Clinical study on primary screening of oral cancer and precancerous lesions by oral cytology

Training high-performance deep learning classifier for diagnosis in oral cytology using diverse annotations

The uncertainty of true labels in medical images hinders diagnosis owing to the variability across professionals when applying deep learning models. We used deep learning to obtain an optimal convolutional neural network (CNN) by adequately annotating data for oral exfoliative cytology considering labels from multiple oral pathologists. Six whole-slide images were processed using QuPath for segmenting them into tiles. The images were labeled by three oral pathologists, resulting in 14,535 images with the corresponding pathologists' annotations. Data from three pathologists who provided the same diagnosis were labeled as ground truth (GT) and used for testing. We investigated six models trained using the annotations of (1) pathologist A, (2) pathologist B, (3) pathologist C, (4) GT, (5) majority voting, and (6) a probabilistic model. We divided the test by cross-validation per slide dataset and examined the classification performance of the CNN with a ResNet50 baseline. Statistical evaluation was performed repeatedly and independently using every slide 10 times as test data. For the area under the curve, three cases showed the highest values (0.861, 0.955, and 0.991) for the probabilistic model. Regarding accuracy, two cases showed the highest values (0.988 and 0.967). For the models using the pathologists and GT annotations, many slides showed very low accuracy and large variations across tests. Hence, the classifier trained with probabilistic labels provided the optimal CNN for oral exfoliative cytology considering diagnoses from multiple pathologists. These results may lead to trusted medical artificial intelligence solutions that reflect diverse diagnoses of various professionals.

Deep multiple instance learning versus conventional deep single instance learning for interpretable oral cancer detection

The current medical standard for setting an oral cancer (OC) diagnosis is histological examination of a tissue sample taken from the oral cavity. This process is time-consuming and more invasive than an alternative approach of acquiring a brush sample followed by cytological analysis. Using a microscope, skilled cytotechnologists are able to detect changes due to malignancy; however, introducing this approach into clinical routine is associated with challenges such as a lack of resources and experts. To design a trustworthy OC detection system that can assist cytotechnologists, we are interested in deep learning based methods that can reliably detect cancer, given only per-patient labels (thereby minimizing annotation bias), and also provide information regarding which cells are most relevant for the diagnosis (thereby enabling supervision and understanding). In this study, we perform a comparison of two approaches suitable for OC detection and interpretation: (i) conventional single instance learning (SIL) approach and (ii) a modern multiple instance learning (MIL) method. To facilitate systematic evaluation of the considered approaches, we, in addition to a real OC dataset with patient-level ground truth annotations, also introduce a synthetic dataset-PAP-QMNIST. This dataset shares several properties of OC data, such as image size and large and varied number of instances per bag, and may therefore act as a proxy model of a real OC dataset, while, in contrast to OC data, it offers reliable per-instance ground truth, as defined by design. PAP-QMNIST has the additional advantage of being visually interpretable for non-experts, which simplifies analysis of the behavior of methods. For both OC and PAP-QMNIST data, we evaluate performance of the methods utilizing three different neural network architectures. Our study indicates, somewhat surprisingly, that on both synthetic and real data, the performance of the SIL approach is better or equal to the performance of the MIL approach. Visual examination by cytotechnologist indicates that the methods manage to identify cells which deviate from normality, including malignant cells as well as those suspicious for dysplasia. We share the code as open source.

DNA-ICM as an adjuvant method applied on oral cytological specimens

OBJECTIVE

This study aimed to evaluate cytology diagnosis accuracy using adjuvant methods in clinical routine for oral cancer.

STUDY DESIGN

This prospective study was conducted on 98 patients with clinically potentially malignant or malignant oral cavity lesions. One oral lesion smear was taken from each patient using a cytobrush before biopsy and stored at PreservCyt Thinprep. Samples were cytologically analyzed, and DNA ploidy measurement was performed on the same slide. The diagnostic methods' accuracy was then calculated.

RESULTS

In clinical inspection, 61 patients had suspicious lesions for malignancy, whereas 37 had potentially malignant disorders. Cytology associated with DNA image cytometry presented a sensitivity of 81.2% and specificity of 90.9%. When analyzing lesions located in high-risk sites to oral malignancies individually, cytology associated with DNA image cytometry presented a sensitivity of 88.2%, specificity of 100.0%, accuracy of 90.0%, and Kappa value of 0.77 (CI 95%: 0.48-1.00).

