Automatic Recognition of Laryngoscopic Images Using a Deep‐Learning Technique

Abstract

Objectives/Hypothesis: To develop a deep-learning-based computer-aided diagnosis system for distinguishing laryngeal neoplasms (benign, precancerous lesions, and cancer) and improve the clinician-based accuracy of diagnostic assessments of laryngoscopy findings. Study Design: Retrospective study. Methods: A total of 24,667 laryngoscopy images (normal, vocal nodule, polyps, leukoplakia and malignancy) were collected to develop and test a convolutional neural network (CNN)-based classifier. A comparison between the proposed CNNbased classifier and the clinical visual assessments (CVAs) by 12 otolaryngologists was conducted. Results: In the independent testing dataset, an overall accuracy of 96.24% was achieved; for leukoplakia, benign, malignancy, normal, and vocal nodule, the sensitivity and specificity were 92.8% vs. 98.9%, 97% vs. 99.7%, 89% vs. 99.3%, 99.0% vs. 99.4%, and 97.2% vs. 99.1%, respectively. Furthermore, when compared with CVAs on the randomly selected test dataset, the CNN-based classifier outperformed physicians for most laryngeal conditions, with striking improvements in the ability to distinguish nodules (98% vs. 45%, P < .001), polyps (91% vs. 86%, P < .001), leukoplakia (91% vs. 65%, P < .001), and malignancy (90% vs. 54%, P < .001). Conclusions: The CNN-based classifier can provide a valuable reference for the diagnosis of laryngeal neoplasms during laryngoscopy, especially for distinguishing benign, precancerous, and cancer lesions.

Deep Learning for Automatic Segmentation of Oral and Oropharyngeal Cancer Using Narrow Band Imaging: Preliminary Experience in a Clinical Perspective

“…Recent studies confirmed the potential of FCNNs in the automatic diagnosis of benign and malignant diseases of the upper aero-digestive tract (22)(23)(24)(25)(26), demonstrating an outstanding Acc, comparable with that of experienced physicians. However, these studies were only focused on tumor detection and did not include OC and anterior OP tumors since the examination was only based on transnasal/transoral flexible video-endoscopy.…”

“…In fact, the operator usually centers the endoscopic image on the lesion to be identified, leading to a significant bias in automatic segmentation by FCNNs. The key role of illumination has also been emphasized by others (22,23), even showing different diagnostic performances in relation to the types of endoscopic device employed (23). In this view, novel advances in the field of image analysis should be supported by a parallel technical evolution of endoscopes, especially in terms of homogeneous illumination, high definition, colors, and optimization of image clarity.…”

Glucagon-like peptide-1 receptor mediates the beneficial effect of liraglutide in an acute lung injury mouse model involving the thioredoxin-interacting protein

“…ARDS can be developed in COVID-19 pneumonia subjects, in which a "cytokine storm" attacks endothelial cells in multiple organs, including lung alveolar epithelial cells. The "cytokine storm" attack will eventually lead to damaged blood gas exchange, inactivated pulmonary surfactant, resulting in the formation of hyaline membrane in the alveolar space, severe pulmonary edema, and disruption of lung parenchyma structure and functions (37,47). Since GLP-1-based drugs exert multiple beneficial effects in the lung, making them potential repurposed drug candidates for treating COVID-19 with or without T2D.…”

Künstliche Intelligenz in der Hals-Nasen-Ohren-Heilkunde

“…In einer aktuellen Studie wurden 19.000 Weißlicht-Endoskopiebilder von Stimmlippenläsionen durch ein Expertengremium aus HNO-Ärzt*innen sowie mithilfe eines Deep-Learning-Algorithmus analysiert und anschließend mit dem histopathologischen Untersuchungsergebnis (benigne Läsion vs. Leukoplakie vs. maligne Neoplasie) verglichen. Dabei schnitt das KI-System kategorieübergreifend signifikant besser ab als der Mensch (Korrektklassifikationsrate von 94 vs. 62 %) [ 25 ].…”