Convolutional Neural Network Performance for Sella Turcica Segmentation and Classification Using CBCT Images.

The present study aims to validate the diagnostic performance and evaluate the reliability of an artificial intelligence system based on the convolutional neural network method for the morphological classification of sella turcica in CBCT (cone-beam computed tomography) images. In this retrospective study, sella segmentation and classification models (CranioCatch, Eskisehir, Türkiye) were applied to sagittal slices of CBCT images, using PyTorch supported by U-Net and TensorFlow 1, and we implemented the GoogleNet Inception V3 algorithm. The AI models achieved successful results for sella turcica segmentation of CBCT images based on the deep learning models. The sensitivity, precision, and F-measure values were 1.0, 1.0, and 1.0, respectively, for segmentation of sella turcica in sagittal slices of CBCT images. The sensitivity, precision, accuracy, and F1-score were 1.0, 0.95, 0.98, and 0.84, respectively, for sella-turcica-flattened classification; 0.95, 0.83, 0.92, and 0.88, respectively, for sella-turcica-oval classification; 0.75, 0.94, 0.90, and 0.83, respectively, for sella-turcica-round classification. It is predicted that detecting anatomical landmarks with orthodontic importance, such as the sella point, with artificial intelligence algorithms will save time for orthodontists and facilitate diagnosis.

Hierarchical CNN-based occlusal surface morphology analysis for classifying posterior tooth type using augmented images from 3D dental surface models.

OBJECTIVE: 3D Digitization of dental model is growing in popularity for dental application. Classification of tooth type from single 3D point cloud model without assist of relative position among teeth is still a challenging task. METHODS: In this paper, 8-class posterior tooth type classification (first premolar, second premolar, first molar, second molar in maxilla and mandible respectively) was investigated by convolutional neural network (CNN)-based occlusal surface morphology analysis. 3D occlusal surface was transformed to depth image for basic CNN-based classification. Considering the logical hierarchy of tooth categories, a hierarchical classification structure was proposed to decompose 8-class classification task into two-stage cascaded classification subtasks. Image augmentations including traditional geometrical transformation and deep convolutional generative adversarial networks (DCGANs) were applied for each subnetworks and cascaded network. RESULTS: Results indicate that combing traditional and DCGAN-based augmented images to train CNN models can improve classification performance. In the paper, we achieve overall accuracy 91.35%, macro precision 91.49%, macro-recall 91.29%, and macro-F1 0.9139 for the 8-class posterior tooth type classification, which outperform other deep learning models. Meanwhile, Grad-cam results demonstrate that CNN model trained by our augmented images will focus on smaller important region for better generality. And anatomic landmarks of cusp, fossa, and groove work as important regions for cascaded classification model. CONCLUSION: The reported work has proved that using basic CNN to construct two-stage hierarchical structure can achieve the best classification performance of posterior tooth type in 3D model without assistance of relative position information. The proposed method has advantages of easy training, great ability to learn discriminative features from small image region.

3D distribution of dental plaque on occlusal surface using 2D-fluorescence-image to 3D-surface registration.

The accumulation of dental plaque on a tooth surface plays a crucial role in developing dental caries. In this paper, fluorescence imaging modality with structured light-based intraoral 3D scanner were combined to investigate the 3D distribution of dental plaque. The traditional fluorescence imaging method only reveals the 2D spatial distribution of the dental plaque on a tooth surface. To visualize the 3D distribution of the dental plaque on an occlusal surface, mapping a 2D fluorescence image to a 3D occlusal surface was investigated. An iterative closest point (ICP)-based contour registration method was proposed. A fluorescence camera was calibrated to obtain intrinsic parameters. The rotation and translation matrices for projecting the 3D occlusal surface were optimized to match the contours of the 2D fluorescence image and the 3D projected model. The 3D distribution of occlusal plaque reveals that dental plaque accumulation relates to the local and global morphology of the tooth surface. Thus, the depth of the pit-and-fissure is not the only parameter used to determine plaque content. The investigation of the 3D distribution of occlusal plaque using 2D-3D registration paves the path for the quantitative analysis of the tooth surface morphology to perform plaque-guided caries risk assessment.

Tissue characterization using optical coherence tomography and cone beam computed tomography: a comparative pilot study.

OBJECTIVES: To evaluate the ability of optical coherence tomography (OCT) in differentiating human oral tissues in comparison with cone beam computed tomography. STUDY DESIGN: In this study, we imaged four types of tissues ex vivo: human enamel, human cortical bone, human trabecular bone, and fatty tissue plus water and air by using OCT (Axsun Inc., Billerica, MA). We then developed a method for qualitative and quantitative analyses of the human specimens. The same types of tissues were also imaged using cone beam computed tomography, and gray-scale values were measured. RESULTS: The qualitative indices (intensity profile, contour plot, and histogram) for OCT images were able to provide information regarding surface characteristics as well as changes in tissue properties at different interfaces. The quantitative index (pixel intensity values) was also able to render information regarding the distribution and density of the pixels in different samples. A similar pattern was observed in the pixel intensity values and gray-scale values in both imaging modalities. CONCLUSIONS: Within the limitations of this ex vivo pilot study, OCT can reliably differentiate between a range of hard and soft tissues.

Coregistered photoacoustic and ultrasound imaging and classification of ovarian cancer: ex vivo and in vivo studies.

