Differentiating between benign and malignant sinonasal lesions using dynamic contrast-enhanced MRI and intravoxel incoherent motion

Deep transfer learning radiomics for distinguishing sinonasal malignancies: a preliminary MRI study

PURPOSE

This study aimed to assess the diagnostic accuracy of combining MRI hand-crafted (HC) radiomics features with deep transfer learning (DTL) in identifying sinonasal squamous cell carcinoma (SCC), adenoid cystic carcinoma (ACC), and non-Hodgkin's lymphoma (NHL) using various machine learning (ML) models.

METHODS

A retrospective analysis of 132 patients (50 with SCC, 42 with NHL, 40 with ACC) was conducted. The dataset was split 80/20 into training and testing cohorts. HC radiomics and DTL features were extracted from T2-weighted, ADC, and contrast-enhanced T1-weighted MRI images. ResNet50, a pre-trained convolutional neural network, was used for DTL feature extraction. LASSO regression was applied to select features and create radiomic signatures. Seven ML models were evaluated for classification performance.

RESULTS

The radiomic signature included 24 hC and 8 DTL features. The support vector machine (SVM) model achieved the highest accuracy (92.6%) in the testing cohort. The SVM model's ROC analysis showed macro-average and micro-average AUC values of 0.98 and 0.99. AUCs for ACC, NHL, and SCC were 0.99, 0.97, and 1.00. K-nearest neighbors (KNN) and XGBoost also showed AUC values above 0.90.

CONCLUSION

Combining MRI-based HC radiomics and DTL features with the SVM model enhanced differentiation between sinonasal SCC, NHL, and ACC.

Correlation between diffusion-weighted image-derived parameters and dynamic contrast-enhanced magnetic resonance imaging-derived parameters in the orofacial region

Background

Diffusion-weighted imaging (DWI) and dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) are widely used in the orofacial region. Furthermore, quantitative analyses have proven useful. However, a few reports have described the correlation between DWI-derived parameters and DCE-MRI-derived parameters, and the results have been controversial.

Purpose

To evaluate the correlation among parameters obtained by DWI and DCE-MRI and to compare them between benign and malignant lesions.

Material and Methods

Fifty orofacial lesions were analysed. The apparent diffusion coefficient (ADC), true diffusion coefficient (D), pseudodiffusion coefficient (D*) and perfusion fraction (f) were estimated by DWI. For DCE-MRI, TK model analysis was performed to estimate physiological parameters, for example, the influx forward volume transfer constant into the extracellular-extravascular space (EES) (Ktrans) and fractional volumes of EES and plasma components (ve and vp).

Results

Both ADC and D showed a moderate positive correlation with ve (ρ = 0.640 and 0.645, respectively). Ktrans showed a marginally weak correlation with f (ρ = 0.296), while vp was not correlated with f or D*; therefore, IVIM perfusion-related parameters and TK model perfusion-related parameters were not straightforward. Both D and ve yielded high diagnostic power between benign lesions and malignant tumours with areas under the curve (AUCs) of 0.830 and 0.782, respectively.

Conclusion

Both D and ve were reliable parameters that were useful for the differential diagnosis. In addition, the true diffusion coefficient (D) was affected by the fractional volume of EES.

Diagnosis and differential diagnosis of dermatofibrosarcoma protuberans: Utility of high-resolution dynamic contrast-enhanced (DCE) MRI

BACKGROUND

Dermatofibrosarcoma protuberans (DFSP) is a kind of low-grade malignant spindle cell neoplasm, the diagnosis, and treatment, which have markedly attracted clinicians' attention for its repeated recurrence. High-resolution magnetic resonance imaging (HR-MRI) has shown unique capabilities in diagnosis of various cutaneous tumors.

MATERIALS AND METHODS

Data of 29 patients with clinically suspected DFSPs and undergoing dynamic contrast-enhanced (DCE) HR-MRI preoperatively were prospectively collected. The HR-MRI qualitative features were evaluated and compared. The DCE-associated quantitative parameters and the time-signal intensity curve (TIC) types were provided using DCE sequences.

RESULTS

A total of 7 DFSPs, nine dermatofibromas (DF, including four cases of cellular variant [CDF]), 12 keloids, and one nodular fasciitis were enrolled. DFSP showed the largest major diameter and the deepest depth. Five DFSPs (71.4%) showed ill-defined margins as well as infiltration of peripheral adipose. All DFSPs showed irregular shape. Most DFSPs presented hyperintensity on T₂ WI (71.4%) and iso-intensity on T₁ WI (85.7%). Six cases (85.7%) had significant enhancement, and six cases (85.7%) had homogeneous enhancement. There were significant differences of K^trans , K_ep , V_e and iAUC values among DFSPs, DFs, and keloids, and DFSP had the highest values for these parameters. Six DFSPs (85.7%) and four CDFs (100%) showed type-III TICs, while the other lesions showed type-Ⅰor type-Ⅱ TICs.