CONCLUSIONS

Association between cytology and DNA image cytometry is an objective and non-invasive diagnostic method that demonstrated high sensitivity and specificity in diagnosing malignant epithelial squamous cell transformation in the oral cavity.

Cytological features of oral malignant lymphoma in scraping liquid-based cytology: Cases of plasmablastic lymphoma and anaplastic lymphoma kinase-positive anaplastic large cell lymphoma

The main purpose of cytological examination in the oral region is to screen for squamous cell carcinoma or intraepithelial neoplasms; thus, the background tends to be considered a deterrent for microscopy. From this perspective, liquid-based cytology (LBC) is favorable for preparing clear samples with few backgrounds. However, background hemocytes are sometimes of critical importance in the diagnosis. We report two cases of oral malignant lymphoma, plasmablastic lymphoma, and anaplastic large cell lymphoma in which careful observation of the background in scraping LBC sample contributed to the early diagnosis. Atypical lymphoid cells were observed only in a very small part of the LBC samples from the presented patients; however, cytological findings, such as large lymphoid cells with outstanding nucleoli, large mitotic cells, or intermediate-to-large lymphoid cells with pleomorphic nuclei were sufficient for obtaining a cytological diagnosis of malignant lymphoma. Although the number and cell size of leukocytes in LBC with Papanicolaou staining were significantly different from those in air-dried conventional smears with Romanovsky staining, which are commonly preferred for the discrimination of hemocytes, the corresponding cytological features could be observed. Therefore, attention should be paid to the background as well as squamous epithelium to prepare for such unexpected cases. The LBC examination with Papanicolaou staining alone can suggest the possibility of malignant lymphoma.

Micronuclei detection in oral cytologic smear: does it add diagnostic value?

BACKGROUND

Screening and early diagnosis of oral squamous cell carcinoma (OSCC) are directly associated with increased survival rate and improved prognosis. Noninvasive diagnostic tools have been implemented in the early detection as toluidine blue staining, optical imaging, and oral cytology. This study aimed to assess and compare the presence of micronuclei (MN) in oral exfoliative cytology of healthy controls, subjects exposed to high-risk factors for oral cancer, subjects with oral potentially malignant lesions (OPMLs), and those with malignant oral lesions.

SUBJECTS AND METHODS

A total number of 92 subjects were divided into 46 healthy controls with no oral mucosal lesions (23 with no evidence of cancer risk factors and 23 with cancer risk factors), 23 with OPMLs and 23 with oral malignant lesions. All the 92 participants were subjected to cytological sampling for detection of MN. The final diagnosis of the oral lesions was confirmed by the histopathological picture and compared to the cytological results.

RESULTS

The results showed that the diagnostic accuracy of MN was higher in OPMLs group (95.2%). The sensitivity of MN test in malignant group was much lower (52.2%); however, all the cytological criteria of malignancy were markedly detected as compared to the OPMLs group.

CONCLUSIONS

Conventional oral cytology supported by MN is highly beneficial as adjunctive tool in the screening for early detection of dysplastic oral lesions.

Uncommon case of cytopathological features of sialadenoma papilliferum

Sialadenoma papilliferum, a benign and rare salivary gland neoplasm, accounts for 0.4%-1.2% of all salivary gland tumors and occurs primarily in minor salivary glands of the oral cavity. Here, we report a case of sialadenoma papilliferum and its associated cytological findings. A papillary tumor was incidentally detected on the palate of an 86-year-old Japanese man. Conventional oral exfoliative cytology was performed; the cytology smear exhibited epithelial clusters composed of atypical epithelial cells with a high nuclear/cytoplasm ratio and arranged in sheet or small papillary-like projections. Cytoplasmic vacuoles were also observed in the papillae. It was difficult to make a definitive diagnosis due to the presence of uncommon cytological features. The excisional biopsy specimen revealed histological features characteristic of sialadenoma papilliferum. Mutational analysis detected BRAFV600E mutation, which confirmed the diagnosis of sialadenoma papilliferum. To the best of our knowledge, no prior cytomorphological evaluations of sialadenoma papilliferum have been reported in detail. Oral exfoliative cytology specimens from salivary gland tumors can demonstrate uncommon cytomorphological features. A differential diagnosis of sialadenoma papilliferum can be based on the observation of mildly atypical epithelial cells that form small papillary-like structures.

Utility of Indigenous Cytology Technique in the Diagnosis of Oral Cavity Lesions using Normal Saline as a Processing Fluid: A Comparative Study

Background

Oral cancer is a major health problem in developing countries. Cytology has been widely accepted as a tool in the early diagnosis of cancer.

