IntroductionFully convoluted neural networks (FCNN) applied to video-analysis are of particular interest in the field of head and neck oncology, given that endoscopic examination is a crucial step in diagnosis, staging, and follow-up of patients affected by upper aero-digestive tract cancers. The aim of this study was to test FCNN-based methods for semantic segmentation of squamous cell carcinoma (SCC) of the oral cavity (OC) and oropharynx (OP).Materials and MethodsTwo datasets were retrieved from the institutional registry of a tertiary academic hospital analyzing 34 and 45 NBI endoscopic videos of OC and OP lesions, respectively. The dataset referring to the OC was composed of 110 frames, while 116 frames composed the OP dataset. Three FCNNs (U-Net, U-Net 3, and ResNet) were investigated to segment the neoplastic images. FCNNs performance was evaluated for each tested network and compared to the gold standard, represented by the manual annotation performed by expert clinicians.ResultsFor FCNN-based segmentation of the OC dataset, the best results in terms of Dice Similarity Coefficient (Dsc) were achieved by ResNet with 5(×2) blocks and 16 filters, with a median value of 0.6559. In FCNN-based segmentation for the OP dataset, the best results in terms of Dsc were achieved by ResNet with 4(×2) blocks and 16 filters, with a median value of 0.7603. All tested FCNNs presented very high values of variance, leading to very low values of minima for all metrics evaluated.ConclusionsFCNNs have promising potential in the analysis and segmentation of OC and OP video-endoscopic images. All tested FCNN architectures demonstrated satisfying outcomes in terms of diagnostic accuracy. The inference time of the processing networks were particularly short, ranging between 14 and 115 ms, thus showing the possibility for real-time application.

Narrow-band imaging (NBI) laryngoscopy is an optical-biopsy technique used for screening and diagnosing cancer of the laryngeal tract, reducing the biopsy risks but at the cost of some drawbacks, such as large amount of data to review to make the diagnosis. The purpose of this paper is to develop a deep-learning-based strategy for the automatic selection of informative laryngoscopic-video frames, reducing the amount of data to process for diagnosis.

Purpose Fast and accurate graft hepaticsteatosis (HS) assessment is of primary importance for lowering liver-dysfunction risks after transplantation. Histopathological analysis of biopsied liver is the goldstandard for assessing HS, despite being invasive and time consuming. Due to the short time availability between liver procurement and transplantation, surgeons perform HS assessment through clinical evaluation (medical history, blood tests) and liver-texture visual analysis. Despite visual analysis being recognized as challenging in the clinical literature, few efforts have been invested to develop computer-assisted solutions for HS assessment. The objective of this paper is to investigate the automatic analysis of liver texture with machinelearning algorithms to automate the HS assessment process and offer support for the surgeon decision process. Methods Forty RGB images of forty different donors were analyzed. The images were captured with an RGB smartphone camera in the operating room (OR). Twenty images refer to livers that were accepted and twenty to discarded livers. Fifteen randomly-selected liver patches were extracted from each image. Patch size was 100×100. This way, a balanced dataset of 600 patches was obtained. Intensity-based features (IN T ), histogram of local binary pattern (H LBP riu2 ), and graylevel co-occurrence matrix (F GLCM ) were investigated. Blood-sample features (Blo) were included in the analysis, too. Supervised and semi-supervised learning approaches were investigated for feature classification. The leave-one-patient-out cross-validation was performed to estimate the classification performance. Results With the best performing feature set (H LBP riu2 +IN T +Blo) and semi-supervised learning, the achieved classification sensitivity, specificity and accuracy were 95%, 81% and 88%, respectively. Conclusions This research represents the first attempt to use machine-learning and automatic texture analysis of RGB images from ubiquitous smartphone cameras for the task of graft HS assessment. The results suggest that is a promising strategy to develop a fully automatic solution to assist surgeons in HS assessment inside the OR.