METHOD AND APPARATUS FOR WIRELESS PORTABLE ULTRASOUND IMAGING

Abstract

Presented is a wireless portable ultrasound acquisition system for dental imaging, having an ultrasound probe with a control switch connected through a cable to a portable ultrasound acquisition system that communicates wirelessly with a smart tablet or a phone display to display the ultrasound images. The system uses ultrasound signals to create images of alveolar bone structure and boundaries of enamel, dentin and gingiva of a patient.

Claims

1. An apparatus for imaging an oral structure of upper and lower jaws at facial and lingual surfaces of a patient, the apparatus comprising: a) an ultrasound probe comprising an array of piezoelectric transducer crystals operating at an ultrasonic frequency of at least 20 megahertz; b) a probe tip configured for housing the array of crystals, the probe tip configured for rotating and bending; c) a gel pad comprising one or both of polymer and hydrogel configured to be disposed on the probe tip and positioned between the array of crystals and the oral structure; d) a battery; and e) a control switch configured for controlling the operation of the apparatus.

2. The apparatus as set forth in claim 1, further comprising a handle, the probe tip rotatably attached to the handle.

3. The apparatus as set forth in claim 1, wherein the gel pad comprises low ultrasonic attenuation at the ultrasonic frequency and is safe for use in the oral structure of the patient, the gel pad configured to cover the array, the gel pad further configured to be shaped to conform to the oral structure to be imaged.

4. The apparatus as set forth in claim 1, comprising an ultrasound data acquisition unit, the acquisition unit comprising: a) a microcontroller or digital signal processor or an application specific integrated circuit (“ASIC”) operatively coupled to the array and configured to control ultrasound signal generation, ultrasound signal acquisition, processing of acquired ultrasound signals and communication of the acquired ultrasound signals; and b) a wireless communications transceiver module operatively coupled to the microcontroller or digital signal processor or ASIC, the transceiver module configured to wirelessly transmit the acquired ultrasound signals to a peripheral smart device comprising a visual display.

5. The apparatus as set forth in claim 4, further comprising a control foot pedal configured for wireless communication with the transceiver module, the foot pedal configured to control the operation of the apparatus.

6. The apparatus as set forth in claim 4, wherein the transceiver module is configured to communicate using one or more of Bluetooth®, Wi-Fi®, Wi-Fi Direct® and ZigBee® communications protocols.

7. The apparatus as set forth in claim 4, wherein the microcontroller or digital signal processor or ASIC is configured to multiplex ultrasound signals transmitted to the array.

8. The apparatus as set forth in claim 4, wherein the microcontroller or digital signal processor or ASIC further comprises an analog-to-digital converter configured to digitize ultrasound signals received from the array.

9. The apparatus as set forth in claim 4, wherein the peripheral smart device comprises one or more of a general purpose computer, a personal digital assistant, a smart phone, a smart television and a computing tablet.

10. The apparatus as set forth in claim 9, wherein the peripheral smart device comprises an iOS® or Android® operating system.

11. The apparatus as set forth in claim 4, wherein the acquisition unit comprises a battery management circuit.

12. The apparatus as set forth in claim 4, wherein the peripheral smart device comprises a memory further comprising software code segments configured to cause the peripheral smart device to carry out one or more steps comprising of: a) enhancing ultrasound signals representing images of alveolar bone structure and boundaries of enamel, dentin and gingiva of a patient using a noise removal filter, a contrast enhancement, an edge enhancement, and machine learning; b) identifying peaks (global maximum) and troughs (global minimum) of one or more of cementoenamel junctions, alveolar bone crests and gingival sulcus of the patient using object detection and recognition; c) calculating changes in bone level or pocket depth of the patient using measurements between ultrasound images of different periods; d) comparing the ultrasound images of the patient with one or more of CBCT images of an oral structure of the patient and enhancing visualization of soft and hard tissues of the oral structure; e) eliminating artifacts caused by multiple reflections of ultrasonic waves in the ultrasonic images of the oral structure; f) calculating ultrasonic velocity for the hard tissues; and g) correcting the detected thickness of the hard tissues.

