture features that reflect the homogeneous phenomenon of vision in the image, including 252 firstorder features, 14 shape features, 336 gray-level cooccurrence matrix (GLCM) features, 224 gray-level run length matrix (GLRLM) features, 224 gray-level size zone matrix (GLSZM) features, 70 neighborhood gray tone difference matrix (NGTDM) features, and 196 gray-level dependence matrix (GLDM) features.
2.2.4 Local and metastatic recurrence-related feature selection and radiomics scoring model establishment
-
1. Feature selection: First, to evaluate the repeatability of radiomics feature extraction, we calculated the interclass correlation coefficient (ICC) of the feature on the basis of the previous two experts' work and selected features with ICC>0.75. Subsequently, in training set, to maintain the comparability of different features and reduce the imbalance in the importance of each feature caused by the differences in the mean and variance of each feature, we normalized the features of the Z score: z= xi-μ/δ. Finally, to reduce the difficulty of model learning and data noise in the later stage, the coefficient of association and analysis of variance were used for feature selection. After removing other redundant and useless features, the features of p<0.05 were retained as an effective factor for the next feature selection.
-
2. Radiomics Scoring Model Establishment: To solve the overfitting problem of high dimensional data, improve the performance of the model, and select the most effective prognostic features, the LASSO regression algorithm was used to select the deviation of the measurement index, and the optimal feature was selected after 10-fold cross verification. In the training group, a logistic regression model with radiomics scoring was constructed based on the selected radiomics features to obtain the regression coefficients. Then, a linear combination of these selected features and their corresponding regression coefficients was used to calculate the Rad-Score for each patient. Then, validation was performed in the validation group. The score value was used for subsequent analysis. The study design for radiomics is shown in Figure 1.
2.2.5 Establishment of radiomics-clinical model and efficacy assessment
2.3 Statistical analysis
3 Results
3.1 Patient characteristics
Factors | Overall (N = 134) | Training set, n (%) | Validation set, n (%) | ||||
---|---|---|---|---|---|---|---|
No recurrence (N = 74) | Recurrence (N = 19) | P-value | No recurrence (N = 36) | Recurrence (N = 5) | P-value | ||
Gender, n (%) | 0.310 | 0.501 | |||||
Female | 15 (11.2%) | 4 (5.4%) | 3 (15.8%) | 8 (22.2%) | 0 (0%) | ||
Male | 119 (88.8%) | 70 (94.6%) | 16 (84.2%) | 28 (77.8%) | 5 (100%) | ||
Age, n (%) | 0.917 | 0.958 | |||||
< 60 | 37 (27.6%) | 19 (25.7%) | 4 (21.1%) | 12 (33.3%) | 2 (40.0%) | ||
≥ 60 | 97 (72.4%) | 55 (74.3%) | 15 (78.9%) | 24 (66.7%) | 3 (60.0%) | ||
Number of tumors, n (%) | 0.990 | 0.029 | |||||
Single | 89 (66.4%) | 48 (64.9%) | 12 (63.2%) | 28 (77.8%) | 1 (20.0%) | ||
Multiple | 45 (33.6%) | 26 (35.1%) | 7 (36.8%) | 8 (22.2%) | 4 (80.0%) | ||
Size of tumors, n (%) | 0.045 | 0.776 | |||||
< 3 | 52 (38.8%) | 35 (47.3%) | 3 (15.8%) | 13 (36.1%) | 1 (20.0%) | ||
≥ 3 | 82 (61.2%) | 39 (52.7%) | 16 (84.2%) | 23 (63.9%) | 4 (80.0%) | ||
Grade, n (%) | 0.070 | 0.308 | |||||
High | 93.0 (69.4%) | 57 (77.0%) | 19 (100%) | 24 (66.7%) | 5 (100%) | ||
Low | 41.0 (30.6%) | 17 (23.0%) | 0 (0%) | 12 (33.3%) | 0 (0%) | ||
pT stage, n (%) | < 0.001 | 0.037 | |||||
≥ 2 | 62 (46.3%) | 27 (36.5%) | 16 (84.2%) | 14 (38.9%) | 5 (100%) | ||
< 2 | 72 (53.7%) | 47 (63.5%) | 3 (15.8%) | 22 (61.1%) | 0 (0%) | ||
pN stage, n (%) | 0.015 | 0.509 | |||||
N0 | 125 (93.3%) | 72 (97.3%) | 15 (78.9%) | 34 (94.4%) | 4(80.0%) | ||
N1/N2 | 9 (6.7%) | 2 (2.7%) | 4 (21.1%) | 2 (5.6%) | 1 (20.0%) | ||
Histology type, n (%) | 0.992 | 0.038 | |||||
Urothelial carcinoma | 129 (96.3%) | 73 (98.6%) | 19 (100%) | 34 (94.4%) | 3 (60%) | ||
Non-urothelial carcinoma | 5 (3.7%) | 1 (1.4%) | 0 (0%) | 2 (5.6%) | 2 (40%) | ||
RFS (months) | < 0.001 | < 0.001 | |||||
Median [Min, Max] | 60 [3, 96] | 51 [36, 95] | 13 [3, 35] | 60 [36, 96] | 13 [6, 29] |
3.2 Selection of radiomics features and establishment of radiomic model
3.3 Establishment of a radiomic-clinical model and its efficacy assessment
Univariate | Multivariate | |||
---|---|---|---|---|
Factors | OR (95% CI) | P | OR (95% Cl) | P |
Gender | 0.305 (0.061–1.671) | 0.144 | ||
Age | 1.295 (0.409–4.971) | 0.678 | ||
Number of tumors | 1.076 (0.362–3.021) | 0.890 | ||
Size of tumors | 4.786 (1.444–21.829) | 0.020 | 1.774 (0.415–9.294) | 0.456 |
Grade | 2.538 (0.637–16.949) | 0.243 | ||
pT stage | 9.259 (2.786–43.478) | 0.001 | 5.682 (1.531–27.778) | 0.015 |
pN stage | 9.600 (1.716–73.979) | 0.013 | 3.924 (0.515–39.299) | 0.201 |
Histology type | 4.056 (0.155–105.917) | 0.330 | ||
Rad-Score | 3.875 (1.852–9.935) | 0.002 | 2.933 (1.336–7.951) | 0.018 |
Training set | Validation set | |||||||
---|---|---|---|---|---|---|---|---|
Model types | AUC | 95% CI | Harrell’s C-index | p-value | AUC | 95% CI | Harrell’s C-index | p-value |
Radiomics clinical model | 0.998 | 0.995–1 | 0.957 | < 0.05 | 0.960 | 0.896–1 | 0.855 | < 0.05 |
Clinical model | 0.871 | 0.803–0.940 | 0.743 | < 0.05 | 0.904 | 0.813–0.995 | 0.807 | < 0.05 |
Radiomics model | 0.852 | 0.778–0.927 | 0.574 | < 0.05 | 0.788 | 0.637–0.939 | 0.607 | < 0.05 |