Assist. Lect. Ahmed Salman Jassim Artificial intelligence (AI) has become one of the most important technological developments in modern medical imaging, particularly in Computed Tomography (CT). CT uses ionizing X-rays to generate detailed cross-sectional images of the human body; therefore, maintaining diagnostic image quality while minimizing radiation exposure remains an important clinical objective. Machine learning and especially deep learning (DL) are increasingly being incorporated into different stages of CT examinations, including patient positioning, protocol optimization, image reconstruction, image analysis, lesion detection, and workflow automation. Recent evidence suggests that AI is particularly valuable for CT radiation-dose optimization and image-quality improvement. PubMed 1. AI-Based CT Image Reconstruction One of the most established applications of AI in CT is Deep Learning Reconstruction (DLR). Reducing the radiation exposure during CT acquisition generally increases image noise, which may compromise visualization of subtle abnormalities. Deep neural networks can be trained to differentiate image noise from true anatomical information and reconstruct cleaner images from low-dose datasets. DLR techniques have demonstrated considerable ability to reduce image noise while preserving clinically useful anatomical details. Several studies have reported advantages over conventional Filtered Back Projection and, in some applications, iterative reconstruction techniques. PubMed 2. Radiation Dose Reduction Radiation-dose optimization is among the most clinically important applications of AI in CT. AI may contribute to dose reduction through automated patient positioning, scan-range optimization, selection of appropriate acquisition parameters, noise suppression, and reconstruction of diagnostic-quality images from lower-dose acquisitions. A 2025 scoping review including 90 studies identified three major AI-related areas for CT dose optimization: patient positioning, scan-range determination, and image reconstruction. PubMed Clinical evidence also indicates that substantial reductions can be achieved in selected applications, although the magnitude of dose reduction varies considerably according to anatomical region, protocol, diagnostic task, scanner, and AI algorithm. Consequently, a single percentage reduction should not be generalized to all CT examinations. PubMed 3. Automated Detection and Image Analysis AI can also analyze CT images and identify suspicious abnormalities. Deep-learning systems may assist with tasks such as lesion detection, anatomical segmentation, lesion-volume measurement, and longitudinal comparison. These applications fall within the broader areas of Computer-Aided Detection and Computer-Aided Diagnosis. Such systems can help radiologists identify potentially important findings, but their output should be interpreted in conjunction with clinical information and expert assessment. Differences between training data and local clinical populations or scanner systems may affect algorithm performance. 4. CT Protocol Selection CT departments use numerous imaging protocols depending on the anatomical region, clinical indication, patient characteristics, and contrast requirements. AI-based decision-support systems can use patient and examination information to assist radiographers or radiologists in selecting appropriate CT protocols and acquisition parameters. This creates an important area for human–AI interaction research, particularly investigations of how the timing and presentation of AI advice influence radiographers' protocol-selection decisions. 5. Automated Patient Positioning Accurate patient centering is an important component of CT image-quality and dose optimization. Computer-vision and AI-based systems can estimate patient position and anatomy and assist with automatic positioning and scan-range selection. Recent literature suggests that these approaches can reduce positioning errors and unnecessary scanning associated with manual procedures. PubMed 6. Workflow Optimization Artificial intelligence may also improve CT workflow through automated examination preparation, reconstruction, anatomical segmentation, measurements, and prioritization of examinations. The objective is not simply to replace manual work. Rather, these tools can automate repetitive processes and allow radiographers and radiologists to focus more attention on patient care, image evaluation, and complex clinical decisions. 7. Challenges and Limitations Despite its potential, AI in CT presents several limitations. AI models require extensive validation across different patient populations, scanners, institutions, and clinical indications. Excessive image denoising can alter image texture, and performance may be reduced for some subtle or low-contrast lesions. Therefore, evaluation of AI reconstruction should include diagnostic-task performance rather than relying exclusively on visual image quality or noise measurements. PubMed Other concerns include algorithmic bias, data privacy, cybersecurity, explainability, integration with existing clinical systems, regulatory oversight, continuous quality assurance, and appropriate staff training. Conclusion Artificial intelligence is becoming an important component of modern CT imaging. Major applications include deep-learning reconstruction, image denoising, radiation-dose optimization, automatic patient positioning, protocol-selection support, automated image analysis, and workflow optimization. Current evidence indicates that AI has considerable potential to improve CT image quality and operational efficiency while reducing radiation exposure. Nevertheless, the magnitude of benefit depends on the clinical application and imaging system. AI should therefore be implemented with clinical validation, quality assurance, and appropriate human oversight. PubMed