External validation and comparison of MR-based radiomics models for predicting pathological complete response in locally advanced rectal cancer: a two-centre, multi-vendor study.

OBJECTIVES: The aim of this study was two-fold: (1) to develop and externally validate a multiparameter MR-based machine learning model to predict the pathological complete response (pCR) in locally advanced rectal cancer (LARC) patients after neoadjuvant chemoradiotherapy (nCRT), and (2) to compare different classifiers' discriminative performance for pCR prediction. METHODS: This retrospective study includes 151 LARC patients divided into internal (centre A, n = 100) and external validation set (centre B, n = 51). The clinical and MR radiomics features were derived to construct clinical, radiomics, and clinical-radiomics model. Random forest (RF), support vector machine (SVM), logistic regression (LR), K-nearest neighbor (KNN), naive Bayes (NB), and extreme gradient boosting (XGBoost) were used as classifiers. The predictive performance was assessed using the receiver operating characteristic (ROC) curve. RESULTS: Eleven radiomics and four clinical features were chosen as pCR-related signatures. In the radiomics model, the RF algorithm achieved 74.0% accuracy (an AUC of 0.863) and 84.4% (an AUC of 0.829) in the internal and external validation sets. In the clinical-radiomics model, RF algorithm exhibited high and stable predictive performance in the internal and external validation datasets with an AUC of 0.906 (87.3% sensitivity, 73.7% specificity, 76.0% accuracy) and 0.872 (77.3% sensitivity, 88.2% specificity, 86.3% accuracy), respectively. RF showed a better predictive performance than the other classifiers in the external validation datasets of three models. CONCLUSIONS: The multiparametric clinical-radiomics model combined with RF algorithm is optimal for predicting pCR in the internal and external sets, and might help improve clinical stratifying management of LARC patients. KEY POINTS: • A two-centre study showed that radiomics analysis of pre- and post-nCRT multiparameter MR images could predict pCR in patients with LARC. • The combined model was superior to the clinical and radiomics model in predicting pCR in locally advanced rectal cancer. • The RF classifier performed best in the current study.

[Intravoxel incoherent motion dwi in renal tumors: differentiation between benign and malignant masses].

OBJECTIVE: To evaluate the performance of intravoxel incoherent motion (IVIM) diffusion-weighted MR imaging in the differential diagnosis of Angiomyolipoma (AML), clear cell renal cell carcinoma (clear cell RCC) and non-clear cell renal cell carcinoma (non-clear cell RCC). METHOD: Forty-five patients were enrolled and classified into three groups according to the surgical pathology results. There were even amount of subjects (15) in each group of AML group, clear cell RCC group and non-clear cell RCC group (including 6 Papillary RCC and 9 Chromophobe RCC). ALL the patients underwent respiratory-triggered DWI on 3.0T MR system using multi b-values of 0, 20, 50, 100, 150, 200, 400, 600, 800 sec/mm2. Pure molecular-based (D), perfusion-related (D*) and vascular fraction (f) were calculated using a bi-exponential model. Comparisons of derived parameters derived by IVIM-DWI among three groups were performed using One-way ANOVA. RESULTS: D in clear cell RCC (1.8±1.5)×10(-3) mm2/s was significantly higher than that in AML (0.7±0.3)×10(-3) mm2/s or non-clear cell RCC (0.9±0.3)×10(-3) mm2/s (P<0.05), while no statistical difference was observed between AML and non-clear cell RCC (P=0.807). D* demonstrated significant difference between AML (33.3±18.2)×10(-3) mm2/s and non-clear cell RCC (17.8±17.3)×10(-3) mm2/s, (P<0.05). No observed significant difference in f was revealed between any two groups. CONCLUSION: Pure water molecules (true diffusion) and capillary microcirculation (pseudo-diffusion) may result in the observed difference of parameters in these three types of renal tumors. D value can contribute to differentiate the clear cell RCC from AML, and between clear cell and non-clear cell RCC as well. D* value may be applied to differentiate AML from non-clear cell RCC.

[Study of ADC value in healthy adults kidney of different age groups].

OBJECTIVE: To evaluate the changes of ADC value on renal parenchyma at different age groups in healthy adults. METHODS: One hundred healthy adults were divided into five groups based on age:namely 18-29 years, 30-39 years, 40-49 years, 50-59 years and 60-69 years. All adults underwent diffusion-weighted magnetic resonance(MR) imaging of the kidneys with b values of 800 s/mm(2). ADC values of renal parenchyma were measured by the manufacturer's software. RESULTS: With the increasing of age , the renal parenchyma ADCs decrease. The ADCs of renal parenchyma in five age groups were (2.07 ± 0.10) ×10(-3)mm(2)/s, (2.06 ± 0.12) ×10(-3)mm(2)/s, (2.03 ± 0.10) ×10(-3)mm(2)/s, (1.98 ± 0.17) ×10(-3)mm(2)/s and (1.94 ± 0.12) ×10(-3)mm(2)/s, which had statistically significant different (F = 3.375, P < 0.05). CONCLUSION: (1)The ADCs of renal parenchyma may be influenced by age , with the increasing of age, ADCs of normal kidney presented tenuous decreasing tendency especially the people aged ≥ 40 years; (2)in clinical application, we ought to pay attention to the influence of age factor. Income data of this study can be took as normal reference value in other more study.

[MR appearance of multifocal clear cell renal cell carcinoma].

