International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056 Volume: 07 Issue: 12 | Dec 2020 www.irjet.net p-ISSN: 2395-0072 © 2020, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 1436 AN EFFICIENT APPROACH FOR THE LIVER LESION DETECTION FROM THE CONTRAST ENHANCED US IMAGES T. Mathan Kumar 1 Dr. G. Nallasivan 2 Dr. M. Vargheese 3 II Year M.E (CSE) Professor/ CSE Professor/ CSE PSN College of Engg & PSN College of Engg & PSN College of Engg & Technology Technology Technology Tirunelveli Tirunelveli Tirunelveli ----------------------------------------------------------------------***----------------------------------------------------------------------- ABSTRACT: This paper proposes an automatic classification method based on machine learning in Ultrasonography of focal liver lesions using image processing techniques. There are different techniques used for the segmentation of lesions from the image. Active contour is one of the active models in segmentation techniques, which makes use of the energy constraints and forces in the image for separation of a region of interest. Active contour defines a separate boundary or curvature for the regions of the target object for segmentation. This method can yield spatial and temporal features based on a discrete wavelet transform. The lesions are classified as benign or malignant liver tumors using support vector machines (SVM) with a combination of selected texture features. The experimental results are consistent with guidelines for diagnosing FLLs. I - INTRODUCTION The differential diagnosis of focal liver lesions includes a broad spectrum of benign, malignant, and infectious etiologies. Focal liver lesions in humans include neoplasms, metastatic lesions, inflammatory masses, and cysts (congenital or acquired); primary neoplasms – both benign and malignant – are 1%-2% of all tumors. Many studies suggest that benign neoplasms are less frequent than malignant tumors. Primary liver neoplasms are in third place, in order of frequency, among malignant intra- abdominal masses in the pediatric population, after Wilms tumors and neuroblastomas, with an incidence of 5-6%. Although liver tumors are the most frequent malignant GI tumors, they are less than 2% of all malignant processes. Most humans with benign or malignant liver masses come into a physical exam with palpable masses. Other symptoms include pain, anorexia, jaundice, paraneoplastic syndromes, hemorrhages, and congestive heart failure. Several factors help when making a differential diagnosis, such as the age of the child, characteristics of the images taken, clinical presentation, levels of alpha-fetoprotein, and whether it is a single or multiple lesions. Liver tumors associated with high serum levels of alpha-fetoprotein include hepatoblastoma and hepatocellular carcinoma. Infantile hemangio endothelioma may have high levels in a minority of lesions (< 3%). The presence of multiple lesions suggests metastatic disease, infantile hemangioendothelioma, abscesses, cat-scratch disease. Adenomas or lymphoproliferative diseases in predisposing conditions, such as Fanconi’s anemia or Gaucher’s disease. Clinical presentation may suggest a specific diagnosis. Several computer-aided diagnosis (CAD) techniques that improve the objectivity of diagnosing FLLs with CEUS have been proposed. For classifying FLLs into four classes (HCC, hepatic hemangioma, FNH, and liver metastasis) using a support vector machine (SVM) with parameters obtained from a TIC analysis of the arterial phase. A neural network with four parameters obtained from a TIC and achieved 93.4% sensitivity and 89.7% specificity for 112 cases. Their method used 43 parameters obtained from max-hold images in CEUS and classified FLLs with six neural networks in a cascade. The method for classifying FLLs as benign or malignant using an enhancement pattern of a differential TIC between the FLL and parenchyma ROIs obtained by ROI tracking based on the scale-invariant feature transform (SIFT) key points detector. The proposed system correctly classified 13 lesions out of 14 cases. a method of benign and malignant classification using TICs of FLL and parenchyma ROIs obtained by ROI tracking based on Compact and Real-time Descriptors (CARD) and mean shape estimation of ROI based on a Generalized Procrustes Analysis (GPA). This method achieved 91.6% accuracy for 107 cases.