13. The apparatus as set forth in claim 12, wherein the software code segments are configured to cause the peripheral smart device to carry out the step of detecting boundary and segments of the oral structure using one or more of multi-label graph cut approach, contrast enhancement, a homomorphic filter, and machine learning.

14. The apparatus as set forth in claim 12, wherein the software code segments are configured to cause the peripheral smart device to carry out the step of extracting interest landmarks of the oral structure using a combination of region extraction, edge detection, local maximum and/or local minimum localization and one or more of adaptive median filtering, homomorphic filtering, and contrast enhancement.

15. The apparatus as set forth in claim 12, wherein the software code segments are configured to cause the peripheral smart device to carry out the step of measuring changes of the oral structure over a period of time using the measurements from ultrasound images of different periods of time.

16. The apparatus as set forth in claim 12, wherein the software code segments are configured to cause the peripheral smart device to carry out the step of fusing the ultrasound images of the oral structure with one or more of CBCT images of the oral structure using a combination of region extraction, edge detection, probability-based set registration, and one or more of adaptive median filtering, homomorphic filtering, and contrast enhancement.

17. The apparatus as set forth in claim 12, wherein the software code segments are configured to cause the peripheral smart device to carry out the step of predicting and removing the multiple reflections artifacts.

18. The apparatus as set forth in claim 12, wherein the software code segments are configured to cause the peripheral smart device to carry out the step of calculating the ultrasonic velocity of the hard tissues.

19. The apparatus as set forth in claim 12, wherein the software code segments are configured to cause the peripheral smart device to carry out the step of correcting the detected thickness of the hard tissues.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0045] FIG. 1A is a block diagram depicting one embodiment of an apparatus for wireless portable ultrasound imaging.

[0046] FIG. 1B is a block diagram depicting another embodiment of the apparatus of FIG. 1A.

[0047] FIG. 2 is a block diagram depicting another embodiment of a data acquisition system for use with the apparatus of FIG. 1A.

[0048] FIG. 3 is an illustration depicting the penetration of ultrasound into hard tissue.

[0049] FIG. 4A is a perspective view depicting one embodiment of an ultrasound probe.

[0050] FIG. 4B is a perspective view depicting another embodiment of the ultrasound probe of FIG. 4A.

[0051] FIG. 5 is a perspective view depicting one embodiment of a wireless portable ultrasound imaging apparatus.

[0052] FIG. 6 is a series of 8 images depicting image preprocessing steps for in-vitro (a) and for in-vivo data (b). (1) Original image; (2) after use of homomorphic filter; (3) after homomorphic filtering and contrast enhancement; (4); after homomorphic filtering, contrast enhancement, and use of adaptive median filter.

[0053] FIG. 7 is a block diagram depicting one embodiment of a semi-automated process to identify CEJ gingival margin and alveolar bone crest.

[0054] FIG. 8 is a series of 3 images depicting one embodiment of region extraction for: (a) Select region manually; (b) corresponding RGB image; (c) region extraction using k-mean clustering.

[0055] FIG. 9 is a series of 3 images depicting one embodiment of CEJ identification process for: (a) Edge detection; (b) local difference calculation; (c) CEJ identification at maximum difference.

[0056] FIG. 10A is a series of 2 images depicting a comparison of CEJ CAL method (red circle) and manual CEJ identification (blue circle) on (a) in-vitro and (b) in-vivo data.

[0057] FIG. 10B is a series of 2 images depicting an example of measuring pocket depth (A), alveolar bone level to the CEJ (B), or gingiva thickness at the CEJ (C) in a human ultrasound image.

[0058] FIG. 11 is a block diagram depicting a flowchart of one embodiment of an ultrasound-CBCT image registration process.

[0059] FIG. 12A is a series of 8 images depicting an example of Coherence Point Drift registration between ultrasound (“US”) and CBCT images.

[0060] FIG. 12B is an image depicting an example of point-based evaluation for the registration.