Most ovarian cancers are diagnosed at advanced stages due to the lack of efficacious screening techniques. Photoacoustic tomography (PAT) has a potential to image tumor angiogenesis and detect early neovascular changes of the ovary. We have developed a coregistered PAT and ultrasound (US) prototype system for real-time assessment of ovarian masses. Features extracted from PAT and US angular beams, envelopes, and images were input to a logistic classifier and a support vector machine (SVM) classifier to diagnose ovaries as benign or malignant. A total of 25 excised ovaries of 15 patients were studied and the logistic and SVM classifiers achieved sensitivities of 70.4 and 87.7%, and specificities of 95.6 and 97.9%, respectively. Furthermore, the ovaries of two patients were noninvasively imaged using the PAT/US system before surgical excision. By using five significant features and the logistic classifier, 12 out of 14 images (86% sensitivity) from a malignant ovarian mass and all 17 images (100% specificity) from a benign mass were accurately classified; the SVM correctly classified 10 out of 14 malignant images (71% sensitivity) and all 17 benign images (100% specificity). These initial results demonstrate the clinical potential of the PAT/US technique for ovarian cancer diagnosis.

Method for estimating closed-form solutions of the light diffusion equation for turbid media of any boundary shape.

This paper reports a method of estimating an approximate closed-form solution to the light diffusion equation for any type of geometry involving Dirichlet's boundary condition with known source location. It is based on estimating the optimum locations of multiple imaginary point sources to cancel the fluence at the extrapolated boundary by constrained optimization using a genetic algorithm. The mathematical derivation of the problem to approach the optimum solution for the direct-current type of diffuse optical systems is described in detail. Our method is first applied to slab geometry and compared with a truncated series solution. After that, it is applied to hemispherical geometry and compared with Monte Carlo simulation results. The method provides a fast and sufficiently accurate fluence distribution for optical reconstruction.

Design of optimal light delivery system for co-registered transvaginal ultrasound and photoacoustic imaging of ovarian tissue.

A hand-held transvaginal probe suitable for co-registered photoacoustic and ultrasound imaging of ovarian tissue was designed and evaluated. The imaging probe consists of an ultrasound transducer and four 1-mm-core multi-mode optical fibers both housed in a custom-made sheath. The probe was optimized for the highest light delivery output and best beam uniformity on tissue surface, by simulating the light fluence and power output for different design parameters. The laser fluence profiles were experimentally measured through chicken breast tissue and calibrated intralipid solution at various imaging depths. Polyethylene tubing filled with rat blood mimicking a blood vessel was successfully imaged up to ∼30 mm depth through porcine vaginal tissue at 750 nm. This imaging depth was achieved with a laser fluence on the tissue surface of 20 mJ/cm(2), which is below the maximum permissible exposure (MPE) of 25 mJ/cm(2) recommended by the American National Standards Institute (ANSI). Furthermore, the probe imaging capability was verified with ex vivo imaging of benign and malignant human ovaries. The co-registered images clearly showed different vasculature distributions on the surface of the benign cyst and the malignant ovary. These results suggest that our imaging system has the clinical potential for in vivo imaging and characterization of ovarian tissues.

Correlating optical coherence elastography based strain measurements with collagen content of the human ovarian tissue.

In this manuscript, the initial feasibility of a catheter based phase stabilized swept source optical coherence tomography (OCT) system was studied for characterization of the strain inside different human ovarian tissue groups. The ovarian tissue samples were periodically compressed with 500 Hz square wave signal along the axial direction between the surface of an unfocused transducer and a glass cover slide. The displacement and corresponding strain were calculated during loading from different locations for each tissue sample. A total of 27 ex vivo ovaries from 16 patients were investigated. Statistically significant difference (p < 0.001) was observed between the average displacement and strain of the normal and malignant tissue groups. A sensitivity of 93.2% and a specificity of 83% were achieved using 25 microstrain (µÎµ) as the threshold. The collagen content of the tissues was quantified from the Sirius Red stained histological sections. The average collagen area fraction (CAF) obtained from the tissue groups were found to have a strong negative correlation (R = -0.75, p < 0.0001) with the amount of strain inside the tissue. This indicates much softer and degenerated tissue structure for the malignant ovaries as compared to the dense, collagen rich structure of the normal ovarian tissue. The initial results indicate that the swept source OCT system can be useful for estimating the elasticity of the human ovarian tissue.

A low-cost photoacoustic microscopy system with a laser diode excitation.

Photoacoustic microscopy (PAM) is capable of mapping microvasculature networks in biological tissue and has demonstrated great potential for biomedical applications. However, the clinical application of the PAM system is limited due to the use of bulky and expensive pulsed laser sources. In this paper, a low-cost optical-resolution PAM system with a pulsed laser diode excitation has been introduced. The lateral resolution of this PAM system was estimated to be 7 µm by imaging a carbon fiber. The phantoms made of polyethylene tubes filled with blood and a mouse ear were imaged to demonstrate the feasibility of this PAM system for imaging biological tissues.

Design of miniaturized illumination for transvaginal co-registered photoacoustic and ultrasound imaging.

A novel lens-array based illumination design for a compact co-registered photoacoustic/ultrasound transvaginal probe has been demonstrated. The lens array consists of four cylindrical lenses that couple the laser beams into four 1-mm-core multi-mode optical fibers with optical coupling efficiency of ~87%. The feasibility of our lens array was investigated by simulating the lenses and laser beam profiles using Zemax. The laser fluence on the tissue surface was experimentally measured and was below the American National Standards Institute (ANSI) safety limit. Spatial distribution of hemoglobin oxygen saturation (sO2) of a mouse tumor was obtained in vivo using photoacoustic measurements at multiple wavelengths. Furthermore, benign and malignant ovaries were imaged ex vivo and evaluated histologically. The co-registered images clearly showed different patterns of blood vasculature. These results highlight the clinical potential of our system for noninvasive photoacoustic and ultrasound imaging of ovarian tissue and cancer detection and diagnosis.