Deep learning algorithms for classification and detection of recurrent aphthous ulcerations using oral clinical photographic images.

Background/purpose: The application of artificial intelligence diagnosis based on deep learning in the medical field has been widely accepted. We aimed to evaluate convolutional neural networks (CNNs) for automated classification and detection of recurrent aphthous ulcerations (RAU), normal oral mucosa, and other common oral mucosal diseases in clinical oral photographs. Materials and methods: The study included 785 clinical oral photographs, which was divided into 251 images of RAU, 271 images of the normal oral mucosa, and 263 images of other common oral mucosal diseases. Four and three CNN models were used for the classification and detection tasks, respectively. 628 images were randomly selected as training data. In addition, 78 and 79 images were assigned as validating and testing data. Main outcome measures included precision, recall, F1, specificity, sensitivity and area under the receiver operating characteristics curve (AUC). Results: In the classification task, the Pretrained ResNet50 model had the best performance with a precision of 92.86%, a recall of 91.84%, an F1 score of 92.24%, a specificity of 96.41%, a sensitivity of 91.84% and an AUC of 98.95%. In the detection task, the Pretrained YOLOV5 model had the best performance with a precision of 98.70%, a recall of 79.51%, an F1 score of 88.07% and an AUC of Precision-Recall curve 90.89%. Conclusion: The Pretrained ResNet50 and the Pretrained YOLOV5 algorithms were shown to have superior performance and acceptable potential in the classification and detection of RAU lesions based on non-invasive oral images, which may prove useful in clinical practice.

[Preliminary establishment of an unsupervised quantification model of Ki-67 proliferation index based on local variable threshold method].

PURPOSE: To develop an effective machine learning method for estimation of Ki-67 cell proliferation index. METHODS: Oral squamous cell carcinoma(OSCC) slices were selected for Ki-67 immunohistochemical staining. The digital pathology images were obtained through whole-slide imaging technology. Variable threshold method based on local statistics was applied to preprocess the images, aiming at reducing the noise in the images. Adaptive threshold method was used to remove the irrelevant light-colored background area in the image, retaining the nucleus part. A threshold method in space was applied to differentiate brown from blue content. Finally, the proliferation index was estimated and compared with manual and the color deconvolution method by paired sample t test and spearman correlation coefficients with SPSS 24.0 software package. RESULTS: A new nucleus detection and classification method was established, which can process pathologic images of different sizes, and effectively detect immunohistochemical brown positive cells and blue negative cells. There was no significant difference between this algorithm and manual counting（P＞0.05）, but the speed was faster. The calculation efficiency advantage was more obvious when processing a large image, and the detection result of Ki-67 proliferation index was better than the commonly used color deconvolution method（P＜0.05）. CONCLUSIONS: The automatic nucleus quantitative analysis method developed in this study can analyze Ki-67 staining of the nucleus in OSCC cells efficiently and calculate the proliferation index, which can be used for auxiliary diagnosis in pathology.

Multi-Site Tumour Sampling Improves the Detection of Intra-Tumour Heterogeneity in Oral and Oropharyngeal Squamous Cell Carcinoma.

Background: Oral squamous cell carcinoma (OSCC) and oropharyngeal squamous cell carcinoma (OPSCC) are very common in head and neck malignancy. Intratumour heterogeneity (ITH) may hamper their responses to treatment. Hence, novel tumour sampling methods that reflect ITH are required. In this study, we investigated the clinical significance of multi-site tumour sampling (MSTS) to detect ITH in OSCC and OPSCC. Methods: One hundred eighty-two paired specimens were sampled by routine sampling (RS) or MSTS, respectively. Histologically, tumour grade, peri-tumoural vascular and lymphatic growth, perineural permeation, tumour necrosis, and muscle invasion were assessed. Immunohistochemically, the positive and average detection rates of P53(mutant), ki67 and CyclinD1 were detected. The exon 9 and exon 20 mutations of PIK3CA gene and the methylation status of the CDKN2A promoter were analysed. Results: Microscopically, the detection rate of perineural permeation, the detection density of peri-tumoural vascular and lymphatic growth, necrosis and muscle invasion in MSTS were significantly more frequent than those in RP (P < 0.05, P < 0.05, P < 0.01, P < 0.01). MSTS resulted in a higher detection rate of P53 (mutant), ki67, and CyclinD1 expression than did RS, but the difference was not significant. MSTS's detection rates in PIK3CA gene mutation and gene methylation sequencing in CDKN2A gene promoter region were both higher than RP (P < 0.05, P < 0.01). To be emphasised, the hotspot mutation H1047Rwas detected in one MSTS specimen (case 24M5) but in no RS specimens. Conclusions: This study verified that MSTS's advantage in the reflection of morphological and molecular characteristics of OSCC and OPSCC. MSTS was more representative than RP. Therefore, MSTS can compensate the RP limitations in ITH detection especially in large tumours.