[0061] FIG. 13A is a block diagram depicting an ultrasonic transducer mounted on a matching layer.

[0062] FIG. 13B is a block diagram depicting the ultrasonic transducer of FIG. 13A where the matching layer is mounted on an aluminum plate.

[0063] FIG. 13C is an image depicting an example of multiple reflections artifacts produced by the ultrasound transducer of FIG. 13A.

[0064] FIG. 13D is an image depicting an example of multiple reflections artifacts produced by the ultrasound transducer of FIG. 13B.

[0065] FIG. 14A is an image depicting an example without multiple reflections removal from a porcine ultrasound RF signal.

[0066] FIG. 14B is an image depicting an example multiple reflections removal from a porcine ultrasound RF signal.

[0067] FIG. 15 is an image flowchart depicting an example of semi-automatic alveolar bone segmentation in ultrasound image using graph cut using one embodiment of a segmentation process: (a) Drawing a region (in yellow color) to determine the region of interest (ROI) on the original ultrasound (“US”) image, (b) after de-noising using homomorphic filtering, (c) with contrast enhancement, (d) after smoothing using Gaussian filtering, (e) multi-label graph cuts segmentation in ROI, (f) alveolar bone extraction, and (g) final result showing delineation (in orange color) of alveolar bone boundary in the original US image.

[0068] FIG. 16 is a series of 2 images depicting an example of automatic alveolar bone segmentation in ultrasound image using machine learning.

DETAILED DESCRIPTION OF EMBODIMENTS

[0069] In this description, references to “one embodiment”, “an embodiment”, or “embodiments” mean that the feature or features being referred to are included in at least one embodiment of the technology. Separate references to “one embodiment”, “an embodiment”, or “embodiments” in this description do not necessarily refer to the same embodiment and are also not mutually exclusive unless so stated and/or except as will be readily apparent to those skilled in the art from the description. For example, a feature, structure, act, etc. described in one embodiment can also be included in other embodiments but is not necessarily included. Thus, the present technology can include a variety of combinations and/or integrations of the embodiments described herein.

[0070] A method and apparatus for wireless ultrasound imaging is provided for qualitative and quantitative assessment of dental conditions and, in particular, the tooth-periodontal complex.

[0071] Referring to FIG. 1A, one embodiment of wireless ultrasound imaging apparatus 100 is shown. In some embodiments, apparatus 100 can comprise ultrasound probe 1 operatively coupled to data acquisition system 2. In some embodiments, probe 1 can comprise control switch 3 disposed thereon for controlling the operation of probe 1. In some embodiments, probe 1 can comprise any high frequency model of ultrasound probe as manufactured by Clarius of Burnaby, British Columbia, Canada. In other embodiments, probe 1 can comprise a customized ultrasound probe. In some embodiments, probe 1 can comprise an array of 1×128 ultrasound piezoelectric transducer crystals in a spaced-apart configuration wherein the array comprises a crystal spacing of 0.1 mm on centre whereby the array comprises a width of 12.8 mm. In some embodiments, data acquisition system 2 can be operatively coupled to battery management system 4 configured to provide electrical power to data acquisition system 2, wherein battery management system 4 can comprise a lithium battery 4a. In some embodiments, data acquisition system 2 can comprise a wireless transceiver and antenna to enable wireless communication with smart device 5, which can further comprise a video display for visually displaying data received from data acquisition system 2. In some embodiments, data acquisition system 2 can further comprise foot pedal switch 6 operatively coupled thereto to enable control of data acquisition system 2 by an operator. Smart device 5 can comprise one or more of a personal computer, a computing tablet, a smart phone and any other electronic computing device capable of wireless communication with other computing devices or a worldwide computing network as well known to those skilled in the art. In other embodiments, data acquisition 2 can comprise of an analog to digital converter, an ultrasound pulse signal generator and two multiplexors, as shown in FIG. 1B.