CONCLUSIONS

DCE-HR-MRI could show the growth characteristics of DFSPs, which was of great value for the diagnosis and differential diagnosis of DFSPs and was helpful for the determination of treatment options, thereby to improve the prognosis of patients.

Diffusion kurtosis imaging and dynamic contrast-enhanced MRI for the differentiation of parotid gland tumors

OBJECTIVE

To assess the usefulness of combined diffusion kurtosis imaging (DKI) and dynamic contrast-enhanced MRI (DCE-MRI) in the differentiation of parotid gland tumors.

METHODS

Seventy patients with 80 parotid gland tumors who underwent DKI and DCE-MRI were retrospectively enrolled and divided into four groups: pleomorphic adenomas (PAs), Warthin tumors (WTs), other benign tumors (OBTs), and malignant tumors (MTs). DCE-MRI and DKI quantitative parameters were measured. The Kruskal-Wallis H test and post hoc test with Bonferroni correction and ROC curve were used for statistical analysis.

RESULTS

WTs demonstrated the highest K_ep value (median 1.89, interquartile range [1.46-2.31] min^-1) but lowest V_e value (0.20, [0.15-0.25]) compared with PAs (K_ep, 0.34 [0.21-0.55] min^-1; V_e, 0.36 [0.24-0.43]), OBTs (K_ep, 1.22 [0.27-1.67] min^-1; V_e, 0.28 [0.25-0.41]), and MTs (K_ep, 0.71 [0.50-1.23] min^-1; V_e, 0.35 [0.26-0.45]) (all p < .05). MTs had the lower D value (1.10, [0.88-1.29] × 10^-3 mm²/s) compared with PAs (1.81, [1.60-2.20] × 10^-3 mm²/s) and OBTs (1.57, [1.32-1.89] × 10^-3 mm²/s) (both p < .05). PAs had the lower K^trans value (0.12, [0.07-0.18] min^-1) compared with OBTs (0.28, [0.11-0.50] min^-1) (p < .05). The cutoff values of combined K_ep and V_e, D, and K^trans to distinguish WTs, MTs, and PAs sequentially were 1.06 min^-1, 0.28, 1.46 × 10^-3 mm²/s, and 0.21 min^-1, respectively (accuracy, 89% [71/80], 91% [73/80], 78% [62/80], respectively).

CONCLUSION

The combined use of DKI and DCE-MRI may help differentiate parotid gland tumors.

KEY POINTS

• The combined use of DKI and DCE-MRI could facilitate the understanding of the pathophysiological characteristics of parotid gland tumors. • A stepwise diagnostic diagram based on the combined use of DCE-MRI parameters and the diffusion coefficient is helpful for accurate preoperative diagnosis in parotid gland tumors and may further facilitate the clinical management of patients.

Dual-energy CT in differentiating benign sinonasal lesions from malignant ones: comparison with simulated single-energy CT, conventional MRI, and DWI

OBJECTIVES

To explore the value of dual-energy CT (DECT) for differentiating benign sinonasal lesions from malignant ones, and to compare this finding with simulated single-energy CT (SECT), conventional MRI (cMRI), and diffusion-weighted imaging (DWI).

METHODS

Patients with sinonasal lesions (38 benign and 34 malignant) who were confirmed by histopathology underwent DECT, cMRI, and DWI. DECT-derived parameters (iodine concentration (IC), effective atomic number (Eff-Z), 40-180 keV (20-keV interval), virtual non-enhancement (VNC), slope (k), and linear-mixed 0.3 (Mix-0.3)), DECT morphological features, cMRI characteristics, and ADC value of benign and malignant tumors were compared using t test or chi-square test. Receiver operating characteristic (ROC) curve was performed to evaluate the diagnostic performance, and the area under the ROC curve (AUC) was compared using the Z test to select the optimal diagnostic approach.

