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.

A nomogram for distinguishing benign and malignant parotid gland tumors using clinical data and preoperative blood markers: development and validation

PURPOSE

This study aimed to construct and validate a nomogram that incorporated clinical data and preoperative blood markers to differentiate BPGTs from MPGTs more efficiently and at low cost.

METHODS

We retrospectively analyzed patients who underwent parotidectomy and histopathological diagnosis at the First Affiliated Hospital of Guangxi Medical University from January 2013 to June 2022. Subjects were randomly divided into training and validation sets with a 7:3 ratio. In the training set, the least absolute shrinkage and selection operator (LASSO) regression analysis was performed to select the most relevant features from 19 variables and built a nomogram using logistic regression. We evaluated the model's performance using receiver-operating characteristic (ROC) curves, calibration curves, clinical decision curve analysis (DCA), and clinical impact curve analysis (CICA).

RESULTS

The final sample consisted of 644 patients, of whom 108 (16.77%) had MPGTs. The nomogram included four features: current smoking status, pain/tenderness, peripheral facial paralysis, and lymphocyte-to-monocyte ratio (LMR). The optimal cut-off value for the nomogram was 0.17. The areas under the ROC curves (AUCs) of the nomogram were 0.748 (95% confidence interval [CI] = 0.689-0.807) and 0.754 (95% CI = 0.636-0.872) in the training and validation sets, respectively. The nomogram also showed good calibration, high accuracy, moderate sensitivity, and acceptable specificity in both sets. The DCA and CICA demonstrated that the nomogram had significant net benefits for a wide range of threshold probabilities (0.06-0.88 for the training set; 0.06-0.57 and 0.73-0.95 for the validation set).

CONCLUSION

The nomogram based on clinical characteristics and preoperative blood markers was a reliable tool for discriminating BPGTs from MPGTs preoperatively.

Parotid gland tumors: comparison of conventional and diffusion-weighted MRI findings with histopathological results

OBJECTIVES

The aim of this study was to investigate the relationship between pathological classification of parotid gland tumors and conventional MRI - diffusion-weighted imaging findings and also contribute the possible effect of apparent diffusion coefficient (ADC) to diagnosis.

METHODS

60 patients with parotid masses diagnosed using histopathology and/or cytology were enrolled in this retrospective study. All patients were evaluated using a 1.5 T MRI. Demographic features, conventional MRI findings, and ADC values (mean, minimum, maximum, and relative) were recorded. MRI findings and ADC values were compared between benign-malignant groups and pleomorphic adenoma vs Warthin's tumor groups.

RESULTS

60 tumors (48 benign, 12 malignant) were evaluated in a total of 60 patients (39 males, 21 females). The mean age was 59 (±14, 18-86) years old; the mean lesion size was 26 (±10, 11-61) mm. On the texture of conventional MRI, T2 dominantly hyperintense/with hypointensity signal was seen in 87% of pleomorphic adenomas and T2 dominantly hypointense/with hyperintesity signal was encountered in 64% of all Warthin's tumors. Seven (28%) Warthin's tumors were misdiagnosed as pleomorphic adenomas and two others (8%) as malignant tumors. The commonly used mean ADC value was 1.6 ± 0.6 × 10^-3 mm² s^-1 for benign tumors, 0.8 ± 0.3 × 10^-3 mm² s^-1 for malign tumors, 1 (0.9-1.8) × 10^-3 mm² s^-1 for Warthin's tumors, and 1.9 ± 0.3 × 10^-3 mm² s^-1 for pleomorphic adenomas. There was a statistically significant difference in ADC values between benign-malignant tumors and pleomorphic adenomas-Warthin's tumors.

CONCLUSIONS

Warthin's tumor may occasionally be misdiagnosed as pleomorphic adenoma and malignant tumor because of variable morphologic features. In addition to benign-malignant differentiation, the added ADC measurement may also be useful for differentiating Warthin's tumors from pleomorphic adenomas.

The diagnostic value of Superb Microvascular Imaging in identifying benign tumors of parotid gland

BACKGROUND

We compared the ultrasound features, superb microvascular imaging (SMI) and micro vessel density (MVD) of pleomorphic adenoma (PA), Warthin's tumor (WT) and basal cell adenoma (BCA) to explore the clinic value of SMI in differential diagnosis of benign tumors of parotid gland.