Objectives

To evaluate the diagnostic utility of four different cytology techniques, that is, modified brush cytology (BR) technique, brush cytology cytocentrifugation (BRCC) technique, modified scrape cytology (SR) technique, and scrape cytology cytocentrifugation (SRCC) technique and correlate the cytopathological diagnosis with the available histopathological diagnosis.

Materials and Methods

It was a prospective observational study of oral cavity lesions conducted from January 2018 to December 2018 at a rural tertiary care referral institute. Smears prepared by four different techniques, that is, BR technique, BRCC technique, SR technique and SRCC technique were evaluated using a scoring system. Normal saline was used as a processing fluid for cytocentrifugation techniques, and the cytological diagnosis was compared with an available histopathological diagnosis for concordance.

Results

Twenty-seven cases of oral cavity lesions were analyzed. Squamous cell carcinoma (55.56%) constituted the most common lesion diagnosed by cytology. Total concordance was 95.65%. Brush cytology techniques were better technique than scrape cytology techniques. Cytocentrifugation techniques were better than modified brush cytology technique and modified scrape cytology technique and the values were statistically highly significant (P<0.0001).

Conclusion

The utility of only normal saline as a processing fluid for cytocentrifugation may be considered an unexplored and prudent endeavor. This indigenously designed technique may be employed to improve the quality of cytological preparation for the evaluation of oral cavity lesions.

Evaluation of oral brush liquid-based cytology for oral squamous cell carcinoma: a comparative study of cytological and histological diagnoses at a single center

PURPOSE

Liquid-based cytology is highly useful in oral cytology. However, there are only few reports on the accuracy of this method. The current study aimed to compare oral liquid-based cytological and histological diagnoses and to evaluate items that should be considered in oral cytological diagnosis for oral squamous cell carcinoma.

METHODS

We included 653 patients who underwent both oral cytological and histological examinations. Data on sex, specimen collection region, cytological and histological diagnoses, and histological images were reviewed.

RESULTS

The overall male-to-female ratio was 1:1.18. The tongue was the most common specimen collection region, followed by the gingiva and buccal mucosa. The most common cytological examination result was negative (66.8%), followed by doubtful (22.7%) and positive (10.3%). The sensitivity, specificity, positive predictive value, and negative predictive value of cytological diagnosis were 69%, 75%, 38%, and 92%, respectively. Approximately 8.3% of patients with a negative cytological diagnosis had a histological diagnosis of oral squamous cell carcinoma. Furthermore, 86.1% of histopathologic images of cytology-negative squamous cell carcinomas exhibited well-differentiated keratinocytes lacking atypia on the surface. The remaining patients developed recurrence, or they had low cell counts.

CONCLUSION

Liquid-based cytology is useful in screening oral cancer. However, a cytological diagnosis of superficial-differentiated oral squamous cell carcinoma is occasionally inconsistent with the histological diagnosis. Therefore, histological and cytological examinations should be performed if tumor-like lesions are suspected clinically.

Cytomorphometric Analysis of Oral Buccal Mucosa of Dental Colleges' Students in Sulaimani City

This study evaluates the cytomorphometric measures of cells obtained from the buccal mucosa of dental students to assess the fluctuation of the cellular characteristics among relatively normal subjects, with any potential correlations with demographic information, different habits, and hormonal disturbance. This prospective study included 100 dental students with no detectable oral alterations submitted to brush cytology. The smears were fixed with 95% ethyl alcohol and stained with hematoxylin and eosin stain. The stained section was observed under an image analyzer for cytomorphometric analysis. Cytopathological observations were recorded, including inflammation, microbial colonies, micronuclei, keratinization, overlapping, and hemorrhage. Chi-square tests were applied for non-parametric variables. One-way analysis of variance (ANOVA) was used to compare the cytometric parameters to habits and hormonal disturbances. A p-value < 0.05 was considered statistically significant. The results showed close proximity among subjects in the matter of cytomorphometric measures; no significant influence of sex, smoking, alcohol drinking habits, and menstruation was found on cytomorphometric diameters or cytopathological observation, and vice versa, while polycystic ovary syndrome impacted nuclear and nuclear-cytoplasmic ratio (p = 0.003, p = 0.02), respectively. Oral exfoliative cytology combined with cytomorphometric analysis for the studied normal individuals can be helpful in various investigations of oral and systemic diseases.