[0072] In some embodiments, apparatus 100 can comprise wireless portable ultrasound acquisition system 2 for dental imaging, comprising an ultrasound probe 1 with a control switch 3, which can be connected through a cable to a portable ultrasound acquisition system that can communicate with a smart tablet or a phone display 5 wirelessly using one or both of Wi-Fi Direct® or Bluetooth®, to display the ultrasound images. The control switch can be used to turn on/off the image acquisition. In addition, pedal 6 can also connects to the ultrasound acquisition system to control image acquisition. In some embodiments, the ultrasound acquisition unit can comprise battery 4, and can be configured to operate in emission and reception. The ultrasound probe can operate at a minimum frequency of 20 MHz and can comprise a small scale multi-array transducer 7 with matching layer. A layer of hydrogel 8 can also be incorporated to act as a delay line between the transducer and gum.

[0073] Referring to FIG. 2, one embodiment of an acquisition system for use with apparatus 100 can comprise of a microcontroller or digital signal processor or an application-specific integrated circuit (“ASIC”) 8 that can generate the ultrasound signals to the probe through D/A converter 10 to multiplexor 9. When the ultrasound signal receives from the probe, it will pass through the multiplexor to a A/D converter and then into the microcontroller or digital signal processor or ASIC. Wireless transceiver 7 can communicate with a smart tablet or a phone display 5 wirelessly using one or both of Wi-Fi Direct® or Bluetooth®, to display the ultrasound images. Battery management circuit 10 can convert the battery power to provide the voltage to the ultrasound apparatus. In some embodiments, recharging circuit 11 can be integral to apparatus 100 to recharge the battery.

[0074] Referring to FIG. 3, part (a) illustrates ultrasound penetrating a hard tissue with the speed of ultrasound υ. Part (b) illustrates the differences on the hard tissue when the speed of ultrasound is {circumflex over (v)}.sub.s instead of υ. Let ultrasound beam incident normally on a plate with thickness h and speed of ultrasound, (FIG. 3, part (a)). The thickness and the speed are related by the equation: h=υ×t/2, where, t is the time taken for ultrasound to traverse the thickness up and down the plate. If the ultrasound scanner uses a different speed, say, {circumflex over (v)}.sub.s, then a different thickness, ĥ, will be determined; i.e., ĥ={circumflex over (v)}.sub.s×t/2. Taking the ratio of two thicknesses yields:

[00001]$\begin{array}{cc}\frac{\hat{h}}{h}=\frac{{\hat{v}}_s}{v}& \left(1\right)\end{array}$

[0075] Consider a segment, {circumflex over (l)}, measured from the ultrasonograph (FIG. 3, part (b)). Its length has been distorted due to {circumflex over (v)}.sub.s used different from υ. The speed-corrected length, l.sub.c, which estimates the actual length, l, is:

[00002]$\begin{array}{cc}{l}_c=C\times \hat{l}& \left(2\right)\\ \mathrm{where}\text{:}& \\ C=\sqrt{si{n}^2\theta +{\left(\frac{v}{v_s}\right)}^{2}co{s}^2\theta }& \left(3\right)\end{array}$

[0076] is the correction factor and θ is the acute angle {circumflex over (l)} makes with the direction perpendicular to the plate (or the direction parallel to the ultrasound beam). The behavior of C in terms of θ is shown (FIG. 3, part (c)).

[0077] In some embodiments, apparatus 100 can provide a portable and an improved ultrasonic imaging system constructed to facilitate imaging the tooth-periodontium complex, qualitative and quantitative assessment of the tooth-periodontal structures of a dental client or a pet animal, in a non-invasive manner.

