Influence of Data Augmentation Strategies on the Segmentation of Oral Histological Images Using Fully Convolutional Neural Networks.

Segmentation of tumor regions in H &E-stained slides is an important task for a pathologist while diagnosing different types of cancer, including oral squamous cell carcinoma (OSCC). Histological image segmentation is often constrained by the availability of labeled training data since labeling histological images is a highly skilled, complex, and time-consuming task. Thus, data augmentation strategies become essential to train convolutional neural networks models to overcome the overfitting problem when only a few training samples are available. This paper proposes a new data augmentation strategy, named Random Composition Augmentation (RCAug), to train fully convolutional networks (FCN) to segment OSCC tumor regions in H &E-stained histological images. Given the input image and their corresponding label, a pipeline with a random composition of geometric, distortion, color transfer, and generative image transformations is executed on the fly. Experimental evaluations were performed using an FCN-based method to segment OSCC regions through a set of different data augmentation transformations. By using RCAug, we improved the FCN-based segmentation method from 0.51 to 0.81 of intersection-over-union (IOU) in a whole slide image dataset and from 0.65 to 0.69 of IOU in a tissue microarray images dataset.

Classification of Multiple H&E Images via an Ensemble Computational Scheme.

In this work, a computational scheme is proposed to identify the main combinations of handcrafted descriptors and deep-learned features capable of classifying histological images stained with hematoxylin and eosin. The handcrafted descriptors were those representatives of multiscale and multidimensional fractal techniques (fractal dimension, lacunarity and percolation) applied to quantify the histological images with the corresponding representations via explainable artificial intelligence (xAI) approaches. The deep-learned features were obtained from different convolutional neural networks (DenseNet-121, EfficientNet-b2, Inception-V3, ResNet-50 and VGG-19). The descriptors were investigated through different associations. The most relevant combinations, defined through a ranking algorithm, were analyzed via a heterogeneous ensemble of classifiers with the support vector machine, naive Bayes, random forest and K-nearest neighbors algorithms. The proposed scheme was applied to histological samples representative of breast cancer, colorectal cancer, oral dysplasia and liver tissue. The best results were accuracy rates of 94.83% to 100%, with the identification of pattern ensembles for classifying multiple histological images. The computational scheme indicated solutions exploring a reduced number of features (a maximum of 25 descriptors) and with better performance values than those observed in the literature. The presented information in this study is useful to complement and improve the development of computer-aided diagnosis focused on histological images.

Correlation of H3K9ac and H4K12ac With Cell Proliferation Marker Ki-67 in Oral Leukoplakia: An Immunohistochemical Study.

This study aimed to analyze the immunohistochemical expression of H3K9ac and H4K12ac in oral leukoplakia (OL) and its association with cell proliferation marker Ki-67 and clinicopathologic data. Paraffin-embedded, formalin-fixed tissue samples from 50 OLs and 15 fragments of the normal oral mucosa (NOM) were submitted to immunohistochemical assay using the streptavidin-biotin-peroxidase method. Quantitative analysis of the antigen-antibody reaction was performed by obtaining integrated optical density (IOD) and the percentage of positive nuclei (PPN) with ImageJ software. OL samples presented higher PPN ( P =0.02) and lower IOD values ( P =0.007) for H4K12ac in comparison to NOM. The area under the receiver operating characteristic curve for PPN and IOD values of H4K12ac immunostaining were 0.70 ( P =0.02) and 0.73 ( P =0.007), respectively. No differences were found between OL and NOM for H3K9ac. Cell proliferation marker Ki-67 had a positive correlation with PPN ( P <0.0001) and IOD ( P =0.0007) for H3K9ac expression and with IOD values ( P =0.002) for H4K12ac expression. The present findings suggest that alterations in the acetylation pattern of H4K12 occur in the early stages of oral carcinogenesis and that both H3K9ac and H4K12ac might have a role in the regulation of epithelial cell proliferation of OL.

Prognostic value of histone H3 serine 10 phosphorylation and histone H4 lysine 12 acetylation in oral squamous cell carcinoma.

