<p>Atherosclerotic plaques are the primary cause of most acute coronary events and are directly associated with coronary artery disease (CAD). Early identification of plaque is essential for timely diagnosis and effective treatment. Intracoronary optical coherence tomography (IOCT) provides high-resolution images with unique capability to reveal subtle structural changes in the arterial wall. This research proposes a hybrid deep-learning framework for plaque detection and classification using IOCT images. The pipeline begins with preprocessing steps, including Gaussian filtering and histogram equalization, to enhance image quality. Hierarchical feature extraction is performed using Convolutional Neural Network (CNN), while Support Vector Machine (SVM) classifier assists in discriminative analysis of normalized grayscale inputs. Auto-thresholding and morphological filtering refine the segmentation process, improving membrane continuity and reducing artifacts. For detection, the Faster R-CNN framework identifies vulnerable plaque regions using multi-scale anchors and shared ResNet-101 features. Candidate regions are generated through a Region Proposal Network (RPN), and redundant bounding boxes are eliminated using Non-Maximum Suppression (NMS), ensuring accurate localization of plaques of varying sizes. Classification is performed using ResNet and DenseNet architectures in both Cartesian and polar domains. A hybrid network integrates plaque detection with lumen contour analysis to enhance overall diagnostic performance. The proposed framework achieves an overall accuracy of 99.05%. The Area Under the Curve (AUC) values are 0.981 and 0.934 for lumen and plaque discrimination, respectively, demonstrating strong discriminative capability. Clinically relevant metrics further confirm reliability, with sensitivity values of 96.8% and 91.0%, and F1-scores of 0.961 and 0.902 for lumen and plaque segmentation, respectively. </p>