[0078] Referring to FIGS. 4A and 4B, embodiments of apparatus 100 are shown, which can comprise an ergonomic probe fitting within a mouth of a patient. In some embodiments, the probe can comprise small scale multi-array transducer 17, disposable gel pad (polymer/hydrogel) 18 with sleeve clip 19 to act as a delay line between the transducer and the gum. Probe head 110 can also comprise a tilt mechanism configured for rotation and tilt, for easy reach to all areas of the mouth and switch 112, which can be disposed on ergonomic handle 111, which can be used to control image acquisition. In some embodiments, apparatus 100 can comprise ultrasound acquisition system 2, which can further comprise a central processing unit (CPU), a pulser, an analog-to-digital converter, a wireless transceiver, which can comprise one or both of Wi-Fi® and Bluetooth® capability, a cable can connect the probe to the ultrasound acquisition system 2. Apparatus 100 can further comprise battery 4 so that the unit can be portable. In some embodiments, apparatus 100 can comprise display unit 5, which can comprise a smart tablet, laptop, smart phone, or smart display with one or both of Wi-Fi and Bluetooth capabilities. In some embodiments, apparatus 100 can comprise pedal switch 6, which can be connected to data acquisition unit 2, either via wire or wireless to control image acquisition.

[0079] In some embodiments, smart device 5 can comprise a memory further comprising a processor and a memory further comprising software code segments configured to cause the smart device to carry out one or more processes on ultrasonic images obtained by apparatus 100 as described herein.

Noise Removal

[0080] In some embodiments, smart device 5 can comprise software code segments configured to cause the smart device to enhance ultrasound signals representing images of alveolar bone structure and boundaries of enamel, dentin, and gingiva of a patient. To accomplish this, there are different noise filtering techniques for ultrasound imaging that can be used as linear filtering (such as Gaussian filter) and nonlinear filtering (such as adaptive median filtering and homomorphic filtering):

[0081] Gaussian filter is a convolution operation that can be applied to each image pixel with a 2×2 Gaussian kernel to remove high-frequency noises (example in FIG. 15).

[0082] The adaptive median filter can operate in a rectangular window area S.sub.xy that can be centered on the pixel (x,y) . The output of the adaptive median filtering is a new value as a replacement to the value of the pixel at (x,y) for each window-filtering time. Adaptive median filter can remove noise while keeping edges relatively sharp.

[0083] The homomorphic filtering is a process that can comprise of three stages: (i) calculating the Fourier transform of the logarithmic compressed image, (ii) applying high-pass filter function and (iii) constructing the inverse Fourier transform of the image. As a result, the homomorphic filtering can normalize the brightness across the image and enhances contrast. In the homomorphic filtering process, the filter is typical in circularly symmetric curve shape, centered at (0,0) coordinates in the frequency domain. Here, a Gaussian high-pass filter can be used to build homomorphic function.

Contrast Enhancement

[0084] Due to the inherent properties of ultrasound images and an approximate selection of the initial region, the region of interest (“ROI”) is inhomogeneous and has low contrast. The reflection from alveolar bone is scattered by the rough surfaces and the corresponding bone boundary is less focused and blurred. Therefore, a linear contrast enhancement approach was applied to enhance the contrast of the images by expanding the original intensity values of the image linearly, thus allowing a better detection of the bone boundary. An example of a noise removal and contrast enhanced image is given in FIG. 6.

[0085] In some embodiments, smart device 5 can comprise software code segments configured to cause the smart device to identify peaks and troughs of one or more of cementoenamel junctions (“CEJ”), gingival margin and alveolar bone crests of a patient using object detection and recognition. FIG. 7 illustrates one embodiment of a semi-automated process to identify CEJ, gingival margin and alveolar bone crests.

Image Preprocessing

[0086] As described above, image enhancement can be accomplished using noise removal with one or more of Gaussian filter, adaptive median filter, homomorphic filtering, and contrast enhancement.

Image Segmentation Using Multi-Label Graph Cut

[0087] To obtain an accurate and reproducible detection of the CEJ location, an initial approximate region of interest consisting of the CEJ and part of enamel and cementum was manually selected and utilized in the proposed approach. A K-means clustering can be used for the identification of foreground and background regions within the initial region of interest. The K-means (K=2) was used to set two pre-classified labels and build the initial graph, since using all of the pixels as the reference for segmentation may slowdown the execution. K-means partition pixel intensities into two initial clusters based on the similarity to the clustering centers. The centers were adjusted based on the average intensity of pixels. This step was repeated until convergence had been reached.

