Utility of the deep learning technique for the diagnosis of orbital invasion on CT in patients with a nasal or sinonasal tumor

Abstract

Background In nasal or sinonasal tumors, orbital invasion beyond periorbita by the tumor is one of the important criteria in the selection of the surgical procedure. We investigated the usefulness of the convolutional neural network (CNN)-based deep learning technique for the diagnosis of orbital invasion, using computed tomography (CT) images. Methods A total of 168 lesions with malignant nasal or sinonasal tumors were divided into a training dataset (n = 119) and a test dataset (n = 49). The final diagnosis (invasion-positive or -negative) was determined by experienced radiologists who carefully reviewed all of the CT images. In a CNN-based deep learning analysis, a slice of the square target region that included the orbital bone wall was extracted and fed into a deep-learning training session to create a diagnostic model using transfer learning with the Visual Geometry Group 16 (VGG16) model. The test dataset was subsequently tested in CNN-based diagnostic models and by two other radiologists who were not specialized in head and neck radiology. At approx. 2 months after the first reading session, two radiologists again reviewed all of the images in the test dataset, referring to the diagnoses provided by the trained CNN-based diagnostic model. Results The diagnostic accuracy was 0.92 by the CNN-based diagnostic models, whereas the diagnostic accuracies by the two radiologists at the first reading session were 0.49 and 0.45, respectively. In the second reading session by two radiologists (diagnosing with the assistance by the CNN-based diagnostic model), marked elevations of the diagnostic accuracy were observed (0.94 and 1.00, respectively). Conclusion The CNN-based deep learning technique can be a useful support tool in assessing the presence of orbital invasion on CT images, especially for non-specialized radiologists.

Advances in Image‐Based Artificial Intelligence in Otorhinolaryngology–Head and Neck Surgery: A Systematic Review

“…Some studies used AI to differentiate anterior ethmoidal artery location, 20 identify middle turbinate pneumatization (concha bullosa), 21 and recognize and calculate the volume of the inferior turbinate and maxillary sinus 53 . Most studies have combined AI with CT 20,21,35,44,46,47,53,56,69,80 . Other studies have combined AI with MRI, 33,39,40,45 whole‐slide imaging (WSI), 14,15,51 endoscopic images, 82,90 or positron emission tomography (PET)‐CT 34 …”

“…Most applications of AI in rhinology are in diagnosing nasal diseases, including nasal polyps, 14,15,51,82 inverted papilloma, 35,46,90 and other sinonasal tumors 39,69 (Supplemental File 3, available online). Some studies used AI to differentiate anterior ethmoidal artery location, 20 identify middle turbinate pneumatization (concha bullosa), 21 and recognize and calculate the volume of the inferior turbinate and maxillary sinus.…”

“…Some studies used AI to differentiate anterior ethmoidal artery location, 20 identify middle turbinate pneumatization (concha bullosa), 21 and recognize and calculate the volume of the inferior turbinate and maxillary sinus. 53 Most studies have combined AI with CT. 20,21,35,44,46,47,53,56,69,80 Other studies have combined AI with MRI, 33,39,40,45 whole-slide imaging (WSI), 14,15,51 endoscopic images, 82,90 or positron emission tomography (PET)-CT. 34 Some studies directly compared AI with specialist physicians and showed equivalence or superiority of AI over specialist physicians 14,69,80,90 ; it was found that AI required much less time to diagnose. 14 Some studies have combined 2 or more algorithms to solve this problem.…”

“…Thus, optimizing AI models using data from multiple centers is important. Nakagawa et al 69 investigated AI for the diagnosis of orbital invasion in patients with nasal or sinonasal tumors using CT images. There were 168 lesions with malignant nasal or sinonasal tumors, divided into a training data set (119) and a test data set (49).…”

Orbital and eyelid diseases: The next breakthrough in artificial intelligence?

“…Han et al (2022) automatically identified the differences in the orbital cavernous venous malformations (OCVM) from orbital CT images by training 13 ML models, including support vector machines (SVMs) and random forests. Nakagawa et al (2022) implemented a VGG-16 network to determine from CT images whether a nasal or sinus tumor invades the periorbital area. The network model achieved a diagnostic accuracy of 0.920, indicating that CNN-based DL techniques can be a useful supporting tool for assessing the presence of orbital infiltration on CT images.…”