Liver CT image
CONCLUSION
Our proposed study is on automatic liver cancer detection in abdominal liver images using optimization techniques. In this first stage of the implemented methodology, performance is analyzed based on entropy, energy, mean, standard deviation, Accuracy, and elapsed time. The new automatic segmentation technique for liver cancer detection is also developed. The new proposed strategy is introduced based on ROI and Adaptive Watershed Algorithm. Moreover, the results of this proposed research result yield clear dimensions about normal and abnormal segmentation of the cancerous region of the liver, enabling the physicians to have a consistent approach. The region growing, intensity-based thresholding and proposed statistical parameter-based segmentation methods can better be used to Segment tumors. Then in the second stage of the research work, CT image segmentation is an essential activity to evaluate the image contains any part of the renal, spleen, etc. For the segmentation of the medical image, contour and clustering methods were extensively employed. We, therefore, provided in this chapter the watershed segmentation of liver cancer. The initial segmentation of the kidney and spleen was done. GLCM was used to remove features from the tumor fragments after the segmentation of the cancerous tissue. Finally, improved residual Google net CNN classifier was employed to classify the liver cancer as hepatocellular carcinoma and metastatic carcinoma. The proposed classifier obtained a higher detection accuracy of 99.32 percent. It ensures that the new segmentation method should be able to identify the exact boundary structure of the cancer area and preserve the essential details that allow it to properly diagnose. The rating system guarantees effective cancer lesion defection without the need for manual method interference. This approach may be very useful for early-stage cancer diagnosis among clinicians
In this research work, we proposed an automatic computer-aided method to detect, classify, and segment the tumor region from the liver CT image. The feature selection was made to select the best-extracted feature patterns from the liver images. The segmented liver tumor region can be diagnosed using ANN and improved residual Google Net CNN classifier for its classification into either benign or malignant stage. The performance is analyzed in terms of specificity, sensitivity, positive and negative predictive values, and Accuracy. The average accuracy value of the proposed liver tumor detection system for malignant images is about 99.32% in accordance with their corresponding CT liver images. This accuracy level of the liver tumor detection system can be improved by implementing soft computing and deep neural network classifier.
FUTURE SCOPE
Further extension of this research work can be done in the following ways:
- This work can be extended to detect and diagnose liver fibrosis on the thermogram images.
- Hepatic cancer can be detected and diagnosed from the thermogram images.
- Fuzzy rules can be built using fuzzy logic to improve the classification accuracy of liver diseases.
- Feasibility of applying this research work for 3D liver images to detect the liver abnormalities can be done.