Edge Detection and Enhancement

[0088] Edges are important for differentiating various types of tissues (gingiva, bone, enamel) in an image. The strength of the edges is calculated by intensity gradient or the change in intensity in the direction of steepest ascent. Edge enhancement can be done with the convolution using first order derivative kernels (Sobel kernel, Canny kernel) or second order derivatives (Laplacian kernel, Log filter).

Feature Selection

[0089] After clustering the region using graph cut segmentation, the function extracts every point in the foreground region, and then detects the edge corresponding to the upper border of the enamel, cementum and alveolar bone. Since enamel, cementum, and alveolar bone are strong ultrasound reflectors, their intensities are very high in comparison with gingiva thus easy to detect. Based on the small V-shaped characteristic of CEJ/gingival margin/alveolar bone crest, our method calculates the absolute value of change along the vertical coordinate axis and then compares to the location of the previous point; the point with largest absolute value of change is seen as CEJ CEJ/gingival margin/alveolar bone crest. In other words, for the upper line of n elements u(i) with i=1,n, the differential was estimated as

[00003]$\begin{array}{cc}{u}^{\prime }\left(i\right)=\{\begin{array}{c}0,i=1\\ u\left(i\right)-u\left(i-1\right),i=\stackrel{\_}{2,n}\end{array}& \end{array}$

[0090] From that, the CEJ/gingival margin/alveolar bone crest was selected as corresponding to the maximum absolute |u′(i)|.sub.max of the differential. Finally, transforming the pixel location from coordinate into the original image, the result of the function marks the CEJ/gingival margin/alveolar bone crest in the original image. FIG. 8 illustrates an example of region extraction. FIG. 9 illustrates an example of CEJ identification process. FIG. 10A illustrates an example of comparing CEJ CAL method and manual CEJ identification on in-vitro (part (a)) and in-vivo data (part (b)).

[0091] In some embodiments, calculating changes in pocket depth (A), alveolar bone level to the CEJ (B), or gingiva thickness at the CEJ (C) of the patient (as shown in FIG. 10B) can be done using measurements between ultrasound images of different periods of time.

[0092] In some embodiments, smart device 5 can compare the ultrasound images of a patient with one or more of CBCT images, and/or MRI images of the oral structure and enhancing visualization of soft and hard tissues of the oral structure by means of Coherence Point Drift (“CPD”) registration.

Region-Growing Segmentation

[0093] This method is a common and effective approach for image segmentation. The user specifies a seed point inside the object to be segmented. Consider a pixel f as a seed point with an intensity I.sub.f. The pixels neighboring f are evaluated to determine if they should also be considered part of the object. To do so, a tolerance, ±t, is set for the lower and upper limit. The “flood fill” region-growing algorithm will add a neighboring pixel q to the pixel f's region if I.sub.q is inside the interval [(I.sub.f−t), (I.sub.f+t)]. The process is repeated recursively for the other neighbors of f to expand from the seed pixel to a coherent region.

Coherence Point Drift (CPD) Registration

[0094] The method considers the alignment of two point sets as a probability density estimation problem. By maximizing the likelihood, the CPD can fit the Gaussian mixture model (“GMM”) centroids of the moving point set to the fixed point set. The GMM probability density function, p, is

[00004]$p\left(x\right)=\omega \frac{1}{N}+\left(1-\omega \right)\underset{m-1}{\overset{M}{.Math.}}\frac{1}{M}p\left(x.Math.m\right)$$\mathrm{where}p\left(x.Math.m\right)=\frac{1}{{\left(2\pi {\sigma }^2\right)}^{D/2}}{\exp }^{\frac{.Math.x-y_m.Math.}{2{\sigma }^2}},$