Diagnostic Accuracy of Screening of Lip and Oral Cavity Cancers or Potentially Malignant Disorders (PMD) by Frontline Workers: A Systematic Review and Meta-Analysis

BACKGROUND

Oral cancer screening strategies help reduce associated mortality and could be performed by a trained frontline health worker (FHW). The present review aims to assess the diagnostic accuracy of commonly used screening modalities for oral cancer performed by FHW in apparently healthy individuals.

METHODS

Electronic databases PubMed, Scopus, Embase, Cochrane Library, and Google Scholar, were searched. The review included studies conducted where apparently healthy adult individuals were screened by the FHW for cancer or PMD of the lip and oral cavity by any of the four commonly used techniques - Conventional Oral Examination (COE), toluidine blue staining (TBS), Oral Cytology (OC), and Chemiluminescent Illumination (CLI).

FINDINGS

A total of 2,413 potentially relevant articles were retrieved from the search, among which five studies for COE were included in the review. Four out of those five studies were done before the year 2000. None of the studies fitted the inclusion criteria for TBS, OC, and CLI. Pooled sensitivity of oral screening by COE performed by an FHW (n=5) was 88.8% (95% CI: 71.6-96.1), whereas pooled specificity was 91.9% (95% CI: 78.3-97.3). On subgroup analysis, the pooled sensitivity and specificity of studies where the prevalence of disease was <50% (n=4) was 84.5% (95% CI: 62.6 - 94.7) and 94.1% (95% CI: 82.2 - 98.2), respectively.

INTERPRETATION

COE by trained FHW had high pooled sensitivity and specificity for screening of oral cancer and PMDs. The screening techniques TBS, OC, and CLI, were not studied for mass screening by trained FHW. COE by trained FHW could be utilized for oral screening in limited-resource settings. However, the FHW should be sufficiently trained to get the desired benefits of early detection.

FUNDING

Department of Health Research, Ministry of Health & Family Welfare, Government of India.

Effective deep learning for oral exfoliative cytology classification

The use of sharpness aware minimization (SAM) as an optimizer that achieves high performance for convolutional neural networks (CNNs) is attracting attention in various fields of deep learning. We used deep learning to perform classification diagnosis in oral exfoliative cytology and to analyze performance, using SAM as an optimization algorithm to improve classification accuracy. The whole image of the oral exfoliation cytology slide was cut into tiles and labeled by an oral pathologist. CNN was VGG16, and stochastic gradient descent (SGD) and SAM were used as optimizers. Each was analyzed with and without a learning rate scheduler in 300 epochs. The performance metrics used were accuracy, precision, recall, specificity, F1 score, AUC, and statistical and effect size. All optimizers performed better with the rate scheduler. In particular, the SAM effect size had high accuracy (11.2) and AUC (11.0). SAM had the best performance of all models with a learning rate scheduler. (AUC = 0.9328) SAM tended to suppress overfitting compared to SGD. In oral exfoliation cytology classification, CNNs using SAM rate scheduler showed the highest classification performance. These results suggest that SAM can play an important role in primary screening of the oral cytological diagnostic environment.

Importance of Cytopathologic Diagnosis in Early Cancer Diagnosis in Resource-Constrained Countries

PURPOSE

The rising cancer burden in low- and middle-income countries (LMICs) stresses already weak health care systems and poses unique challenges. In resource-constrained LMICs and in circumstances where most patients must pay out of pocket for diagnostic tests, these may not be available or affordable for many. Cytopathology provides a simple, inexpensive, standardized, and low-technology diagnostic procedure that is increasingly used as an effective tool to address the hurdles faced in cancer control programs in LMICs. This review explores the potential role of cytopathology in LMICs in reducing the cancer burden.

METHODS

This review studied the existing literature across the globe regarding the utilization of cytopathology as a diagnostic or screening tool for various types of malignancies as well as its advantages and disadvantages, depending on the local situation.

RESULTS

Apart from the usefulness of cytopathology, this review also sheds light on the barriers to using cytopathology in LMICs. Most recently, SARS-CoV-2 has produced several unique challenges for cytopathology. These are being met with innovative measures to combat the effects of the pandemic and ensure the safe delivery of essential cytopathology services.

CONCLUSION

The usefulness of cytopathologic techniques has been demonstrated via various studies, even during the recent pandemic. If cytology is to be used appropriately, the focus needs to be on integrating it into the national cancer screening and diagnostic programs as well as providing well-trained human resources.