FORM 2
THE PATENT ACT 1970
(39 of 1970)
&
The Patent Rule, 2003
Complete Specification
(see section 10 and Rule 13)
TITLE OF INVENTION: Lung cancer detection using modified Gabor filter (MGF), Gradient operators and morphological segmentation tool

2. PREAMBLE TO DESCRIPTION:
Lung Cancer is a disease of irregular cells multiplying and growing into a tumor. It's hard to believe, but lung cancer is the primary cause of cancer deaths among both women and men. Every year more people die of lung cancer than of colon, prostate and Breast cancers. Some important facts of lung cancer are Excluding Skin cancer, lung cancer is the second most common cancer in both women and men. Statistics from Indian Council of Medical Research (ICMR) recommend that lung cancer is fast turning into a plague in India. It is a high mortality cancer due to poor access to affordable health care and diagnosis at late stage .Globally lung cancer accounts for 8.4% percent of all cancers in women and 14.5% in men. For lung cancer Smoking is the single largest contributor. Other causes are exposure to carcinogenic toxins like radon, asbestos, radiation and air pollutants. Exposure to women to smoke from the burning of charcoal for cooking is also a cause of lung cancer. In recent times, image processing measures are frequently used in a number of medical areas for enlargement of the image in preceding recognition and managing periods, where the instant aspect is really important to determine the abnormality problems in objective figures, mainly in a variety of malignancy tumours such as lung cancer, breast cancer etc..
3. Description
In order to overcome the drawbacks of existing system, new method is proposed for the detection of lung cancer in early stages. This new method involves Modified Gabor filter, gradient operators and Morphological segmentation tool and this proposed method gives promising results compared to other methods. Figure shows the Architecture of lung cancer detection system.


Image Enhancement using Modified Gabor Filter (MGF):
The Gabor filter is applied to 2D (two-dimensional) X-ray images for the analysis and they related directly to Gabor wavelets. Image enhancement is the initial and important stage in lung cancer detection; it takes X-ray lung image as the input from the database. In Gabor filter the number of rotations and dilations are present and these are time taking. This results in blurred image which is not appropriate for next stages in image processing. To overcome this drawback a Modified Gabor Filter (MGF) approach has been implemented. The modifications are as follows.
• At the initial stage spatial aspect ratio is not considered, this leads to reduction of distortion of image at the beginning.
• The spatial aspect ratio is considered at the kernel size directly.
• This helps to get clear images and significant change in PSNR (Peak Signal to Noise ratio) values.
• The above steps provide improved overall response:
Feature extraction using Gradient operators
The output of Image Enhancement is given as input to the feature extraction. In this stage features of the lung image will be extracted by using gradient operators.
Modulated intensity gradient based segmentation
Modulated intensity gradient method involves convolving image with the gradient operators. High value of gradient magnitude can be points with sudden change between intensity of two regions and these points are called edge pixels. To form closed boundaries these points linked together. Usually Laplace operator, laplacian of Gaussian (LOG), sobel operator, canny operator etc are used as operator in this gradient based segmentation method. In Digital image processing edge detection is very useful. Out of numerous methods, in this research work canny operator is used instead of sobel operator because Sobel operator is simple, but its accuracy suffers in noisy environment but canny operator or canny edge detector has many advantages such as smoothing effect to remove the noise present in the images, through non-maximal suppression improves the signal to noise ratio. The only disadvantage is it is time taking because use of complex algorithms used in canny operators.
Texture based segmentation
The term texture is difficult to define, but it represents aspects of surface pattern such as colour, regularity, coarseness and directionality. Texture is a phenomenon that is widespread, hard to define and easy to recognise. Features of the texture can be used for segmentation. The combination of Texture gradient and modulated intensity gives the total gradient. Here texture and intensity gradient information is combined to obtain final gradient, capturing all perceptual edges in the image. Here the Dual tree Complex Wavelet transform (CWT) is used

for feature extraction. The multidimensional (M-D) dual tree CWT is based on a computationally competent, independent Filter Bank (FB).
Image segmentation using Morphological Segmentation tool (i.e. texture watershed)
The output of feature extraction is given to the segmentation stage .In this stage the objects are separated from the background, as well as from each other. The term watershed refers to a edge that divides areas drained by different river systems. A catchment basin is the geographical area draining into a river and initially watershed algorithm was implemented to find the water level and water basis from the satellite images. Segmentation means process of partitioning a binary image or gray scale image into multiple segments i.e. set of pixels. Segmentation of non trivial images is one of the difficult and important tasks in image processing. Segmentation accuracy determines the failure or success of automated analysis procedures. The watershed transform is a morphological segmentation tool, when applied to the gray scale image or gradient image the resulting regions corresponds to local troughs in the gradient, while the boundary or watershed pixels corresponds to peaks. Traditionally the
gradient image used is computed as where 1is the gray scale intensity. This captures
the perception that region boundaries are likely where the intensity gradient is large. However, this formulation ignores the fact that humans are able to discriminate not just between regions of homogenous intensity, but also between those of homogenous texture. The current method follows the approach of which integrates the measure of spatial variations in texture with the traditional intensity gradient. The watershed transform is also called as texture watershed and it consist of number of conceptual stages.
• Compute a texture representation that characterizes a local area surrounding each pixel. • Post process the texture features to make them suitable for meaningful gradient
extraction. • Generate gradient images for each of the texture features, as well as for gray scale
intensity and potentially colour. • Normalise/weight the contribution of each gradient image. • Combine the various gradient images to form the single valued gradient surface. • Segment by applying the watershed transform to this surface.