[0095] where D is the dimension of the point sets, N and M are the number of points in the point sets, and the weight, w (0≤w≤1)), provides a flexible control in the presence of severe outliers and missing points. In the rigid registration, the coherence constraint was imposed by re-parameterization of GMM centroid locations with rigid parameters and derive a closed form solution of the maximization step of the expectation-maximization (“EM”) algorithm in arbitrary dimensions. The EM algorithm used for optimization of the likelihood function can comprise of two steps: E-step to compute the probabilities and M-step to update the transformation. Another advantage of the CPD is that it can preserve the topological structure of the point sets because the GMM centroids are moved coherently as a group. FIG. 11 illustrates one embodiment CPD registration between ultrasound and CBCT images. FIG. 11 shows a flow chart of the image registration process. The CBCT images can first be processed by adaptive low-pass Wiener filter to remove the white Gaussian noise. A side-tracing strategy can be employed to identify the points, which have the maximum intensity difference between adjacent pixels along each row from left to right, to be selected as the surfaces of the tooth and bone. The detected surface can be curve-fitted to remove outliers. The set of points thus obtained can be used as a reference or fixed-point set. For ultrasound images, the regions for the hard tissues can be extracted from the images by a region growing segmentation method. Then, the local contrast of the extracted images can be adaptively enhanced. In the next step, the tooth and bone surfaces from the ultrasound images can be determined when searching the largest intensity pixels along each row. Similar to CBCT, the detected surfaces in ultrasound images can be curve-fitted to remove outliers. The set of points thus obtained can be used as a moving point set. Cubic curve fitting can be used for both CBCT and ultrasound images.

An Example of Coherence Point Drift Registration Between US and CBCT

[0096] FIG. 12A illustrates an example of a CBCT point set and a correspondent US point set: part (a) represents the original CBCT image; part (b) represents the CBCT image after denoising; part (c) represents CBCT Point detection on the tooth and bone surface; part (d) represents the removal of the outliers in CBCT using the curve fitting; part (e) represents the original US image; part (f) represents the US image after segmentation and adaptive local contrast enhancement; part (g) represents US Point detection on the tooth and bone surfaces; and part (h) represents the removal of the outliers (yellow dots) in US using the curve fitting.

[0097] FIG. 12B illustrates an example of point-based evaluation for the registration.

[0098] In some embodiments, smart device 5 can comprise software code segments configured to cause the smart device to eliminate artifacts caused by multiple reflections of ultrasonic waves in the ultrasonic raw signals by means of predictive deconvolution.

[0099] FIG. 13A illustrates an ultrasonic transducer 30 comprising individual transducing elements 32 mounted just on matching layer 34. FIG. 13B illustrates ultrasonic transducer 30 with matching layer 34 mounted on an aluminum plate 36. Reverberation within the matching layer 34 can create secondary echoes or multiple reflections, which can obscure the primary reflections. The multiple reflections can be separated in time by a constant which equals to the two-way travel time within the matching layer. The multiples can be predictable and can repeat themselves at a constant time interval. The time constant can be used as an input parameter to a predictive deconvolution filter. The filter can then be used to convolve the ultrasound data. The output can be the primary ultrasound data without multiple artefact.

[0100] FIG. 13C illustrates an example of multiple reflections artifacts from ultrasound transducer 30 as shown in FIG. 13A. FIG. 13D illustrates an example of multiple reflections artifacts from ultrasound transducer 30 as shown in FIG. 13B. FIG. 14A illustrates an example without multiple reflections removed from porcine ultrasound RF signals, where the reflections are circled in red, whereas FIG. 14B illustrates an example with the multiple reflections removed from porcine ultrasound RF signals.

[0101] In some embodiments, smart device 5 can comprise software code segments configured to cause the smart device to calculate the velocity of ultrasound signals in hard tissues of the patient and to correct the detected thickness of the alveolar bone of the patient. The corrected velocity is:

[00005]$v_{\mathrm{corrected}}=2\times h_{\mathrm{corrected}}/t=2\times C\times h_{\mathrm{measured}}/t=C\times V_{\mathrm{measured}}$

[0102] where the corrected thickness is:

[00006]$h_{\mathrm{corrected}}=C\times h_{\mathrm{measured}}$

Image Segmentation Using Machine Learning

[0103] The proposed machine learning method primarily consists of an encoder and a decoder component to capture the image features, and to construct and localize the segmentation labels, respectively. All the parameters of the neural networks were initialized and computed using the training data, where the parameter values were updated iteratively to minimize a cost function. Although not used for computing the neural net parameters, the validation set was also utilized during training to determine when to stop the parameter update to prevent overfitting.