RESULTS

Significantly higher DECT-derived single parameters (IC, Eff-Z, 40 keV, 60 keV, 80 keV, slope (k), Mix-0.3) were found in malignant lesions than those of benign sinonasal lesions (all p < 0.004, Bonferroni correction). Combined quantitative parameters (IC, Eff-Z, 40 keV, 60 keV, 80 keV, slope (k)) can improve the diagnostic efficiency for discriminating these two entities. Combination of DECT quantitative parameters and morphological features can further improve the overall diagnostic performance, with AUC, sensitivity, specificity, and accuracy of 0.935, 96.67%, 90.00%, and 93.52%. Moreover, the AUC of DECT was higher than those of Mix-0.3 (simulated SECT), cMRI, DWI, and cMRI+DWI.

CONCLUSIONS

Compared with simulated SECT, cMRI, and DWI, DECT appears to be a more accurate imaging technique for differentiating benign from malignant sinonasal lesions.

KEY POINTS

• DE can differentiate benign sinonasal lesions from malignant ones based on DECT-derived qualitative parameters. • DECT appears to be more accurate in the diagnosis of sinonasal lesions when compared with simulated SECT, cMRI, and DWI.

Intravoxel Incoherent Motion Magnetic Resonance Imaging for Prediction of Induction Chemotherapy Response in Locally Advanced Hypopharyngeal Carcinoma: Comparison With Model-Free Dynamic Contrast-Enhanced Magnetic Resonance Imaging

BACKGROUND

Multiparametric intravoxel incoherent motion (IVIM) provides diffusion and perfusion information for the treatment prediction of cancer. However, the superiority of IVIM over dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI) in locally advanced hypopharyngeal carcinoma (LAHC) remains unclear.

PURPOSE

To compare the diagnostic performance of IVIM and model-free DCE in assessing induction chemotherapy (IC) response in patients with LAHC.

STUDY TYPE

Prospective.

POPULATION

Forty-two patients with LAHC.

FIELD STRENGTH/SEQUENCE

3.0 T MRI, including IVIM (12 b values, 0-800 seconds/mm² ) with a single-shot echo planar imaging sequence and DCE-MRI with a volumetric interpolated breath-hold examination sequence. IVIM MRI is a commercially available sequence and software for calculation and analysis from vendor.

ASSESSMENT

The IVIM-derived parameters (diffusion coefficient [D], pseudodiffusion coefficient [D*], and perfusion fraction [f]) and DCE-derived model-free parameters (Wash-in, time to maximum enhancement [Tmax], maximum enhancement [Emax], area under enhancement curve [AUC] over 60 seconds [AUC₆₀ ], and whole area under enhancement curve [AUC_w ]) were measured. At the end of IC, patients with complete or partial response were classified as responders according to the Response Evaluation Criteria in Solid Tumors.

STATISTICAL TESTS

The differences of parameters between responders and nonresponders were assessed using Mann-Whitney U tests. The performance of parameters for predicting IC response was evaluated by the receiver operating characteristic curves.

RESULTS

Twenty-three (54.8%) patients were classified as responders. Compared with nonresponders, the perfusion parameters D*, f, f × D*, and AUC_w were significantly higher whereas Wash-in was lower in responders (all P-values <0.05). The f × D* outperformed other parameters, with an AUC of 0.84 (95% confidence interval [CI]: 0.69-0.93), sensitivity of 79.0% (95% CI: 54.4-93.9), and specificity of 82.6% (95% CI: 61.2-95.0).

DATA CONCLUSION

The IVIM MRI technique may noninvasively help predict the IC response before treatment in patients with LAHC.

LEVEL OF EVIDENCE

2 TECHNICAL EFFICACY: Stage 2.

Dynamic Contrast-Enhanced MRI Can Quantitatively Discriminate the Original Site From Peripheral Portion of Sinonasal Inverted Papillomas

BACKGROUND

Identification of the original site of sinonasal inverted papillomas (SIPs) is difficult but essential for reducing the recurrence rate. Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) may provide information about tissue perfusion and permeability to solve this problem.

PURPOSE

To investigate the accuracy of DCE-MRI parameters in discriminating between regions of interest (ROIs) in the original site and peripheral portion.

STUDY TYPE

Retrospective.

POPULATION

Ninety consecutive patients with pathologically proven SIP.

FIELD STRENGTH/SEQUENCE

3.0T/DCE-MRI using fast-spoiled gradient recalled (FSPGR) T₁ -weighted images with fat saturation.

ASSESSMENT

ROIs were placed in the original site and the peripheral portion of SIP by two radiologists according to surgical records. Maximum slope of increase (MaxSlope), contrast-enhancement ratio (CER), bolus arrival time (BAT), initial area under the signal intensity-time curve (IAUGC), volume transfer constant (K^trans ), volume of the extravascular extracellular space (v_e ), and rate constant (K_ep ) were calculated and repeated again with a month interval by a radiologist.

