Deep learning for the automatic detection and segmentation of parotid gland tumors on MRI.

OBJECTIVES: Parotid gland tumors (PGTs) often occur as incidental findings on magnetic resonance images (MRI) that may be overlooked. This study aimed to construct and validate a deep learning model to automatically identify parotid glands (PGs) with a PGT from normal PGs, and in those with a PGT to segment the tumor. MATERIALS AND METHODS: The nnUNet combined with a PG-specific post-processing procedure was used to develop the deep learning model trained on T1-weighed images (T1WI) in 311 patients (180 PGs with tumors and 442 normal PGs) and fat-suppressed (FS)-T2WI in 257 patients (125 PGs with tumors and 389 normal PGs), for detecting and segmenting PGTs with five-fold cross-validation. Additional validation set separated by time, comprising T1WI in 34 and FS-T2WI in 41 patients, was used to validate the model performance. RESULTS AND CONCLUSION: To identify PGs with tumors from normal PGs, using combined T1WI and FS-T2WI, the deep learning model achieved an accuracy, sensitivity and specificity of 98.2% (497/506), 100% (119/119) and 97.7% (378/387), respectively, in the cross-validation set and 98.5% (67/68), 100% (20/20) and 97.9% (47/48), respectively, in the validation set. For patients with PGTs, automatic segmentation of PGTs on T1WI and FS-T2WI achieved mean dice coefficients of 86.1% and 84.2%, respectively, in the cross-validation set, and of 85.9% and 81.0%, respectively, in the validation set. The proposed deep learning model may assist the detection and segmentation of PGTs and, by acting as a second pair of eyes, ensure that incidentally detected PGTs on MRI are not missed.

B1 Power Modification for Amide Proton Transfer Imaging in Parotid Glands: A Strategy for Image Quality Accommodation and Evaluation of Tumor Detection Feasibility.

BACKGROUND: In the application of APTw protocols for evaluating tumors and parotid glands, inhomogeneity and hyperintensity artifacts have remained an obstacle. This study aimed to improve APTw imaging quality and evaluate the feasibility of difference B1 values to detect parotid tumors. METHODS: A total of 31 patients received three APTw sequences to acquire 32 lesions and 30 parotid glands (one patient had lesions on both sides). Patients received T2WI and 3D turbo-spin-echo (TSE) APTw imaging on a 3.0 T scanner for three sequences (B1 = 2 µT, 1 µT, and 0.7 µT in APTw 1, 2, and 3, respectively). APTw image quality was evaluated using four-point Likert scales in terms of integrity and hyperintensity artifacts. Image quality was compared between the three sequences. An evaluable group and a trustable group were obtained for APTmean value comparison. RESULTS: Tumors in both APT2 and APT3 had fewer hyperintensity artifacts than in APT1. With B1 values decreasing, tumors had less integrity in APTw imaging. APTmean values of tumors were higher than parotid glands in traditional APT1 sequence though not significant, while the APTmean subtraction value was significantly different. CONCLUSIONS: Applying a lower B1 value could remove hyperintensity but could also compromise its integrity. Combing different APTw sequences might increase the feasibility of tumor detection.

Early detection of nasopharyngeal carcinoma: performance of a short contrast-free screening magnetic resonance imaging.

