Convolutional neural network for automatic maxillary sinus segmentation on cone-beam computed tomographic images

Abstract

An accurate three-dimensional (3D) segmentation of the maxillary sinus is crucial for multiple diagnostic and treatment applications. Yet, it is challenging and time-consuming when manually performed on a cone-beam computed tomography (CBCT) dataset. Recently, convolutional neural networks (CNNs) have proven to provide excellent performance in the field of 3D image analysis. Hence, this study developed and validated a novel automated CNN-based methodology for the segmentation of maxillary sinus using CBCT images. A dataset of 264 sinuses were acquired from 2 CBCT devices and randomly divided into 3 subsets: training, validation, and testing. A 3D U-Net architecture CNN model was developed and compared to semi-automatic segmentation in terms of time, accuracy, and consistency. The average time was significantly reduced (p-value < 2.2e−16) by automatic segmentation (0.4 min) compared to semi-automatic segmentation (60.8 min). The model accurately identified the segmented region with a dice similarity co-efficient (DSC) of 98.4%. The inter-observer reliability for minor refinement of automatic segmentation showed an excellent DSC of 99.6%. The proposed CNN model provided a time-efficient, precise, and consistent automatic segmentation which could allow an accurate generation of 3D models for diagnosis and virtual treatment planning.

Three-dimensional maxillary virtual patient creation by convolutional neural network-based segmentation on cone-beam computed tomography images

“…Segmentations in Table 3 illustrate the types of required refinements per segmented structure. According to previous validation studies’ classification [ 12 , 13 ], minor refinements have no or slight clinical relevance, and the present qualitative analysis assumes that this clinical impact depends on the number of minor refinements needed. In daily practice, the clinical relevance of such refinements might differ depending on the task at hand, such as visualization, diagnosis, treatment planning, and patient education.…”

“…The quantitative assessment revealed that the sum of mean time required for automatic MVP segmentation (1.7 min) was slightly higher compared to the sum of the previously documented timing for each structure segmentation which totaled 1.3 min (maxillofacial complex: 39.1, maxillary sinus: 24.4, all teeth: 13.7 s) [ 8 , 12 , 13 ]. This minimal difference could be attributed to some technical variabilities, such as nonuser active processes, which impact the segmentation time even if the same AI tool is run several times, making it a challenge to keep the time constant [ 16 ].…”

“…This virtual patient could be applied for digital virtual planning of several treatment procedures, not only in general dentistry but also in maxillofacial surgery; Ear, Nose, and Throat (ENT); neurosurgery; and ophthalmology. Therefore, we aimed to assess the qualitative and quantitative performance of integrated CNN models of three previously validated individual networks for the creation of a segmented maxillary virtual patient (MVP) consisting of maxillary skeletal complex, maxillary sinuses, and teeth from CBCT images [ 8 , 12 , 13 ]. We hypothesized that the three integrated CNN models would reveal a similar performance as the individuals’ ones, along with a strong interobserver agreement in terms of time-efficiency and consistency for creating a segmented MVP.”…”

AI-based MRI auto-segmentation of brain tumor in rodents, a multicenter study

“…Technically, 3D convolution followed by 3D max-pooling was adopted in its encoder path, with 3D up-sampling together with the spatial information during encoding in decoder path. This architecture has been tested in various scenarios of medical imaging, with robust performance [ 16 , 21 ]. Additionally, it is noteworthy that novel variants of U-Net have been proposed, which are believed to be methodologically superior, including UNeXt, nnU-Net, cascaded U-Net, U-NetCC, double U-Net, and recurrent residual U-Net [ 1 , 11 , 13 – 15 , 28 ].…”

The Use of CBCT in Evaluating the Health and Pathology of the Maxillary Sinus

“…Nowadays, there are many research works on artificial intelligence including machine learning that can help clinicians detect abnormal findings in the maxillary sinus from CBCT images [ 87 ]. To begin with, convolutional neural network (CNN) could be used to automatically segment the maxillary sinus into the bone, air and lesion [ 88 , 89 ], such as MRC [ 90 ] or even complete opacification [ 91 ]. CNN could also be used to classify if a maxillary sinus was healthy or with sinusitis [ 92 ], and to detect the location of the maxillary sinus floor and automatically measure the associated alveolar ridge height for implant planning [ 93 ].…”