STATISTICAL TESTS

Univariate and multivariate analysis was used to determine the best diagnostic parameters, and their performances in discrimination were evaluated by receiver operating characteristic (ROC) curves. Reproducibility was estimated by the intraclass correlation coefficient (ICC).

RESULTS

MaxSlope, CER, IAUGC, K^trans , and v_e were significantly lower (P < 0.05) in the original site than the peripheral portion of SIPs. CER (odds ratio [OR] = 0.227, 95% confidence interval [95% CI] = 0.073-0.704) and v_e (OR = 0.048, 95% CI = 0.004-0.527) were the best indicators for identifying the original ROIs. The combination of CER and v_e had the best diagnostic performance in the discrimination between the ROIs (the area under the curve [AUC]: 0.937; 95% CI: 0.896-0.974).

DATA CONCLUSION

DCE-MRI derived parameter values differed between the original site and the peripheral portion of SIPs. The model combining CER and v_e appears to be able to accurately distinguish the original from peripheral ROIs.

LEVEL OF EVIDENCE

4 TECHNICAL EFFICACY STAGE: 2.

Clinical value of high-resolution dynamic contrast-enhanced (DCE) MRI in diagnosis of cutaneous squamous cell carcinoma

BACKGROUND

While uncomplicated cases of skin squamous cell carcinoma (cSCC) can be treated with surgery topical therapy alone, more objective and non-invasive examination methods are needed to guide clinicians to make more detailed biopsy and surgical plans for lesions with atypical or subcutaneous growth. High-resolution magnetic resonance imaging (HR-MRI) is a novel skin imaging method.

MATERIALS AND METHODS

Prospective collection of 19 patients with clinically suspected cSCC. All patients underwent high-resolution DCE-MRI using a 70-mm microscopy coil before operation. The imaging features and results of surgical pathology were recorded. K^trans , K_ep , V_e values, and the time-signal curve (TIC) types were determined using DCE images.

RESULTS

16 cases of cSCC, 3 cases of acanthoma. The subcutaneous invasion of all lesions was clearly displayed, of which 8 lesions invaded the subcutaneous fat layer, 5 invaded the muscle layer, 1 invaded the periosteum, 2 invaded the cap fascia, and the layer of all lesions invasion judged by HR-MR imaging was consistent with the postoperative pathology. The main manifestations of cSCC were ill-defined margin, obvious inhomogeneous enhancement, higher perfusion parameters value and type-III TIC, while acanthoma showed well-defined and type-I TIC. Some imaging findings (such as boundary, enhancement) and DCE perfusion parameters of the two groups overlap.

CONCLUSION

High-resolution DCE-MRI can fully and directly display the subcutaneous invasion of cSCC, and more work needs to be done to prove its value. Next, we will expand the sample size, and further explore its value in the differential diagnosis and prognosis evaluation of cSCC from acanthoma or other skin tumors.

Diffusion Kurtosis Imaging and Intravoxel Incoherent Motion in Differentiating Nasal Malignancies

OBJECTIVES/HYPOTHESIS

To evaluate the usefulness of diffusion kurtosis imaging (DKI) and intravoxel incoherent motion (IVIM) in the differentiation of sinonasal malignant tumors (SNMTs) with different histological types.

STUDY DESIGN

Retrospective observational and diagnostic study.

METHODS

Sixty-five patients with SNMTs who underwent DKI and IVIM were enrolled in this retrospective study, including 27 squamous cell carcinomas (SCCs), 13 olfactory neuroblastomas (ONBs), 14 malignant melanomas (MMs) and 11 lymphomas. The kurtosis (K) and diffusion coefficient (Dk) from DKI and the pure diffusion coefficient (D), pseudodiffusion coefficient (D*), perfusion fraction (f), and the product of D* and f (f∙D*) from IVIM were measured. Kruskal-Wallis and Dunn multiple comparison tests with Bonferroni correction, receiver operating characteristic curve, and logistic regression analyses were used for statistical analysis.

RESULTS

Lymphomas demonstrated the highest K values but lowest Dk, D, D*, f, and f∙D* values among these four malignant tumors. ONBs exhibited high K values and MMs had highest D*, f, and f∙D* values. The cutoff value of ≤0.887 × 10^-3 mm² /sec for f∙D* provided a sensitivity, specificity, and an accuracy of 100%, 98.1%, and 98.5%, respectively, for differentiating lymphomas from the other three entities. The combination of f∙D* and D values showed a sensitivity of 92.9% and a specificity of 92.5% for the discrimination of MMs from ONBs and SCCs. The K value was useful for differentiating ONBs from SCCs, with a threshold value of 0.942 (sensitivity, 84.6%; specificity, 63.0%).