BACKGROUND: Although contrast-enhanced magnetic resonance imaging (MRI) detects early-stage nasopharyngeal carcinoma (NPC) not detected by endoscopic-guided biopsy (EGB), a short contrast-free screening MRI would be desirable for NPC screening programs. This study evaluated a screening MRI in a plasma Epstein-Barr virus (EBV)-DNA NPC screening program. METHODS: EBV-DNA-screen-positive patients underwent endoscopy, and endoscopy-positive patients underwent EGB. EGB was negative if the biopsy was negative or was not performed. Patients also underwent a screening MRI. Diagnostic performance was based on histologic confirmation of NPC in the initial study or during a follow-up period of at least 2 years. RESULTS: The study prospectively recruited 354 patients for MRI and endoscopy; 40/354 (11.3%) endoscopy-positive patients underwent EGB. Eighteen had NPC (5.1%), and 336 without NPC (94.9%) were followed up for a median of 44.8 months. MRI detected additional NPCs in 3/18 (16.7%) endoscopy-negative and 2/18 (11.1%) EGB-negative patients (stage I/II, n = 4; stage III, n = 1). None of the 24 EGB-negative patients who were MRI-negative had NPC. MRI missed NPC in 2/18 (11.1%), one of which was also endoscopy-negative. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of MRI, endoscopy, and EGB were 88.9%, 91.1%, 34.8%, 99.4%, and 91.0%; 77.8%, 92.3%, 35.0%, 98.7%, and 91.5%; and 66.7%, 92.3%, 31.6%, 98.1%, and 91.0%, respectively. CONCLUSION: A quick contrast-free screening MRI complements endoscopy in NPC screening programs. In EBV-screen-positive patients, MRI enables early detection of NPC that is endoscopically occult or negative on EGB and increases confidence that NPC has not been missed.

Dual-Level Augmentation Radiomics Analysis for Multisequence MRI Meningioma Grading.

BACKGROUND: Preoperative, noninvasive prediction of meningioma grade is important for therapeutic planning and decision making. In this study, we propose a dual-level augmentation strategy incorporating image-level augmentation (IA) and feature-level augmentation (FA) to tackle class imbalance and improve the predictive performance of radiomics for meningioma grading on Magnetic Resonance Imaging (MRI). METHODS: This study recruited 160 consecutive patients with pathologically proven meningioma (129 low-grade (WHO grade I) tumors; 31 high-grade (WHO grade II and III) tumors) with preoperative multisequence MRI imaging. A dual-level augmentation strategy combining IA and FA was applied and evaluated in 100 repetitions in 3-, 5-, and 10-fold cross-validation. RESULTS: The best area under the receiver operating characteristics curve of our method in 100 repetitions was ≥0.78 in all cross-validations. The corresponding cross-validation sensitivities (cross-validation specificity) were 0.72 (0.69), 0.76 (0.71), and 0.63 (0.82) in 3-, 5-, and 10-fold cross-validation, respectively. The proposed method achieved significantly better performance and distribution of results, outperforming single-level augmentation (IA or FA) or no augmentation in each cross-validation. CONCLUSIONS: The dual-level augmentation strategy using IA and FA significantly improves the performance of the radiomics model for meningioma grading on MRI, allowing better radiomics-based preoperative stratification and individualized treatment.

Radiomics Analysis in Characterization of Salivary Gland Tumors on MRI: A Systematic Review.

Radiomics analysis can potentially characterize salivary gland tumors (SGTs) on magnetic resonance imaging (MRI). The procedures for radiomics analysis were various, and no consistent performances were reported. This review evaluated the methodologies and performances of studies using radiomics analysis to characterize SGTs on MRI. We systematically reviewed studies published until July 2023, which employed radiomics analysis to characterize SGTs on MRI. In total, 14 of 98 studies were eligible. Each study examined 23-334 benign and 8-56 malignant SGTs. Least absolute shrinkage and selection operator (LASSO) was the most common feature selection method (in eight studies). Eleven studies confirmed the stability of selected features using cross-validation or bootstrap. Nine classifiers were used to build models that achieved area under the curves (AUCs) of 0.74 to 1.00 for characterizing benign and malignant SGTs and 0.80 to 0.96 for characterizing pleomorphic adenomas and Warthin's tumors. Performances were validated using cross-validation, internal, and external datasets in four, six, and two studies, respectively. No single feature consistently appeared in the final models across the studies. No standardized procedure was used for radiomics analysis in characterizing SGTs on MRIs, and various models were proposed. The need for a standard procedure for radiomics analysis is emphasized.