METHODS

The vascular distributions and grade by color doppler flow imaging (CDFI) and SMI, as well as vascular index (VI) of 249 parotid gland masses from 217 patients were analyzed.

RESULTS

The internal echogenicity of BCA are more homogeneous in comparing with WT and PA(P < 0.05). By SMI, the vascular distribution and vascular grade in PA were mainly peripheral (33.1%) and avascular (25.7%), Grade 1 (27.8%) and Grade 0 (25.7%). WT were mainly central (31.3%) and mixed distribution (34.9%), in Grade 3 (37.3%) and Grade 2 (36.2%). BCA was mainly peripheral (33.3%) and mixed distribution (33.3%), in Grade 2 (33.3%) and Grade 3 (33.3%). The overall detection rate of SMI for vascular Grade 2 and 3 was significantly higher than that of CDFI (P < 0.05). Both VI and MVD were lowest in PA, highest in WT (P < 0.001). The VI by SMI was correlated with MVD (P < 0.001). The correlation index between vascular distribution and grade by SMI and MVD were significantly higher than CDFI.

CONCLUSION

SMI can provide low-velocity blood flow information, which is helpful for the differential diagnosis of common benign tumors of parotid gland, and is expected to be more widely used.

Oncological outcomes of parotid gland malignancies: a retrospective analysis of 74 patients

INTRODUCTION

Salivary gland malignancies are rare neoplasms whose management has been evolving over the last two decades. Nevertheless, patient outcomes have not improved accordingly.

OBJECTIVE

In the present paper, factors and variables that could influence Overall, Disease-Specific and Disease-Free Survival, and Loco-Regional Control were analyzed.

METHODS

Chart data from 74 patients who underwent parotid gland surgery were retrospectively analyzed and stratified for tumor histology, grading, size, pT stage, pN stage, extracapsular spread, involved salivary gland lobe, and age at diagnosis. Major outcomes were estimated at 5 years by Kaplan-Meier curves.

RESULTS

Advanced stage, high grade, and lymph nodes involvement greatly impaired patient outcomes. Furthermore, in our cohort, the age at diagnosis ≥ 55 was a cause of poorer disease survival likely due to a different distribution in tumor histotypes between older and younger patients. Despite the two groups were homogeneous for the numerosity of squamous cell carcinomas, older patients were more rarely affected by mucoepidermoid and acinic cell carcinomas, which have generally better prognosis. Finally, patients aged ≥ 55 had a more frequent pathological involvement of the deep lobe of the parotid gland if compared to the younger counterpart.

CONCLUSION

The rarity of some salivary gland tumor histotypes requires further high-number series to fully understand the prognostic factors for both patient survival and recurrence development. In our cohort, the age at diagnosis ≥ 55 raises concerns that play crucial roles in disease survival shortening.

[Cyto-histological correlation in the management of parotid gland tumors: A retrospective study of 160 cases]

INTRODUCTION

The aim of this study was to correlate the cytological and histological results and evaluate the diagnostic performance of fine-needle aspiration cytology (FNAC) in the management of parotid gland tumors.

METHODS

This retrospective study included 160 patients with a parotid gland tumor who underwent fine-needle aspiration and parotidectomy surgery between January 2005 and August 2016 at the Limoges university hospital center.

RESULTS

On 160 fine-needle aspirations performed, fine-needle aspiration diagnoses were: 77 benign lesions, 35 malignant lesions and 48 non-diagnostic cases. Final histological diagnosis revealed there were 113 benign lesions and 47 malignant lesions. A hundred and one cytological diagnoses were accurate over 112 contributive fine-needle aspirations: seven false-negative cases and 4 false-positive cases were observed. The sensitivity, specificity, and accuracy were 82, 95 and 90% respectively for fine-needle aspiration, and 83, 95 and 92% respectively for association of fine-needle aspiration and magnetic resonance imaging. Diagnostic concordance between fine-needle aspiration and final histology for malignant lesions was 78%. A greater number of contributive fine-needle aspirations was observed among experimented operators.

DISCUSSION

Fine-needle aspiration is a reliable, safe and effective diagnostic tool that allows good differentiation between malignant and benign diagnosis in the preoperative management of parotid gland tumors. The association of fine-needle aspiration and magnetic resonance imaging (MRI) can improve diagnostic performance.