CONCLUSIONS

The combined use of DKI and IVIM is helpful for differentiating among four histological types of SNMTs.

LEVEL OF EVIDENCE

3 Laryngoscope, 2019.

Whole-tumor histogram analysis of monoexponential and advanced diffusion-weighted imaging for sinonasal malignant tumors: Correlations with histopathologic features

BACKGROUND

The histopathological basis of monoexponential diffusion-weighted imaging (DWI), intravoxel incoherent motion (IVIM), and diffusion kurtosis imaging (DKI) in the characterization of sinonasal malignant tumors is still unclear.

PURPOSE

To explore the correlations of histogram metrics from monoexponential DWI, IVIM, and DKI with histopathologic features in sinonasal malignant tumors.

STUDY TYPE

Retrospective.

SUBJECTS

In all, 76 patients with sinonasal malignant tumors.

FIELD STRENGTH/SEQUENCE

Fourteen different b values (b = 0, 50, 100, 150, 200, 250, 300, 350, 400, 800, 1000, 1500, 2000, and 2500 sec/mm² ) were used to perform different DWI models at 3.0T.

ASSESSMENT

The whole-tumor histogram metrics were calculated on the apparent diffusion coefficient (ADC), pure diffusion coefficient (D), pseudodiffusion coefficient (D*), perfusion fraction (f), diffusion kurtosis (K), and diffusion coefficient (Dk) maps. Histopathologic features, including nuclear, cytoplasmic, cellular, stromal fractions, and the nuclear-to-cytoplasmic (N/C) ratio, were measured.

STATISTICAL TESTS

Spearman correlations and stepwise multiple linear regression analyses were performed to determine the correlations between histogram metrics and histopathologic features.

RESULTS

ADC, Dk, and f histogram metrics showed significant correlations with investigated histopathologic features; D and K histogram metrics were significantly correlated with cellular, stromal, and nuclear fractions (all P < 0.05). Significant correlations between the 75th percentile of D and cytoplasmic fraction and between the kurtosis of K and the N/C ratio were observed (P < 0.05). The skewness of Dk, K, and the 75th percentile of D were independently associated with cellular and nuclear fractions; the skewness of Dk and K were independently associated with stromal fraction (P < 0.05).

DATA CONCLUSION

Monoexponential and advanced DWI histogram parameters were significantly correlated with histopathologic features in sinonasal malignancies.

LEVEL OF EVIDENCE

3 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2020;51:273-285.

Intravoxel incoherent motion MRI for differentiating sinonasal small round cell malignant tumours (SRCMTs) from Non-SRCMTs: comparison and correlation with dynamic contrast-enhanced MRI

AIM

To investigate the value of intravoxel incoherent motion (IVIM) in the differentiation of sinonasal small round cell malignant tumours (SRCMTs) from non-SRCMTs and to compare and correlate these results with those of dynamic contrast-enhanced (DCE) magnetic resonance imaging (MRI).

MATERIALS AND METHODS

Ninety patients with histologically confirmed sinonasal malignant tumours (53 SRCMTs and 37 non-SRCMTs) who underwent conventional MRI, IVIM, and DCE-MRI before treatment were enrolled. The IVIM and DCE-MRI parameters were measured. Statistical analyses were performed using Student's t-tests, receiver operating characteristic (ROC) curve analyses, and Spearman's correlation coefficients.

RESULTS

A lower pure diffusion coefficient (D) value and a higher pseudo-diffusion coefficient (D*) value were found in the sinonasal SRCMTs than in the non-SRCMTs (p<0.001 and p=0.011, respectively). Moreover, the mean extravascular extracellular space volume ratio (V_e) of the SRCMTs was significantly lower than that of the non-SRCMTs (p=0.020). ROC curve analysis showed that the diagnostic performance of D outperformed those of the other perfusion and diffusion parameters. A cut-off D value of 0.56 ×10^-3 mm²/s yielded a sensitivity of 80.4%, a specificity of 75%, and an accuracy of 78.2%, with an AUC of 0.825. Significant but poor-to-fair correlations were found between the parameters from IVIM and DCE-MRI.

CONCLUSIONS

The D and D* values of IVIM and the V_e value of DCE-MRI are helpful in distinguishing sinonasal SRCMTs from non-SRCMTs, with the D values having the best diagnostic efficiency.