AIMS: Studies on epigenetics in oral squamous cell carcinoma (OSCC) are rare. Histone modifications comprise epigenetic mechanisms that perform a key role in gene transcription and may regulate tumour development. Thus, the aim of this study was to determine whether two post-translational histone modifications, i.e. phosphorylation of serine 10 in histone H3 and acetylation of lysine 12 in histone H4, have prognostic value for OSCC patients. METHODS AND RESULTS: Paraffin-embedded tissue samples of 90 patients diagnosed with OSCC were obtained and subjected to immunohistochemical staining with antibodies against histone H3 with phosphorylation of serine 10 (H3S10ph) and histone H4 with acetylation of lysine 12 (H4K12ac). The associations of H3S10ph and H4K12ac expression levels with clinicopathological factors were determined. Five-year survival analysis and univariate and multivariate analyses were also performed. Both H3S10ph and H4K12ac were expressed in the nuclei of tumour cells. A low median of H3S10ph expression was significantly associated with cervical lymph node metastasis. Tumours with high H4K12ac expression were significantly associated with gender, alcohol consumption, and cervical lymph node metastasis. H4K12ac was also shown to have independent prognostic value in the multivariate analysis. Tumours with high H3S10ph expression, size >40 mm, an advanced stage and the presence of cervical lymph node metastases were associated with a better 5-year survival rate. Tumours with low H4K12ac expression, size >40 mm, an advanced stage and cervical lymph node metastasis were associated with a better 5-year survival rate. CONCLUSIONS: These findings suggest that H3S10ph, and mainly H4K12ac, may play a role in OSCC progression and the occurrence of cervical lymph node metastasis. Also, the expression level of H4K12ac could be an independent prognostic factor for OSCC patients.

Ameloblastic carcinoma: a Brazilian collaborative study of 17 cases.

AIMS: Ameloblastic carcinoma (AMECA) is an odontogenic malignancy that combines the histological features of ameloblastoma and cytological atypia. Because of its rarity, it poses difficulties in diagnosis. The aim of this study was to investigate the socio-demographic data, histopathology, immunohistochemical features, treatment and outcomes of 17 cases. METHODS AND RESULTS: Descriptive statistical analyses were used to portray the clinicopathological data collected, retrospectively. Log-rank tests were performed to determine new prognostic factors. Lesions were immunostained for Ki67, p16, p53, and cytokeratins (CKs), and compared with solid/multicystic ameloblastomas (n = 15). AMECA was mostly diagnosed at a late stage, affecting the posterior mandible of male patients in their fifth decade of life. Recurrence was diagnosed in nearly 90% of treated patients, and metastasis occurred in four patients. The mean number of Ki67-positive cells was 86.4 ± 66 per field. Tumours were focally positive for CK7, CK8, CK14, and CK18, and diffusely positive for CK19, p53, and p16. AMECA showed increased immunoexpression of CK18, CK19, p16, p53 and Ki67 as compared with benign cases. CONCLUSIONS: Our study has contributed to the improved characterization of the epidemiology, prognostic markers, treatment options and outcomes of AMECA. Current criteria must be reviewed to simplify the diagnostic process for these neoplasms.

A novel highly reactive Fab antibody for breast cancer tissue diagnostics and staging also discriminates a subset of good prognostic triple-negative breast cancers.

The discovery of novel markers for breast cancer (BC) has been recently relied on antibody combinatorial libraries and selection through phage display. We constructed a recombinant Fab library, and after selections against BC tissues, the FabC4 clone was thoroughly investigated by immunohistochemistry in 232 patients with long-term follow-up. The FabC4 ligand was determined by mass spectrometry. The FabC4 expression was associated with younger age, lack of progesterone receptor, higher histological grades and non-luminal subtypes, and it also identified a subset of good prognostic triple-negative BCs, possibly targeting a conformational epitope of Cytokeratin-10 (CK10). This new CK10-epitope specific antibody may open new possibilities in diagnostic and therapeutic strategies.