Recommendations for Epstein-Barr virus-based screening for nasopharyngeal cancer in high- and intermediate-risk regions.

A meeting of experts was held in November 2021 to review and discuss available data on performance of Epstein-Barr virus (EBV)-based approaches to screen for early stage nasopharyngeal carcinoma (NPC) and methods for the investigation and management of screen-positive individuals. Serum EBV antibody and plasma EBV DNA testing methods were considered. Both approaches were found to have favorable performance characteristics and to be cost-effective in high-risk populations. In addition to endoscopy, use of magnetic resonance imaging (MRI) to investigate screen-positive individuals was found to increase the sensitivity of NPC detection with minimal impact on cost-effectiveness of the screening program.

Discriminating between benign and malignant salivary gland tumors using diffusion-weighted imaging and intravoxel incoherent motion at 3 Tesla.

PURPOSE: The purpose of this study was to retrospectively evaluate the diagnostic performances of diffusion-weighted imaging (DWI) and intravoxel incoherent motion (IVIM) for discriminating between benign and malignant salivary gland tumors (SGTs). MATERIALS AND METHODS: Sixty-seven patients with 71 SGTs who underwent MRI examination at 3 Tesla were included. There were 34 men and 37 women with a mean age of 57 ± 17 (SD) years (age range: 20-90 years). SGTs included 21 malignant tumors (MTs) and 50 benign SGTs (33 pleomorphic adenomas [PAs] and 17 Warthin's tumors [WTs]). For each SGT, DWI and IVIM parameters, mean, skewness, and kurtosis of apparent diffusion coefficient (ADC), pure diffusion coefficient (D), pseudo-diffusion coefficient (D*) and perfusion volume fraction (f) were calculated and further compared between SGTs using univariable analysis. Areas under the curves (AUC) of receiver operating characteristic of significant parameters were compared using the Delong test. RESULTS: Significant differences in ADCmean, Dmean and D*mean were found between SGTs (P < 0.001). The highest AUC values were obtained for ADCmean (0.949) for identifying PAs and D*mean (0.985) for identifying WTs and skewness and kurtosis did not outperform mean. To discriminate benign from malignant SGTs with thresholds set to maximize Youden index, IVIM and DWI produced accuracies of 85.9% (61/71; 95% CI: 75.6-93.0) and 77.5% (55/71; 95% CI: 66.0-86.5) but misdiagnosed MTs as benign in 28.6% (6/21) and 61.9% (13/21) of SGTs, respectively. After maximizing specificity to 100% for benign SGTs, the accuracies of IVIM and DWI decreased to 76.1% (54/71; 95% CI: 64.5-85.4) and 64.8% (46/71; 95% CI: 52.5-75.8) but no MTs were misdiagnosed as benign. IVIM and DWI correctly diagnosed 66.0% (33/50) and 50.0% (25/50) of benign SGTs and 46.5% (33/71) and 35.2% (25/71) of all SGTs, respectively. CONCLUSION: IVIM is more accurate than DWI for discriminating between benign and malignant SGTs because of its advantage in detecting WTs. Thresholds set by maximizing specificity for benign SGTs may be advantageous in a clinical setting.

Radiomics for Discriminating Benign and Malignant Salivary Gland Tumors; Which Radiomic Feature Categories and MRI Sequences Should Be Used?

