Added value of susceptibility-weighted imaging to diffusion-weighted imaging in the characterization of parotid gland tumors

Current status of diffusion-weighted imaging in differentiating parotid tumors

Recently, diffusion-weighted imaging (DWI) is an essential magnetic resonance imaging (MRI) protocol for head and neck imaging in clinical practice as it plays an important role in lesion detection, tumor extension evaluation, differential diagnosis, therapeutic effect prediction, therapy evaluation, and recurrence diagnosis. Especially in the parotid gland, several studies have already attempted to achieve accurate differentiation between benign and malignant tumors using DWI. A conventional single-shot echo-planar-based DWI is widely used for head and neck imaging, whereas advanced DWI sequences, such as intravoxel incoherent motion, diffusion kurtosis imaging, periodically rotated overlapping parallel lines with enhanced reconstruction, and readout-segmented echo-planar imaging (readout segmentation of long variable echo-trains), have been used to characterize parotid tumors. The mean apparent diffusion coefficient values are easily measured and useful for assessing cellularity and histological characteristics, whereas advanced image analyses, such as histogram analysis, texture analysis, and machine and deep learning, have been rapidly developed. Furthermore, a combination of DWI and other MRI protocols has reportedly improved the diagnostic accuracy of parotid tumors. This review article summarizes the current state of DWI in differentiating parotid tumors.

Differential diagnosis of parotid gland tumours: Application of SWI combined with DWI and DCE-MRI

BACKGROUND

Parotid tumours (PTs) have a variety of pathological types, and the surgical procedures differ depending on the tumour type. However, accurate diagnosis of PTs from the current preoperative examinations is unsatisfactory.

METHODS

This retrospective study was approved by the Ethics Committee of our hospital, and the requirement for informed consent was waived. A total of 73 patients with PTs, including 55 benign and 18 malignant tumours confirmed by surgical pathology, were enrolled. All patients underwent diffusion-weighted imaging (DWI), dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI), susceptibility-weighted imaging (SWI), T2-weighted imaging (T2WI), and T1-weighted imaging (T1WI). The signal uniformity and capsule on T2WI, apparent diffusion coefficient (ADC) derived from DWI, semi-quantitative parameter time-intensity curve (TIC) pattern, and quantitative parameters including transfer constant (K_trans), extravascular extracellular volume fraction (V_e), wash-out constant (K_ep) calculated from DCE-MRI, and intratumoural susceptibility signal (ITSS) obtained from SWI were assessed and compared between benign and malignant PTs. Logistic regression analysis was used to select the predictive parameters for the classification of benign and malignant parotid gland tumours, and receiver operating characteristic (ROC) curve analysis was used to evaluate their diagnostic performance.

RESULTS

Malignant PTs tended to exhibit a type C TIC pattern, whereas benign tumours tended to be type A and B (p < 0.001). Benign PTs had less ITSS than malignant tumours (p < 0.001). Multivariate analyses showed that ADC, V_e, and ITSS were predictors of tumour classification. ROC analysis showed that the area under the curve (AUC) of ADC, V_e, ITSS, and ADC combined with V_e were 0.623, 0.615, 0.826, and 0.782, respectively, in differentiating between malignant and benign PTs. When ITSS was added, the AUCs of ADC, V_e, and ADC combined with V_e increased to 0.882, 0.848, and 0.930, respectively.

CONCLUSION

SWI offers incremental diagnostic value to DWI and DCE-MRI in the characterisation of parotid gland tumours.

TransMed: Transformers Advance Multi-Modal Medical Image Classification

Over the past decade, convolutional neural networks (CNN) have shown very competitive performance in medical image analysis tasks, such as disease classification, tumor segmentation, and lesion detection. CNN has great advantages in extracting local features of images. However, due to the locality of convolution operation, it cannot deal with long-range relationships well. Recently, transformers have been applied to computer vision and achieved remarkable success in large-scale datasets. Compared with natural images, multi-modal medical images have explicit and important long-range dependencies, and effective multi-modal fusion strategies can greatly improve the performance of deep models. This prompts us to study transformer-based structures and apply them to multi-modal medical images. Existing transformer-based network architectures require large-scale datasets to achieve better performance. However, medical imaging datasets are relatively small, which makes it difficult to apply pure transformers to medical image analysis. Therefore, we propose TransMed for multi-modal medical image classification. TransMed combines the advantages of CNN and transformer to efficiently extract low-level features of images and establish long-range dependencies between modalities. We evaluated our model on two datasets, parotid gland tumors classification and knee injury classification. Combining our contributions, we achieve an improvement of 10.1% and 1.9% in average accuracy, respectively, outperforming other state-of-the-art CNN-based models. The results of the proposed method are promising and have tremendous potential to be applied to a large number of medical image analysis tasks. To our best knowledge, this is the first work to apply transformers to multi-modal medical image classification.

