Deep learning-based multimodal segmentation of oropharyngeal squamous cell carcinoma on CT and MRI using self-configuring nnU-Net.

PURPOSE: To evaluate deep learning-based segmentation models for oropharyngeal squamous cell carcinoma (OPSCC) using CT and MRI with nnU-Net. METHODS: This single-center retrospective study included 91 patients with OPSCC. The patients were grouped into the development (n = 56), test 1 (n = 13), and test 2 (n = 22) cohorts. In the development cohort, OPSCC was manually segmented on CT, MR, and co-registered CT-MR images, which served as the ground truth. The multimodal and multichannel input images were then trained using a self-configuring nnU-Net. For evaluation metrics, dice similarity coefficient (DSC) and mean Hausdorff distance (HD) were calculated for test cohorts. Pearson's correlation and Bland-Altman analyses were performed between ground truth and prediction volumes. Intraclass correlation coefficients (ICCs) of radiomic features were calculated for reproducibility assessment. RESULTS: All models achieved robust segmentation performances with DSC of 0.64 ± 0.33 (CT), 0.67 ± 0.27 (MR), and 0.65 ± 0.29 (CT-MR) in test cohort 1 and 0.57 ± 0.31 (CT), 0.77 ± 0.08 (MR), and 0.73 ± 0.18 (CT-MR) in test cohort 2. No significant differences were found in DSC among the models. HD of CT-MR (1.57 ± 1.06 mm) and MR models (1.36 ± 0.61 mm) were significantly lower than that of the CT model (3.48 ± 5.0 mm) (p = 0.037 and p = 0.014, respectively). The correlation coefficients between the ground truth and prediction volumes for CT, MR, and CT-MR models were 0.88, 0.93, and 0.9, respectively. MR models demonstrated excellent mean ICCs of radiomic features (0.91-0.93). CONCLUSION: The self-configuring nnU-Net demonstrated reliable and accurate segmentation of OPSCC on CT and MRI. The multimodal CT-MR model showed promising results for the simultaneous segmentation on CT and MRI. CLINICAL RELEVANCE STATEMENT: Deep learning-based automatic detection and segmentation of oropharyngeal squamous cell carcinoma on pre-treatment CT and MRI would facilitate radiologic response assessment and radiotherapy planning. KEY POINTS: • The nnU-Net framework produced a reliable and accurate segmentation of OPSCC on CT and MRI. • MR and CT-MR models showed higher DSC and lower Hausdorff distance than the CT model. • Correlation coefficients between the ground truth and predicted segmentation volumes were high in all the three models.

Comparison of the Therapeutic Efficacy and Technical Outcomes between Conventional Fixed Electrodes and Adjustable Electrodes in the Radiofrequency Ablation of Benign Thyroid Nodules.

OBJECTIVE: This study aimed to compare therapeutic efficacy and technical outcomes between adjustable electrode (AE) and conventional fixed electrode (FE) for radiofrequency ablation (RFA) of benign thyroid nodules. MATERIALS AND METHODS: Between 2013 and 2021, RFA was performed on histologically proven benign thyroid nodules. For the AE method, AE length ≥ 1 cm with higher power and < 1 cm with lower power were utilized for ablating feeding vessels and nodules, especially those near anatomical structures, respectively. The therapeutic efficacy (volume reduction rate [VRR], complication rate, and regrowth rate) and technical outcomes (total energy delivery, ablated volume/energy, RFA time, and ablated volume/time) of FE and AE were compared. Continuous parameters were compared using a two-sample t-test or Mann-Whitney U test, and categorical parameters were compared using a chi-squared test or Fisher's exact test. RESULTS: A total of 182 nodules (FE: 92 vs. AE: 90) in 173 patients (mean age ± standard deviation, 47.0 ± 14.7 years; female, 90.8% [157/173]; median follow-up, 726 days [interquartile range, 441-1075 days]) were analyzed. The therapeutic efficacy was comparable, whereas technical outcomes were more favorable for AE. Both electrodes demonstrated comparable overall median VRR (FE: 92.4% vs. AE: 84.9%, P = 0.240) without immediate major complications. Overall regrowth rates were comparable between the two groups (FE: 2.2% [2/90] vs. AE: 1.1% [1/90], P > 0.99). AE demonstrated a shorter median RFA time (FE: 811 vs. AE: 627 seconds, P = 0.009). Both delivered comparable median energy (FE: 42.8 vs. AE: 29.2 kJ, P = 0.069), but AE demonstrated higher median ablated volume/energy and median ablated volume/time (FE: 0.2 vs. AE: 0.3 cc/kJ, P < 0.001; and FE: 0.7 vs. AE: 1.0 cc/min, P < 0.001, respectively). CONCLUSION: Therapeutic efficacy between FE and AE was comparable. AE demonstrated better technical outcomes than FE in terms of RFA time, ablated volume/energy, and ablated volume/time.

