<p>Vascular calcification often associated with cardiovascular diseases is considered a prominent cause of mortality globally. Coronary artery calcification scores are valuable in predicting cardiovascular events but are underutilized due to limited availability, high cost as well as radiation exposure. Type 2 Diabetes mellitus exacerbates the cardiovascular risk through chronic inflammation as well as insulin resistance. To overcome these shortcomings, this research paper proposes a new a novel Gated Attention GoogleNet⁺⁺ based Early Warning System for early detection and risk assessment of cardiovascular patients. The proposed model incorporates Coronary Computed Tomography Angiography images and clinical data collected from Type 2 diabetes mellitus patients. Data pre-processing steps namely normalization, resizing, noise removal, data augmentation, and data splitting are employed to clean the data thereby ensuring data quality. A Gated Attention UNet model is developed for accurate segmentation of calcification regions in computed tomography angiography images via gated spatial convolution and attention gate to focus on the most critical features. To refine feature maps, a GoogleNet⁺⁺model is employed in the extraction of features thereby enhancing the classification by integrating Exponential Linear Units and employing a Convolutional Block Attention Module. A Multilayer Perceptron network is then employed to predict cardiovascular conditions and assess the calcification scores that categorize the risk as high, moderate, and low. The early warning system generates alerts based on pre-defined thresholds that ensure timely intervention and prevention of cardiovascular risks. The results then demonstrate exceptional performances with high accuracy of about 98.82%, and recall of 98.11%. Thus, the results highlight the potential of providing an effective, non-invasive, and accessible solution for early cardiovascular risk detection.</p>