The lack of a consistent MRI radiomic signature, partly due to the multitude of initial feature analyses, limits the widespread clinical application of radiomics for the discrimination of salivary gland tumors (SGTs). This study aimed to identify the optimal radiomics feature category and MRI sequence for characterizing SGTs, which could serve as a step towards obtaining a consensus on a radiomics signature. Preliminary radiomics models were built to discriminate malignant SGTs (n = 34) from benign SGTs (n = 57) on T1-weighted (T1WI), fat-suppressed (FS)-T2WI and contrast-enhanced (CE)-T1WI images using six feature categories. The discrimination performances of these preliminary models were evaluated using 5-fold-cross-validation with 100 repetitions and the area under the receiver operating characteristic curve (AUC). The differences between models' performances were identified using one-way ANOVA. Results show that the best feature categories were logarithm for T1WI and CE-T1WI and exponential for FS-T2WI, with AUCs of 0.828, 0.754 and 0.819, respectively. These AUCs were higher than the AUCs obtained using all feature categories combined, which were 0.750, 0.707 and 0.774, respectively (p < 0.001). The highest AUC (0.846) was obtained using a combination of T1WI + logarithm and FS-T2WI + exponential features, which reduced the initial features by 94.0% (from 1015 × 3 to 91 × 2). CE-T1WI did not improve performance. Using one feature category rather than all feature categories combined reduced the number of initial features without compromising radiomic performance.

Radiomics for Discrimination between Early-Stage Nasopharyngeal Carcinoma and Benign Hyperplasia with Stable Feature Selection on MRI.

Discriminating early-stage nasopharyngeal carcinoma (NPC) from benign hyperplasia (BH) on MRI is a challenging but important task for the early detection of NPC in screening programs. Radiomics models have the potential to meet this challenge, but instability in the feature selection step may reduce their reliability. Therefore, in this study, we aim to discriminate between early-stage T1 NPC and BH on MRI using radiomics and propose a method to improve the stability of the feature selection step in the radiomics pipeline. A radiomics model was trained using data from 442 patients (221 early-stage T1 NPC and 221 with BH) scanned at 3T and tested on 213 patients (99 early-stage T1 NPC and 114 BH) scanned at 1.5T. To verify the improvement in feature selection stability, we compared our proposed ensemble technique, which uses a combination of bagging and boosting (BB-RENT), with the well-established elastic net. The proposed radiomics model achieved an area under the curve of 0.85 (95% confidence interval (CI): 0.82−0.89) and 0.80 (95% CI: 0.74−0.86) in discriminating NPC and BH in the 3T training and 1.5T testing cohort, respectively, using 17 features selected from a pool of 422 features by the proposed feature selection technique. BB-RENT showed a better feature selection stability compared to the elastic net (Jaccard index = 0.39 ± 0.14 and 0.24 ± 0.06, respectively; p < 0.001).

Automatic detection and segmentation of morphological changes of the maxillary sinus mucosa on cone-beam computed tomography images using a three-dimensional convolutional neural network.

OBJECTIVES: To propose and evaluate a convolutional neural network (CNN) algorithm for automatic detection and segmentation of mucosal thickening (MT) and mucosal retention cysts (MRCs) in the maxillary sinus on low-dose and full-dose cone-beam computed tomography (CBCT). MATERIALS AND METHODS: A total of 890 maxillary sinuses on 445 CBCT scans were analyzed. The air space, MT, and MRCs in each sinus were manually segmented. Low-dose CBCTs were divided into training, training-monitoring, and testing datasets at a 7:1:2 ratio. Full-dose CBCTs were used as a testing dataset. A three-step CNN algorithm built based on V-Net and support vector regression was trained on low-dose CBCTs and tested on the low-dose and full-dose datasets. Performance for detection of MT and MRCs using area under the curves (AUCs) and for segmentation using Dice similarity coefficient (DSC) was evaluated. RESULTS: For the detection of MT and MRCs, the algorithm achieved AUCs of 0.91 and 0.84 on low-dose scans and of 0.89 and 0.93 on full-dose scans, respectively. The median DSCs for segmenting the air space, MT, and MRCs were 0.972, 0.729, and 0.678 on low-dose scans and 0.968, 0.663, and 0.787 on full-dose scans, respectively. There were no significant differences in the algorithm performance between low-dose and full-dose CBCTs. CONCLUSIONS: The proposed CNN algorithm has the potential to accurately detect and segment MT and MRCs in maxillary sinus on CBCT scans with low-dose and full-dose protocols. CLINICAL RELEVANCE: An implementation of this artificial intelligence application in daily practice as an automated diagnostic and reporting system seems possible.