Glioma grading using multiparametric MRI: head-to-head comparison among dynamic susceptibility contrast, dynamic contrast-enhancement, diffusion-weighted images, and MR spectroscopy.

PURPOSE: To assess the diagnostic accuracy of dynamic susceptibility contrast, dynamic contrast-enhancement, MR spectroscopy (MRS), and diffusion-weighted imaging for differentiating high-grade (HGGs) from low-grade gliomas (LGGs). METHODS: Seventy-two patients (16 LGGs, 56 HGGs) with pathologically confirmed gliomas were retrospectively included. From three-dimensionally segmented tumor, histogram analyses of relative cerebral blood volume (rCBV), volume transfer constant (Ktrans), and apparent diffusion coefficient (ADC) were performed. Choline-to-creatinine ratio (Cho/Cr) was calculated using MRS. Logistic regression analyses were performed to differentiate HGGs (grade ≥ 3) from LGGs (grade ≤ 2). Areas under the receiver operating characteristics curves (AUC) were plotted. Subgroup analysis was performed between IDH-wildtype glioblastomas and IDH-mutant astrocytomas. Pairwise Spearman's correlation coefficients (ρ) were computed. RESULTS: HGGs had higher 95th percentile rCBV, Ktrans and Cho/Cr (P < 0.01) than LGGs. AUC of 95th percentiles of rCBV and Ktrans were 0.79 (95% CI, 0.67-0.91) and 0.74 (95% CI, 0.59-0.88), respectively. AUC of 5th percentile of ADC was 0.63 (95% CI, 0.48-0.79), and that of Cho/Cr was 0.67 (95% CI, 0.52-0.81). IDH-wildtype glioblastomas and IDH-mutant astrocytomas showed significantly different 95th percentile rCBV (P = 0.04) and Ktrans (P < 0.01), with Ktrans showing the highest AUC (0.73, 95% CI 0.57-0.89) in IDH status prediction. Moderate correlations were observed between 95th percentile rCBV and Ktrans (ρ = 0.47), Cho/Cr (ρ = 0.40), and 5th percentile ADC (ρ = -0.36) (all P < 0.01). CONCLUSIONS: The 95th percentile rCBV may be most helpful in discriminating HGGs from LGGs. The 95th percentile Ktrans may aid predicting IDH status of diffuse gliomas.

MR-self Noise2Noise: self-supervised deep learning-based image quality improvement of submillimeter resolution 3D MR images.

OBJECTIVES: The study aimed to develop a deep neural network (DNN)-based noise reduction and image quality improvement by only using routine clinical scans and evaluate its performance in 3D high-resolution MRI. METHODS: This retrospective study included T1-weighted magnetization-prepared rapid gradient-echo (MP-RAGE) images from 185 clinical scans: 135 for DNN training, 11 for DNN validation, 20 for qualitative evaluation, and 19 for quantitative evaluation. Additionally, 18 vessel wall imaging (VWI) data were included to evaluate generalization. In each scan of the DNN training set, two noise-independent images were generated from the k-space data, resulting in an input-label pair. 2.5D U-net architecture was utilized for the DNN model. Qualitative evaluation between conventional MP-RAGE and DNN-based MP-RAGE was performed by two radiologists in image quality, fine structure delineation, and lesion conspicuity. Quantitative evaluation was performed with full sampled data as a reference by measuring quantitative error metrics and volumetry at 7 different simulated noise levels. DNN application on VWI was evaluated by two radiologists in image quality. RESULTS: Our DNN-based MP-RAGE outperformed conventional MP-RAGE in all image quality parameters (average scores = 3.7 vs. 4.9, p < 0.001). In the quantitative evaluation, DNN showed better error metrics (p < 0.001) and comparable (p > 0.09) or better (p < 0.02) volumetry results than conventional MP-RAGE. DNN application to VWI also revealed improved image quality (3.5 vs. 4.6, p < 0.001). CONCLUSION: The proposed DNN model successfully denoises 3D MR image and improves its image quality by using routine clinical scans only. KEY POINTS: • Our deep learning framework successfully improved conventional 3D high-resolution MRI in all image quality parameters, fine structure delineation, and lesion conspicuity. • Compared to conventional MRI, the proposed deep neural network-based MRI revealed better quantitative error metrics and comparable or better volumetry results. • Deep neural network application to 3D MRI whose pulse sequences and parameters were different from the training data showed improvement in image quality, revealing the potential to generalize on various clinical MRI.

Diagnosis of nasal bone fractures on plain radiographs via convolutional neural networks.

