Inventing New Technique to Detect Early Infected Brain Tumor and Calculating the Tumor Area from MRI
Assis. prof. Dr. Aliaa Hussein Ali &
Assis. prof. Dr. Sabah Nuri Mizhir
One of the most challenging tasks in the medical science is the detection of cancerous tumors at early stage. Thus, it is necessary to have high accuracy requirements in diagnosis. The difficulty arises due to overlapping of normal with abnormal tissue in gray level of medical images. So, it is mandatory for the surgeon or the therapist of Radioisotopes to accurately delineate the boundaries of abnormal tissues from healthy ones; so as to remove the tumor without damaging the healthy tissues. The proposed computer system provide us with techniques to analyze and identify MRI by using digital images processing. Segmentation and classification scheme was used to detect brain tumor. In addition, a statistical structure analysis which depends on the structural analysis in the brain tissue in both normal and abnormal tissues was calculated. Depending on the outlines of contour process, an algorithm was used to color and transfer images from RGB system to (L*a*b*). The aim of this algorithm is to separate the color component, because the overlap between the RGB colors. After that, K-mean clusters was used to reach a high classification rang of tissues. After separating tissues by K-mean clusters and calculating some statistical properties of examination tissues, tissues were analyzed by Co-occurrence matrix. This process was proved to be the best as it is effective in separating the normal tissues from the abnormal ones. The result shows that both the tumor diagnosed by the proposed programs and the tumor clinically diagnosed by oncologist is identical for all samples which were tested in Al-kadhimiya Hospital. Also, the work included these result.
1. The Thresholding algorithm which was applied on MRI was able to detect abnormal tissues (tumor) and was able to separate the abnormal tissues from normal ones.
2. Removing the noise of MRI by using the wavelet transformation . Using the clustering method which is the modified K-mean that depend on the color and distance , best result can be obtained.
3.The Co-occurrence matrix of normal and abnormal brain tissues was calculated. The values of Energy tissues record the highest in the sold tissues and the lowest in healthy tissues. The results were obtained when normal and abnormal tissues were colored by (L*a*b*). and separated each layer from others.
4.The values of Entropy record the highest in healthy tissues and the lowest in sold tissues and they inversely proportion to the energy.
5. The values of correlation record the highest in healthy tissues and the lowest in the tumors.
6. The values of Contrast and Standard deviation record the highest in the healthy tissues and the lowest values in tumors.
7. The values of Homogeneity are relatively same for all brain tissues. In addition, the aforementioned values can be used in diagnosing the purity of tissues.