Current Applications of Deep Learning and Radiomics on CT and CBCT for Maxillofacial Diseases.

The increasing use of computed tomography (CT) and cone beam computed tomography (CBCT) in oral and maxillofacial imaging has driven the development of deep learning and radiomics applications to assist clinicians in early diagnosis, accurate prognosis prediction, and efficient treatment planning of maxillofacial diseases. This narrative review aimed to provide an up-to-date overview of the current applications of deep learning and radiomics on CT and CBCT for the diagnosis and management of maxillofacial diseases. Based on current evidence, a wide range of deep learning models on CT/CBCT images have been developed for automatic diagnosis, segmentation, and classification of jaw cysts and tumors, cervical lymph node metastasis, salivary gland diseases, temporomandibular (TMJ) disorders, maxillary sinus pathologies, mandibular fractures, and dentomaxillofacial deformities, while CT-/CBCT-derived radiomics applications mainly focused on occult lymph node metastasis in patients with oral cancer, malignant salivary gland tumors, and TMJ osteoarthritis. Most of these models showed high performance, and some of them even outperformed human experts. The models with performance on par with human experts have the potential to serve as clinically practicable tools to achieve the earliest possible diagnosis and treatment, leading to a more precise and personalized approach for the management of maxillofacial diseases. Challenges and issues, including the lack of the generalizability and explainability of deep learning models and the uncertainty in the reproducibility and stability of radiomic features, should be overcome to gain the trust of patients, providers, and healthcare organizers for daily clinical use of these models.

A convolutional neural network combined with positional and textural attention for the fully automatic delineation of primary nasopharyngeal carcinoma on non-contrast-enhanced MRI.

BACKGROUND: Convolutional neural networks (CNNs) have the potential to automatically delineate primary nasopharyngeal carcinoma (NPC) on magnetic resonance imaging (MRI), but currently, the literature lacks a module to introduce valuable pre-computed features into a CNN. In addition, most CNNs for primary NPC delineation have focused on contrast-enhanced MRI. To enable the use of CNNs in clinical applications where it would be desirable to avoid contrast agents, such as cancer screening or intra-treatment monitoring, we aim to develop a CNN algorithm with a positional-textural fully-connected attention (FCA) module that can automatically delineate primary NPCs on contrast-free MRI. METHODS: This retrospective study was performed in 404 patients with NPC who had undergone staging MRI. A proposed CNN algorithm incorporated with our positional-textural FCA module (Aproposed ) was trained on manually delineated tumours (M1st ) to automatically delineate primary NPCs on non-contrast-enhanced T2-weighted fat-suppressed (NE-T2W-FS) images. The performance of Aproposed , three well-established CNNs, Unet (Aunet ), Attention-Unet (Aatt ) and Dense-Unet (Adense ), and a second manual delineation repeated to evaluate human variability (M 2 nd ) were measured by comparing to the reference standard M 1 st to obtain the Dice similarity coefficient (DSC) and average surface distance (ASD). The Wilcoxon rank test was used to compare the performance of Aproposed against Aunet , Aatt , Adense and M 2 nd . RESULTS: Aproposed showed a median DSC of 0.79 (0.10) and ASD of 0.66 (0.84) mm. It performed better than the well-established networks Aunet [DSC =0.75 (0.12) and ASD =1.22 (1.73) mm], Aatt [DSC =0.75 (0.10) and ASD =0.96 (1.16) mm] and Adense [DSC =0.71 (0.14) and ASD =1.67 (1.92) mm] (all P<0.01), but slightly worse when compared to M 2 nd [DSC =0.81 (0.07) and ASD =0.56 (0.80) mm] (P<0.001). CONCLUSIONS: The proposed CNN algorithm has potential to accurately delineate primary NPCs on non-contrast-enhanced MRI.