This study aimed to assess the performance of deep learning (DL) algorithms in the diagnosis of nasal bone fractures on radiographs and compare it with that of experienced radiologists. In this retrospective study, 6713 patients whose nasal radiographs were examined for suspected nasal bone fractures between January 2009 and October 2020 were assessed. Our dataset was randomly split into training (n = 4325), validation (n = 481), and internal test (n = 1250) sets; a separate external dataset (n = 102) was used. The area under the receiver operating characteristic curve (AUC), sensitivity, and specificity of the DL algorithm and the two radiologists were compared. The AUCs of the DL algorithm for the internal and external test sets were 0.85 (95% CI, 0.83-0.86) and 0.86 (95% CI, 0.78-0.93), respectively, and those of the two radiologists for the external test set were 0.80 (95% CI, 0.73-0.87) and 0.75 (95% CI, 0.68-0.82). The DL algorithm therefore significantly exceeded radiologist 2 (P = 0.021) but did not significantly differ from radiologist 1 (P = 0.142). The sensitivity and specificity of the DL algorithm were 83.1% (95% CI, 71.2-93.2%) and 83.7% (95% CI, 69.8-93.0%), respectively. Our DL algorithm performs comparably to experienced radiologists in diagnosing nasal bone fractures on radiographs.

Image Quality of High-Resolution 3-Dimensional Neck MRI Using CAIPIRINHA-VIBE and GRASP-VIBE: An Intraindividual Comparative Study.

OBJECTIVES: Acquiring high-quality magnetic resonance imaging (MRI) of the head and neck region is often challenging due to motion and susceptibility artifacts. This study aimed to compare image quality of 2 high-resolution three-dimensional (3D) MRI sequences of the neck, controlled aliasing in parallel imaging results in higher acceleration (CAIPIRINHA)-volumetric interpolated breath-hold examination (VIBE), and golden-angle radial sparse parallel imaging (GRASP)-VIBE. MATERIALS AND METHODS: One hundred seventy-three patients indicated for contrast-enhanced neck MRI examination were scanned using 3 T scanners and both CAIPIRINHA-VIBE and GRASP-VIBE with nearly isotropic 3D acquisitions (<1 mm in-plane resolution with analogous acquisition times). Patients' MRI scans were independently rated by 2 radiologists using a 5-grade Likert scale for overall image quality, artifact level, mucosal and lesion conspicuity, and fat suppression degree at separate anatomical regions. Interobserver agreement was calculated using the Cohen κ coefficient. The quality ratings of both sequences were compared using the Mann-Whitney U test. Nonuniformity and contrast-to-noise ratio values were measured in all subjects. Separate MRI scans were performed twice for each sequence in a phantom and healthy volunteer without contrast injection to calculate the signal-to-noise ratio (SNR). RESULTS: The scores of overall image quality, overall artifact level, motion artifact level, and conspicuity of the nasopharynx, oropharynx, oral cavity, hypopharynx, and larynx were all significantly higher in GRASP-VIBE than in CAIPIRINHA-VIBE (all P 's < 0.001). Moderate to substantial interobserver agreement was observed in overall image quality (GRASP-VIBE κ = 0.43; CAIPIRINHA-VIBE κ = 0.59) and motion artifact level (GRASP-VIBE κ = 0.51; CAIPIRINHA-VIBE κ = 0.65). Lesion conspicuity was significantly higher in GRASP-VIBE than in CAIPIRINHA-VIBE ( P = 0.005). The degree of fat suppression was weaker in the lower neck regions in GRASP-VIBE (3.90 ± 0.72) than in CAIPIRINHA-VIBE (4.97 ± 0.21) ( P < 0.001). The contrast-to-noise ratio at hypopharyngeal level was significantly higher in GRASP-VIBE (6.28 ± 4.77) than in CAIPIRINHA-VIBE (3.14 ± 9.95) ( P < 0.001). In the phantom study, the SNR of GRASP-VIBE was 12 times greater than that of CAIPIRINHA-VIBE. The in vivo SNR of the volunteer MRI scan was 13.6 in CAIPIRINHA-VIBE and 20.7 in GRASP-VIBE. CONCLUSIONS: Both sequences rendered excellent images for head and neck MRI scans. GRASP-VIBE provided better image quality, as well as mucosal and lesion conspicuities, with less motion artifacts, whereas CAIPIRINHA-VIBE provided better fat suppression in the lower neck regions.

Deep convolutional neural network for classification of thyroid nodules on ultrasound: Comparison of the diagnostic performance with that of radiologists.