3. FIELD OF INVENTION: In medical field i.e. Lung cancer Detection
3.1 BACKGROUND OF INVENTION:
In early stages it is important to detect the lung tumour with minimum time delay and give a improved solution to reduce the lung cancer .Detection of lung cancer consist of three stages like Image enhancement, feature extraction and Image segmentation. Research work aiming Image precision and superiority is the interior aspects of this research, picture superiority, growth as well as dimensions are depending on the development phase where small pre processing methods are used based on Modified Gabor filter within Gaussian policies for Image enhancement. For extracting required features gradient operators are used. For segmentation stage Morphological segmentation tool used which consists of number of conceptual stages. Several researchers developed Image processing techniques for the detection of lung cancer. For Image enhancement earlier researchers used kalman filters, Hessian Based filters but these methods have drawbacks like poor and non uniform response for images of varying sizes and varying contrast. Some of the researchers used interpolation methods which is complex and time taking. Other methods for enhancement are by using Discrete Cosine Transform (DCT), Fast Fourier Transform (FFT) , Auto Enhancement Algorithm (AEA) ,Discrete Wavelet Transform(DWT) all these methods are time consuming and less accurate. To overcome these drawbacks a modified Gabor filter is used for Image Enhancement. For feature extraction earlier researchers employed methods like Binarization, Gray level Co-occurrence matrix (GLCM) and Masking approach. Depending on these methods, conclusion is prepared whether the lung has nodule or not but these methods have drawbacks like visibility of the geometric and intensity based statistical features are not clear and physical measurements i.e. shape measurements are not clearly visible that will decide the appearance of the object. To overcome these drawbacks gradient operators are used for feature extraction which consist of modulated Intensity gradient and Texture gradient. Segmentation algorithms like watershed algorithm, Thresholding approach are having many drawbacks like
o Local structures of images such as boundaries and flat areas are appeared differently, o Image segmentation is a low level procedure in image analysis. So, segmentation
algorithms partition images using low level concepts of images such as texture,
pixel colour, pixel intensity. o Another drawback is the lack of exact criteria for evaluating segmentation
algorithms.
To overcome above drawbacks texture watershed segmentation method is used for segmentation. In this research work X-ray lung images and CT scan lung images are collected from various hospitals and they are analyzed by using Image processing techniques. This new method gives promising results compare to other methods. Hence, this new technique can be used for early detection of lung cancer and also it is helpful to medical practitioners, medical equipments manufacturing industries.

SUMMARY OF INVENTION:
The modifications in Gabor filter (MGF) include the consideration of spatial aspect ratio at the
kernel size directly instead of at the initial stage. This modification results in reduction of
distortion of the image at the beginning and helps to obtain clear images at the initial stage.
Applying dual tree CWT on the gradient operators for extracting features it has many
advantages like smoothing effect to remove noise present in the images through non-maximal
suppression and improves the signal to noise ratio. For the image segmentation texture
| watershed is used, in this stage maxima is calculated (i.e. lowest deepest point) to apply
i watershed. Finally the segmented image is superimposing with the original image, depending
on the segmented image the presence of lung cancer is identified.

4.CLAIMS:
By using the MGF, gradient operators and morphological segmentation tool an efficient lung cancer detection system is developed.
1. The modifications in Gabor filter (MGF) include the consideration of spatial aspect ratio at the kernel size directly instead of at the initial stage. This modification results in reduction of distortion of the image at the beginning and helps to obtain clear images at the initial stage.
2. Applying dual tree CWT on the gradient operators for extracting features it has many advantages like smoothing effect to remove noise present in the images through non-maximal j suppression and improves the signal to noise ratio. Here real part and imaginary parts of the : image are added to get the total gradient image.
3. For the image segmentation watershed transform is used i.e. texture watershed, in this stage maxima is calculated (i.e. lowest deepest point) to apply watershed. Finally the segmented image is superimposing with the original image, depending on the segmented image the presence of lung cancer is identified.
4. For analysis a normal X-ray lung image and abnormal lung image is selected from the database and results are found to be fruitful. The presence of the cancer cells is decided by the next stage by using neural networks. Hence, this proposed method of lung cancer detection using MGF, gradient operators and morphological segmentation tool helps in early detection of cancerous cells present in lungs.