OBJECTIVE: To document the magnetic resonance (MR) imaging features of multifocal clear cell renal cell carcinoma (RCC). METHODS: The MR findings of 11 cases of pathologically-proved multifocal clear cell RCCs were reviewed retrospectively from January 2008 to December 2010. All patients underwent MR in a 1.5 T or a 3.0 T scanner. The MR features of the lesions were analyzed, with emphasis on the location, the signal intensity on T2-weighted imaging (T2WI), in regards to pseudocapsule, lipid component, hemorrhage, cystic degeneration, and enhancement pattern. Meanwhile, from these six aspects, the MR appearance of renal cell carcinomas occurring in the same patient were compared. RESULTS: 24 tumors (mean diameter, 3.1 cm) in the 11 patients (6 men, 5 women; age range, 36-69 years; mean age, 50.6 years) were detected. Multiple lesions in bilateral kidneys were found in 9 patients (9/11), while in unilateral kidney 2 patients (2/11). Hyperintensity on T2WI were observed in 16 lesions (16/24), while hypointensity 8 lesions (8/24); pseudocapsule was seen in all lesions; hemorrhage in 4 lesions (4/24), lipid component in 12 lesions (12/24) and cystic degeneration in 18 lesions (18/24), persistent enhancement in 22 lesions (22/24) and rapid wash-out pattern in 2 lesions (2/24). All lesions showed moderate or intense enhancement. The MR appearances of lesions in the same patient were identical in 5 cases (5/11), different in 6 cases (6/11). CONCLUSION: Multifocal clear cell RCCs occur more commonly in bilateral kidneys simultaneously than in unilateral kidney. The lesions in the same patient can show different MR imaging features, which probably facilitate the preoperative determination.

[Magnetic resonance imaging findings of Chromophobe renal cell carcinoma].

OBJECTIVE: To document the MRI features of Chromophobe cell carcinoma and to explore whether MR features vary with the tumor size. METHODS: MRI features of 34 patients (16 male, 18 female, age range from 24 - 61, average age is 49 years old), totally 35 focuses with histologically proved chromophobe cell carcinoma were studied retrospectively. All the patients underwent MR plain and dynamic contrast-enhanced scan before their surgery. Variation of signal intensity (3D LAVA) and ADC values of the lesions, and the analysis of images of T1WI, T2WI and chemical shift were all evaluated on the GE Advantage 4.4 work station. All lesions were categorized into 2 groups (tumor diameter ≤ 4 cm, or > 4 cm). The difference of imaging characteristics between these two groups was analyzed using Fisher exact test. Signal intensity variation and ADC values were analyzed by using one-way ANOVA method. RESULTS: None of the 35 cases contained fat or lipid. On T2WI 27 cases showed slightly low signal intensity (77.1%). In all cases, 4 appeared local cystic change (11.4%); 4 spotty hemorrhage (11.4%); 5 necrosis (covering less than 20% of whole tumor) (14.2%); 30 clear pseudo capsule (85.7%); 29 round-like lesions (the difference among the length, width and height was within 0.5 cm) (82.8%); and no cases showed signs of invasiveness or metastasis. The average changes of signal intensity of all the 35 cases were 119.8% on cortex period, 176.4% on medulla period and 154.5% on delayed phase. The mean ADC value of tumor was 1.08 ± 0.28×10(-3)mm(2)/s. 35 lesions were divided into two groups , 21 in group 1(diameter ≤ 4 cm) and 14 in group 2 (diameter > 4 cm). Cystic degeneration was seen in 0/21 in group 1 versus 4/14 in group 2 respectively, and hemorrhage 0/21 versus 4/14, necrosis 0/21 versus 5/14 , central scar 2/21 versus 8/14. The difference of these findings between two groups demonstrated statistical significance (P < 0.05). Variation of signal intensity and ADC values in two groups has no statistical significance. CONCLUSION: The MR features of Chromophobe renal cell carcinoma were hypointensity on T2WI, clear pseudocapsule, However, cystic degeneration, hemorrhage, necrosis and central scar are more common in larger tumors.

Coronal diffusion-weighted magnetic resonance imaging of the kidney: agreement with axial diffusion-weighted magnetic imaging in terms of apparent diffusion coefficient values.

BACKGROUND: Coronal diffusion-weighted magnetic resonance imaging (DW-MRI) and apparent diffusion coefficient (ADC) values have gradually become applied (following conventional axial DW-MRI) in the renal analysis. To explore whether data obtained using coronal DW-MRI are comparable with those derived using axial DW-MRI, this preliminary study sought to assess the agreement in renal ADC values between coronal DW-MRI and axial DW-MRI. METHODS: Thirty-four healthy volunteers were enrolled in the study; written consents were obtained. All subjects underwent respiratory-triggered axial and coronal DW-MRI using a 1.5-MR system with b values of 0 and 800 s/mm 2 . The signal-to-noise ratios (SNRs) of the two DW-MRI sequences were measured and statistically compared using the paired t-test. The extent of agreement of ADC values of the upper pole, mid-pole, and lower pole of the kidney; the mean ADC values of the left kidney and right kidney; and the mean ADC values of the bilateral kidneys were evaluated via calculation of intraclass correlation coefficients (ICCs) or Bland-Altman method between the two DW-MRI sequences. RESULTS: The SNR of coronal DW-MR images was statistically inferior to that of axial DW-MR images (P < 0.001). The ICCs of the ADC values of each region of interest, and the mean ADC values of bilateral kidneys, between the two sequences, were greater than 0.5, and the mean ADCs of the bilateral kidneys demonstrated the highest ICC (0.869; 95% confidence interval: 0.739-0.935). In addition, 94.1% (32/34), 94.1% (32/34), and 97.1% (31/34) of the ADC bias was inside the limits of agreement in terms of the mean ADC values of the left kidneys, right kidneys, and bilateral kidneys when coronal and axial DWI-MRI were compared. CONCLUSIONS: ADC values derived using coronal DW-MRI exhibited moderate-to-good agreement to those of axial DW-MRI, rendering the former an additional useful DW-MRI method, and causing the ADC values derived using the two types of DW-MRI to be comparable.