Convolutional neural network in nasopharyngeal carcinoma: how good is automatic delineation for primary tumor on a non-contrast-enhanced fat-suppressed T2-weighted MRI?

PURPOSE: Convolutional neural networks (CNNs) show potential for delineating cancers on contrast-enhanced MRI (ce-MRI) but there are clinical scenarios in which administration of contrast is not desirable. We investigated performance of the CNN for delineating primary nasopharyngeal carcinoma (NPC) on non-contrast-enhanced images and compared the performance to that on ce-MRI. MATERIALS AND METHODS: We retrospectively analyzed primary NPC in 195 patients using a well-established CNN, U-Net, for tumor delineation on the non-contrast-enhanced fat-suppressed (fs)-T2W, ce-T1W and ce-fs-T1W images. The CNN-derived delineations were compared to manual delineations to obtain Dice similarity coefficient (DSC) and average surface distance (ASD). The DSC and ASD on fs-T2W were compared to those on ce-MRI. Primary tumor volumes (PTVs) of CNN-derived delineations were compared to that of manual delineations. RESULTS: The CNN for NPC delineation on fs-T2W images showed similar DSC (0.71 ± 0.09) and ASD (0.21 ± 0.48 cm) to those on ce-T1W images (0.71 ± 0.09 and 0.17 ± 0.19 cm, respectively) (p > 0.05), and lower DSC but similar ASD to ce-fs-T1W images (0.73 ± 0.09, p < 0.001; and 0.17 ± 0.20 cm, p > 0.05). The CNN overestimated PTVs on all sequences (p < 0.001). CONCLUSION: The CNN showed promise for NPC delineation on fs-T2W images in cases where it is desirable to avoid contrast agent injection. The CNN overestimated PTVs on all sequences.

Convolutional neural network for discriminating nasopharyngeal carcinoma and benign hyperplasia on MRI.

OBJECTIVES: A convolutional neural network (CNN) was adapted to automatically detect early-stage nasopharyngeal carcinoma (NPC) and discriminate it from benign hyperplasia on a non-contrast-enhanced MRI sequence for potential use in NPC screening programs. METHODS: We retrospectively analyzed 412 patients who underwent T2-weighted MRI, 203 of whom had biopsy-proven primary NPC confined to the nasopharynx (stage T1) and 209 had benign hyperplasia without NPC. Thirteen patients were sampled randomly to monitor the training process. We applied the Residual Attention Network architecture, adapted for three-dimensional MR images, and incorporated a slice-attention mechanism, to produce a CNN score of 0-1 for NPC probability. Threefold cross-validation was performed in 399 patients. CNN scores between the NPC and benign hyperplasia groups were compared using Student's t test. Receiver operating characteristic with the area under the curve (AUC) was performed to identify the optimal CNN score threshold. RESULTS: In each fold, significant differences were observed in the CNN scores between the NPC and benign hyperplasia groups (p < .01). The AUCs ranged from 0.95 to 0.97 with no significant differences between the folds (p = .35 to .92). The combined AUC from all three folds (n = 399) was 0.96, with an optimal CNN score threshold of > 0.71, producing a sensitivity, specificity, and accuracy of 92.4%, 90.6%, and 91.5%, respectively, for NPC detection. CONCLUSION: Our CNN method applied to T2-weighted MRI could discriminate between malignant and benign tissues in the nasopharynx, suggesting that it as a promising approach for the automated detection of early-stage NPC. KEY POINTS: • The convolutional neural network (CNN)-based algorithm could automatically discriminate between malignant and benign diseases using T2-weighted fat-suppressed MR images. • The CNN-based algorithm had an accuracy of 91.5% with an area under the receiver operator characteristic curve of 0.96 for discriminating early-stage T1 nasopharyngeal carcinoma from benign hyperplasia. • The CNN-based algorithm had a sensitivity of 92.4% and specificity of 90.6% for detecting early-stage nasopharyngeal carcinoma.