TransMed: Transformers Advance Multi-Modal Medical Image Classification

Over the past decade, convolutional neural networks (CNN) have shown very competitive performance in medical image analysis tasks, such as disease classification, tumor segmentation, and lesion detection. CNN has great advantages in extracting local features of images. However, due to the locality of convolution operation, it cannot deal with long-range relationships well. Recently, transformers have been applied to computer vision and achieved remarkable success in large-scale datasets. Compared with natural images, multi-modal medical images have explicit and important long-range dependencies, and effective multi-modal fusion strategies can greatly improve the performance of deep models. This prompts us to study transformer-based structures and apply them to multi-modal medical images. Existing transformer-based network architectures require large-scale datasets to achieve better performance. However, medical imaging datasets are relatively small, which makes it difficult to apply pure transformers to medical image analysis. Therefore, we propose TransMed for multi-modal medical image classification. TransMed combines the advantages of CNN and transformer to efficiently extract low-level features of images and establish long-range dependencies between modalities. We evaluated our model on two datasets, parotid gland tumors classification and knee injury classification. Combining our contributions, we achieve an improvement of 10.1% and 1.9% in average accuracy, respectively, outperforming other state-of-the-art CNN-based models. The results of the proposed method are promising and have tremendous potential to be applied to a large number of medical image analysis tasks. To our best knowledge, this is the first work to apply transformers to multi-modal medical image classification.

Quantitative Analysis and Pathological Basis of Signal Intensity on T2-Weighted MR Images in Benign and Malignant Parotid Tumors

Objective

To investigate the value of the signal intensity on T2-weighted magnetic resonance (MR) imaging using quantitative analysis in the differentiation of parotid tumors.

Materials and Methods

MR data of 80 pleomorphic adenomas (PAs), 68 Warthin tumors (WTs), and 34 malignant tumors (MTs) confirmed by surgery and histology were retrospectively analyzed. The signal intensities of tumor, normal parotid gland, spinal cord, and buccal subcutaneous fat were measured, and the signal intensity ratios (SIRs) between the tumor and the three references were calculated. Receiver operating characteristic curve was used to determine the optimal threshold and diagnostic efficiency of SIR for differentiating PAs, WTs, and MTs.

Results

The area under the curve (AUC) of tumor to parotid gland SIR (SIR_P), tumor to spinal cord SIR (SIR_C), and tumor to buccal subcutaneous fat SIR (SIR_F) for differentiating PAs and WTs was 0.922, 0.918, and 0.934, respectively. The sensitivity and specificity at an optimal SIR threshold were 86.3% and 91.2%, 80.0% and 97.1%, and 85.0% and 94.1%, respectively. The AUC of SIR_P, SIR_C, and SIR_F for distinguishing PAs from MTs was 0.793, 0.802, and 0.774, respectively. The sensitivity and specificity at an optimal SIR threshold was 86.3% and 61.8%, 80.0% and 73.5%, and 82.5% and 73.5%, respectively. The AUC of SIR_P, SIR_C, and SIR_F for distinguishing WTs from MTs was 0.716, 0.709, and 0.759, respectively. The sensitivity and specificity at an optimal SIR threshold were 61.8% and 82.4%, 55.9% and 82.4%, and 64.7% and 86.8%, respectively.

Conclusion

SIR_P, SIR_C, and SIR_F on T2-weighted MR images had high diagnostic efficiency for differentiating between PAs and WTs, while SIR_P and SIR_C for differentiating between PAs and MTs, and SIR_F for differentiating between WTs and MTs had relatively high diagnostic efficiency.

Diagnostic value of maximum signal intensity on T1-weighted MRI images for differentiating parotid gland tumours along with pathological correlation

AIM

To investigate the efficacy of the maximum signal intensity of tumour on T1-weighted magnetic resonance imaging (MRI) images for differentiating Warthin's tumours (WTs) from pleomorphic adenomas (PAs) and malignant tumours (MTs).

MATERIALS AND METHODS

One hundred and fifty-four histopathologically confirmed parotid tumours, including 76 PAs, 45 WTs, and 33 MTs, were analysed. MRI results were compared with pathological findings. The maximum signal intensity of tumour and the average signal intensity of spinal cord were measured on T1-weighted images, then the tumour-to-spinal cord signal intensity ratio (T1-max-SIR) was calculated. The distribution of T1-max-SIRs among the three groups of tumours was analysed using the Mann-Whitney U-test. Receiver operating characteristic curves were generated to assess the ability of T1-max-SIRs to differentiate parotid tumours. In addition, the interobserver agreement between readers was assessed using interclass correlation coefficient (ICC).

