<p>Automated radiology report generation can significantly enhance the efficiency and consistency of clinical workflows, while decreasing the workload of radiologists “Producing high-fidelity medical reports from radiological images is the main challenge involved, however, such as mapping intricate visual features to domain-specific textual descriptions accurately. These challenges are further worsened by imaging modality variations (e.g., CT, X-ray), poor image quality, imbalanced data, and annotation discrepancies. Current models have a hard time in detecting nuanced abnormalities and reporting them in clinically meaningful terms. extract my abstract and edit. To overcome such limitations, we introduce a new system called Automatic Radiology Report Generation using Contextual Learning (ARRG-CL), which combines CLIP (Contrastive Language-Image Pretraining) for shared image-text matching with BERT (Bidirectional Encoder Representations from Transformers) for contextual text comprehension. In order to generate radiology reports, this work investigates a structured integration of pretrained CLIP and BERT models, focusing on contextual improvement and embedding-level fusion as opposed to end-to-end retraining. The ARRG-CL framework achieved strong medical report generation performance, with BLEU-4 (0.858/0.884), ROUGE (0.905/0.92), and CIDEr (0.765/0.76) scores on the Open-I and ROCO datasets, respectively as opposed to conventional CNN-RNN architectures that generally achieve lower report similarity scores. Standard, publicly accessible evaluation scripts with the same tokenization parameters and full-report level comparison were used to calculate all stated evaluation scores; post-hoc filtering and manual adjustment were not used. This paper introduces a new multimodal architecture that closes the gap between clinical natural language processing and computer vision, providing a scalable and high-performing medical report generation solution.</p>