DESC:OTITIS DETECTION USING NON-INVASIVE METHOD:

Otitis media is also known as the middle ear infection majorly occurs with children of age between 0 and 17 years. Infants and toddlers cannot even express their feelings through normal human communication. Hence, it is important to device the system that is non-invasive and instantaneous to accurately detect and report the infection. As we all know that middle ear infection happens because of fluid trapped in the middle ear and that allows bacteria or virus to grow to cause the infection. This swells the ear and prevents air from entering into the middle ear causing hearing loss.

Technology Overview:
The system built for detecting otitis media is divided into following segments.
AMI OTOSCOPE can take pictures of the middle ear and transmits data to AMI OTOSCOPE application through BTtechnology.AMI OTOSCOPE application algorithm can detect inflammation and infection in the ear

AMI Otoscope overview
AMI device ( Integrated device) has built in camera and flash with an accessory ear-lens attachment (AMI Otoscope) that takes pictures of the Tympanic Membrane (TM).
The ear lens accessory contains a lens for magnification and fiber optics that transmits the flash light into the ear.
To avoid the user seeing his tympanic membrane on the smart phone/Bluetooth enabled computer, only a wire frame view of the picture is shown to the user. This is done to avoid the user from having bad physiological effects.

Otitis Media:
Otitis media signifies middle ear inflammation in general terms and is classified as AOM (Acute Otitis Media), OME (Otitis Media with Effusion) and NOE (No Effusion).
The presence of Middle ear infection creates a few changes in the Tympanic membrane which can be used to detect Otitis. The tympanic membrane abnormalities due to the middle ear infection include change in the color, transparency and shape (Bulging / inflammation), in addition to effusion of middle ear.

Image processing technique overview:

Segmentation:
Before processing the captured image for Otitis media detection, it is first segmented to extract the relevant regions which are useful for the algorithm for computation. Active contours based algorithms are implemented in the grayscale version of the input image. A set of forces that outline the gradients and edges are formed. The algorithm iteratively grows the contour and stops at a pre-defined convergence criterion. The result of this is an outline that covers the relevant region in the image. This outline is used to generate the mask which when applied to the input image gives us a segmented image containing only pixels of our region of interest.

Concavity:

A gray scale version of segmented image is used to find the concavity. A concave region in the center of tympanic membrane is located to detect bulging in the middle ear. Sliding window is done to obtain the local circular neighborhood XR(r,?) of radius R. This circular neighborhood is then transformed into polar co-ordinates to obtain XR(r,?) with r e {1,2,3,…,R}, ? e [0,2p] and
r = v(m-mc)2 + (n-nc)2 , ? = arctan (n-nc)/(m-mc), where (mc , nc) are the center co-ordinates of the neighborhood XR.
The center area of the concave region is dark and gets brighter while moving towards the outer periphery. That is, as the radius increases, the brightness increases as shown in the figure above. This color contrast will be more clearly visible in the grayscale and WND-CHARM image classifier is used to make the color difference between center and outer periphery
more obvious.

Translucency:
Tympanic membrane turning gray is a primary characteristic of NOE. While the TM turns gray, it appears to be translucent which is used to measure the greyness of the TM. A snapshot of TM is taken at different lighting conditions and viewing angles and is used to structure the TM. Gray level clusters in the translucent region are extracted by segmenting the image. These extracted pixels are clustered using K-means clustering. Euclidean distances are computed to the center of cluster. If any K distance is lesser than the set threshold, then it is considered to be translucent. The overall translucency is defined as the mean of the binary image Xt.

Amber Level:

OME is predominantly Amber or pale yellow. The amber and non-amber region is differentiated using color assignment techniques and the result is a binary image with amber and non-amber regions. The mean of amber region is calculated to get the amber levels of the image.

Bubble presence:

Visible air fluids levels or bubbles behind the TM are an indication of OME. Red and Green channels from the original image is given to Canny edge detection algorithm creating a binary image Xb. Filtering and morphological operations are performed on the obtained binary image to enhance edge detection and obtain smoother boundaries. Mean of the binary image Xbis calculatedto define the level of bubble presence.

Light:
Illuminated image of the TM is taken for detection of AOM. The distinct bulging of AOM results in non-uniform illumination of the TM (in contrast to the uniform illumination in NOE). Contrast enhancement on the grayscale image is done to make the non-uniform lighting prominent. Ratio of the means of bright region to the dark region is taken to define the non-uniformity. AOM is confirmed to exist beyond a set threshold ratio limit.

Conclusion:

Therefore based on the bulging (Concavity), translucency, light, grayscale variance and bubble presence, the existence of middle ear infection is detected and is classified as per the below table.

Symptoms based on classification of Otitis:

Parameter AOM OME NOE
Color Pale yellow, White, Substantially red. Blue, Gray, White, Amber Gray, Pink
Position Bulged Neutral, Retracted Neutral, Retracted
Translucency Opaque Opaque, Semi-opaque Translucent

,CLAIMS:
a.A method of detecting otitis media is divided into following segments that can take pictures of the middle ear and transmits data to AMI OTOSCOPE application through BTtechnology.

b.First of its kind to have a built in camera and flash with an accessory ear-lens attachment that takes pictures of the Tympanic Membrane and process through application algorithm can detect inflammation and infection in the ear

c. First of its kind to have the image screened and processed to special algorithm and sent to the specifically designated smart phone/Bluetooth enabled computer which can store the results including image of the Tympanic membrane for future reference.