In past two decades, the Intra Vascular Ultrasound (IVUS) imaging technique is utilized for detecting the calcified plaque in the coronary artery aided with deep learning techniques. However, the existing approaches fall on computational complexity, time consumption and poor accuracy of plaque localization. In this paper, a deep CNN based plaque detection framework is proposed to tackle the issues in the existing approaches. Initially, the region between lumen and media contour is segmented with the terminus sector segmenting approach based on pixel concentration divergence and the texture features. Then the discriminative features are extracted through the meticulous feature extraction approach via construction of contraction path in multi-scale CNN. Finally, the location of the plaque has been identified with unambiguous detection and localization approach, whereas the expansion path is constructed for plaque detection. The experimental result shows the effectiveness of the proposed framework and the results are compared with other existing approaches.