[0104] In some embodiments, smart device 5 can comprise software code segments configured to detect boundary and segments of the oral structure using multi-label graph cut optimization approach or machine learning. FIG. 15 illustrates an example of semi-automatic alveolar bone segmentation in ultrasound image using graph cut. The overall system diagram showing the proposed segmentation process: (a) Drawing a region (in yellow color) to determine the ROI on the original ultrasound image, (b) after de-noising using homomorphic filtering, (c) with contrast enhancement, (d) after smoothing using Gaussian filtering, (e) multi-label graph cuts segmentation in ROI, (f) alveolar bone extraction, and (g) final result showing delineation (in orange color) of alveolar bone boundary in the original ultrasound image. FIG. 16 illustrates an example of automatic alveolar bone segmentation in ultrasound image using machine learning.

[0105] The various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the embodiments disclosed herein can be implemented as electronic hardware, computer software, or combinations of both. To clearly illustrate this interchangeability of hardware and software, various illustrative components, blocks, modules, circuits, and steps have been described above generally in terms of their functionality. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the overall system. Skilled artisans can implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the embodiments described herein.

[0106] Embodiments implemented in computer software can be implemented in software, firmware, middleware, microcode, hardware description languages, or any combination thereof. A code segment or machine-executable instructions can represent a procedure, a function, a subprogram, a program, a routine, a subroutine, a module, a software package, a class, or any combination of instructions, data structures, or program statements. A code segment can be coupled to another code segment or a hardware circuit by passing and/or receiving information, data, arguments, parameters, or memory contents. Information, arguments, parameters, data, etc. can be passed, forwarded, or transmitted via any suitable means including memory sharing, message passing, token passing, network transmission, etc.

[0107] The actual software code or specialized control hardware used to implement these systems and methods is not limiting of the embodiments described herein. Thus, the operation and behavior of the systems and methods were described without reference to the specific software code being understood that software and control hardware can be designed to implement the systems and methods based on the description herein.

[0108] When implemented in software, the functions can be stored as one or more instructions or code on a non-transitory computer-readable or processor-readable storage medium. The steps of a method or algorithm disclosed herein can be embodied in a processor-executable software module, which can reside on a computer-readable or processor-readable storage medium. A non-transitory computer-readable or processor-readable media includes both computer storage media and tangible storage media that facilitate transfer of a computer program from one place to another. A non-transitory processor-readable storage media can be any available media that can be accessed by a computer. By way of example, and not limitation, such non-transitory processor-readable media can comprise RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, or any other tangible storage medium that can be used to store desired program code in the form of instructions or data structures and that can be accessed by a computer or processor. Disk and disc, as used herein, include compact disc (CD), laser disc, optical disc, digital versatile disc (DVD), floppy disk, and Blu-ray disc where disks usually reproduce data magnetically, while discs reproduce data optically with lasers. Combinations of the above should also be included within the scope of computer-readable media. Additionally, the operations of a method or algorithm can reside as one or any combination or set of codes and/or instructions on a non-transitory processor-readable medium and/or computer-readable medium, which can be incorporated into a computer program product.

[0109] Although a few embodiments have been shown and described, it will be appreciated by those skilled in the art that various changes and modifications can be made to these embodiments without changing or departing from their scope, intent or functionality. The terms and expressions used in the preceding specification have been used herein as terms of description and not of limitation, and there is no intention in the use of such terms and expressions of excluding equivalents of the features shown and described or portions thereof, it being recognized that the invention is defined and limited only by the claims that follow.