Myb Immunohistochemical Staining and Fluorescence in situ Hybridization in Salivary Rare Basaloid Lesions.

Objective: Salivary rare basaloid lesions, including cribriform type basal cell adenoma (cBCA), BCA with incomplete capsule (iBCA), sialoblastoma (SB), and intercalated duct hyperplasia (IDH), could easily be misdiagnosed as adenoid cystic carcinoma (AdCC). We aim to identify an approach for differential diagnosis and to establish an optimal workflow concerning the diagnosis of these lesions. Material and methods: A panel of antibodies (MYB, ß-catenin, CD117, SOX10, ki67, P63, calponin) and fluorescence in situ hybridization (FISH)-MYB were utilized to distinguish above salivary basaloid diseases from AdCC. Results: Histologically, the striking diagnostic features of cBCA, iBCA, SB, and IDH are composed of basaloid tumor cells, well-defined encapsulation, or lack of destructive invasion. Immunohistochemically, Myb immune-labeling could effectively make a distinction among cBCA, iBCA, SB, and IDH from AdCC, except in SB. cBCA and iBCA typically expressed ß-catenin in the nuclei of tumor cells. There was no statistical significance in the ki67 index between SB and AdCC, but their indices were significantly higher than those of iBCA and IDH (p < 0.05, p < 0.05, respectively). P63 and calponin immune-expression were observed in the basaloid or myoepithelial cells. CD117 were observed positively in cBCA, iBCA, SB, and AdCC, except in IDH. SOX10 were observed positively in all cases. No cases had fusion of MYB and NFIB detectable by FISH, except in AdCC. Conclusion: Considering their sensitivity and specificity, FISH-Myb and an immunohistochemical panel of MYB/ß-catenin/ki67 would be an optimal choice for the differential diagnosis of these basaloid lesions. Clinical relevance: Some salivary basaloid tumor or tumor-like lesions have overlapping features with AdCC. Through this present research, we suggested that the panel IHC of MYB, ßcatenin, and ki67 combined with FISH-Myb should be an optimal choice for differential diagnosis among those lesions.

Clinicopathologic Analysis of Oral Squamous Cell Carcinoma After 125I Interstitial Brachytherapy.

Oral squamous cell carcinoma is a common malignancy in the head and neck region. Brachytherapy after radical surgery has achieved much progress as a cancer adjuvant treatment. This study focused on the pathologic characteristics of the patients with oral squamous cell carcinoma who underwent seed implantation after radical surgery, and the relationship of these characteristics with prognosis. Thus, 76 patients with oral squamous cell carcinoma, who were treated with surgery and subsequently with iodine 125 (125I) radioactive seed implantation, were recruited in this study. We summarized the demographic information, tumor size, location, clinical stage, prognosis, and pathologic characteristics, and discussed the correlations between prognosis and histologic features of oral squamous cell carcinoma after seed implantation. The data showed that the median age was 64 years old, the male/female ratio was 47/29, and the frequent location of the carcinoma was the tongue (35.5%). The median follow-up time was 126 months, and of the patients, 52 (68.4%) exhibited recurrent tumors. The 5-year survival rate was 81.5%, and the local control rate in 6 months was 95.3%. Microscopically, 25 cases demonstrated lymph node metastasis, there was obvious necrosis in 13 cases, and 55 cases exhibited confirmed adjacent tissue invasion including muscle, gland, vessel, nerve, and bone infiltration. Among those, vascular infiltration (13 cases) was significantly correlated with tumor recurrence ( P < .05). This study suggests that detailed pathologic diagnosis and microscopic description, especially of vascular infiltration, was valuable in the prognosis prediction of brachytherapy.

An Extracranial Metastasis of Glioblastoma Mimicking Mucoepidermoid Carcinoma.

BACKGROUND: Glioblastoma (GBM) is the most common and aggressive primary malignant tumor of the brain and central nervous system. Extracranial metastases of GBM are rare, with few case reports published to date. The tumor cells of GBM show strong immunopositivity for glial fibrillary acid protein. CASE DESCRIPTION: A 47-year-old man without comorbidities presented with a 1-year history of an augmenting right parotid lump. A right total parotidectomy with selective neck dissection was performed. The hematoxylin-eosin-stained slice of a parotid lymph node collected intraoperatively revealed destruction of normal lymph node structure by medium-sized pleomorphic cells scattered in groups; their cytoplasm was lightly stained and pale. There were abundant myxoid stroma in the interstitial tissue. This characteristic mimicked mucoepidermoid carcinoma. An immunohistochemistry test demonstrated that the tumor cells were positive for glial fibrillary acid protein. A diagnosis of extracranial metastasis of GBM was made after confirmation with postoperative pathologic examination and the review of the intracranial resection specimen. CONCLUSIONS: We believe that this is the first reported case of extracranial metastasis of GBM resembling mucoepidermoid carcinoma in the microscope features. Pathologists and clinicians should be alert to this rare lesion and consider this differential diagnosis after excluding other common parotid lesions.