PURPOSE: This study aimed to train and validate deep learning (DL) models for differentiating malignant from benign thyroid nodules on US images and compare their performance with that of radiologists. METHODS: Images of thyroid nodules in patients who underwent US-guided fine-needle aspiration biopsy at our institution between January 2010 and March 2020 were retrospectively reviewed. Four radiologists independently classified the images. Images of thyroid nodules were trained using three different image classification DL models (VGG16, VGG19, and ResNet). The diagnostic performances of the DL models were calculated for the internal and external datasets and compared with the diagnoses of the four radiologists. Pairwise comparisons of the AUCs between the radiologists and DL models were made using bootstrap-based tests. RESULTS: In total, 15,409 images from 7,321 patients (mean age, 60 ± 13 years; malignant nodules, 20.7%) were randomly grouped into training (n = 12,327) and validation (n = 3,082) sets. Independent internal (n = 432; 197 patients) and external (n = 168; 59 patients) test sets were also acquired. The DL models demonstrated a higher diagnostic performance than the radiologists in the internal test set (AUC, 0.83 - 0.86 vs. 0.71 - 0.76, P < 0.05), but not in the external test set. The VGG16 model demonstrated the highest diagnostic performance in internal (AUC, 0.86; sensitivity, 91.8%; specificity, 73.2%) and external (AUC: 0.83; sensitivity: 78.6%; specificity: 76.8%) test sets. However, no statistical differences were found in the AUCs among the DL models. CONCLUSIONS: The DL models demonstrated comparable diagnostic performance to radiologists in distinguishing benign from malignant thyroid nodules on US images and may play a potential role in augmenting radiologists' diagnosis of thyroid nodules.

Magnetic Resonance Imaging-Based Radiomics for the Prediction of Progression-Free Survival in Patients with Nasopharyngeal Carcinoma: A Systematic Review and Meta-Analysis.

Advanced non-metastatic nasopharyngeal carcinoma (NPC) has variable treatment outcomes. However, there are no prognostic biomarkers for identifying high-risk patients with NPC. The aim of this systematic review and meta-analysis was to comprehensively assess the prognostic value of magnetic resonance imaging (MRI)-based radiomics for untreated NPC. The PubMed-Medline and EMBASE databases were searched for relevant articles published up to 12 August 2021. The Transparent Reporting of a Multivariable Prediction Model for Individual Prognosis or Diagnosis (TRIPOD) checklist was used to determine the qualities of the selected studies. Random-effects modeling was used to calculate the pooled estimates of Harrell's concordance index (C-index) for progression-free survival (PFS). Between-study heterogeneity was evaluated using Higgins' inconsistency index (I2). Among the studies reported in the 57 articles screened, 10 with 3458 patients were eligible for qualitative and quantitative data syntheses. The mean adherence rate to the TRIPOD checklist was 68.6 ± 7.1%. The pooled estimate of the C-index was 0.762 (95% confidence interval, 0.687-0.837). Substantial between-study heterogeneity was observed (I2 = 89.2%). Overall, MRI-based radiomics shows good prognostic performance in predicting the PFS of patients with untreated NPC. However, more consistent and robust study protocols are necessary to validate the prognostic role of radiomics for NPC.

Evaluating the added benefit of CT texture analysis on conventional CT analysis to differentiate benign ovarian cysts.

PURPOSE: We aimed to evaluate the benefit of adding CT texture analysis on conventional CT features of benign adnexal cystic lesions, especially in identifying mucinous cystadenoma. METHODS: This retrospective study included patients who underwent surgical removal of benign ovarian cysts (44 mucinous cystadenomas, 32 serous cystadenomas, 16 follicular/simple cysts and 43 endometriotic cysts) at our institution between January 2015 and November 2017. The CT images were independently reviewed by an abdominal radiologist (reviewer 1) and a resident (reviewer 2). Both reviewers recorded the conventional characteristics and performed texture analysis. Based on reviewer 1's results, two decision trees for differential diagnosis were developed. Reviewer 2's results were then applied to the decision trees. The diagnostic performances of each reviewer with and without the decision trees were compared. RESULTS: Several conventional features and texture analysis parameters showed significant differences between mucinous cystadenomas and other benign adnexal cysts. The first decision tree selected septum number and thickness as significant features, whereas the second decision tree selected septum number and the mean values at spatial scaling factor (SSF) 0. Reviewer 1's performance did not change significantly with or without the use of the decision trees. Reviewer 2's interpretations were significantly less sensitive than reviewer 1's interpretations (p = 0.001). However, when aided by the first and second decision trees, Reviewer 2's interpretations were significantly more sensitive than reviewer 1's interpretations (86.4%, p < 0.001; 72.7%, p = 0.001). CONCLUSION: This study suggests the benefit of CT texture analysis on conventional images to differentiate mucinous cystadenoma from other benign adnexal cysts, particularly for less experienced radiologists.