RESULTS

T1-max-SIRs were higher in WTs than in PAs (p<0.001) and MTs (p<0.001), and no significant difference was found between PAs and MTs (p=0.151). The area under the curve (AUC) of T1-max-SIRs for differentiating WTs from PAs was 0.901, with a sensitivity of 91.1% and a specificity of 82.9%. The AUC of T1-max-SIRs for differentiating WTs from MTs was 0.851, with a sensitivity of 88.9% and a specificity of 78.8%. Readers had excellent interobserver agreement on T1-max-SIRs (ICC = 0.989; 95% confidence interval, 0.985-0.992).

CONCLUSIONS

T1-max-SIRs can be useful for differentiating WTs from PAs and MTs with high diagnostic efficiency.

Performance of diffusion-weighted imaging for the diagnosis of parotid gland malignancies: A meta-analysis

PURPOSE

This study aimed to assess the diagnostic performance of diffusion-weighted imaging (DWI) for parotid gland malignancies.

METHODS

Four databases (PubMed, the Cochrane Library, Embase, and Web of Science) were searched systematically and retrospectively by two researchers until May 18, 2020. The methodological quality of the studies was assessed using the Quality Assessment of Diagnostic Accuracy Studies-2 tool. A bivariate random effects model was used to pool the sensitivity and specificity data for the apparent diffusion coefficient (ADC). Summary receiver operating characteristic curve was constructed, and the area under the curve (AUC) was calculated. The positive (LR+) and negative likelihood ratios (LR-) were also calculated. Subgroup and meta-regression analyses were performed to evaluate heterogeneity within studies.

RESULTS

Sixteen studies involving 1004 patients were included. The pooled sensitivity, specificity, and AUC for the ADC to distinguish malignant from begin parotid lesions were 89 %, 76 %, and 0.91, respectively. The LR + was 3.7 and LR- was 0.15, respectively. Subgroup analyses revealed that the applied cut-off b values and study size were sources of heterogeneity for the ADC. There were publication bias concerns.

CONCLUSIONS

Our meta-analysis suggests that the ADC value provides excellent sensitivity and moderate specificity for the diagnosis of malignant lesions in the parotid gland. However, substantial heterogeneity was found. Therefore, additional larger, prospective studies in combination with standard techniques focusing on parotid tumors should be conducted to determine the true performance of DWI for the differential diagnosis of parotid lesions.

Feasibility study of using simultaneous multi-slice RESOLVE diffusion weighted imaging to assess parotid gland tumors: comparison with conventional RESOLVE diffusion weighted imaging

Background

To evaluate the feasibility of using simultaneous multi-slice (SMS) readout segmentation of long variable echo-trains (RESOLVE) diffusion-weighted imaging (DWI) to assess parotid gland tumors, compared with conventional RESOLVE DWI.

Methods

From September 2018 to December 2018, 20 consecutive patients with parotid tumors who underwent MRI scan for pre-surgery evaluation were enrolled. SMS-RESOLVE DWI and conventional RESOLVE DWI were scanned with matched imaging parameters, respectively. The scan time of two DWI sequences was recorded. Qualitative (anatomical structure differentiation, lesion display, artifact, and overall image quality) and quantitative (apparent diffusion coefficient, ADC; ratio of signal-to-noise ratio, SNR ratio; ratio of contrast-to-noise ratio, CNR ratio) assessments of image quality were performed, and compared between SMS-RESOLVE DWI and conventional RESOLVE DWI by using Paired t-test. Two-sided P value less than 0.05 indicated significant difference.

Results

The scan time was 3 min and 41 s for SMS-RESOLVE DWI, and 5 min and 46 s for conventional RESOLVE DWI. SMS-RESOLVE DWI produced similar qualitative image quality with RESOLVE DWI (anatomical structure differentiation, P = 0.164; lesion display, P = 0.193; artifact, P = 0.330; overall image quality, P = 0.083). Meanwhile, there were no significant difference on ADC_Lesion (P = 0.298), ADC_Masseter (P = 0.122), SNR ratio (P = 0.584) and CNR ratio (P = 0.217) between two DWI sequences.

Conclusion

Compared with conventional RESOLVE DWI, SMS-RESOLVE DWI could provide comparable image quality using markedly reduced scan time. SMS could increase the clinical usability of RESOLVE technique for DWI of parotid gland.