From Human Observation to Automated Detection: How Are Artificial Intelligence Algorithms Transforming the Future of Diagnostic Radiology? — Maryam Reda Al-Sur‘a Abstract The integration of artificial intelligence into diagnostic radiology represents one of the most significant technological developments in modern medical imaging. Traditionally, radiological image interpretation has relied on radiologists’ visual expertise and clinical skills to identify anatomical changes, detect pathological lesions, and support medical decisions. With rapid advances in machine learning, particularly deep learning, computer systems have become capable of analyzing medical images, recognizing complex patterns, and identifying regions that may indicate pathological changes. These developments have expanded the applications of artificial intelligence across conventional radiography, computed tomography, magnetic resonance imaging, and mammography. Artificial intelligence systems offer promising capabilities for detecting pulmonary nodules, intracranial hemorrhage, bone fractures, and other radiological findings. They may also improve workflow efficiency and support quantitative image measurements. However, the clinical value of these technologies depends on their diagnostic reliability, their ability to perform across different clinical settings, the interpretability of their results, and their integration into everyday medical practice. This article reviews the historical development of artificial intelligence in diagnostic radiology, its major applications, potential benefits, and the scientific and ethical challenges associated with its use. It emphasizes that artificial intelligence should support medical expertise rather than replace specialized clinical judgment. Keywords: Artificial intelligence, diagnostic radiology, deep learning, medical imaging, computer-aided detection, radiological diagnosis, convolutional neural networks. 1. Introduction Medical imaging is a cornerstone of modern healthcare. It provides essential information for disease detection and diagnosis, treatment planning, and monitoring disease progression. Radiological image interpretation depends on the specialist’s ability to recognize subtle anatomical changes, distinguish normal structures from pathological lesions, and relate imaging findings to the patient’s clinical information. However, the continuous increase in the number of imaging examinations and the complexity of the resulting images has created new challenges concerning the speed, accuracy, and consistency of interpretation. In this context, artificial intelligence has emerged as a promising technological approach to enhancing medical image analysis. Unlike some traditional image-processing techniques that rely on predefined rules, modern artificial intelligence algorithms can learn visual patterns from large collections of labeled medical images. Deep learning, a branch of machine learning, is one of the most important techniques used in this field. It enables computer models to extract features at multiple levels, ranging from edges and textures to more complex patterns associated with certain pathological findings [1, 2]. Artificial intelligence applications have evolved from computer-aided detection systems into a broader range of tools. These include lesion localization, tissue segmentation, classification of findings, quantitative measurements, and prioritization of examinations according to urgency. Nevertheless, technological progress does not automatically lead to improved health outcomes. Each system should be evaluated according to its intended diagnostic task, the data used for training, and the clinical environment in which it will operate. 2. From Conventional Radiology to Artificial Intelligence Historically, radiological diagnosis relied on direct visual interpretation of images by radiologists. The introduction of digital imaging and picture archiving and communication systems (PACS) improved image storage, retrieval, and sharing, while also supporting the development of workflows within healthcare institutions. The emergence of computer-aided detection and diagnosis (CAD) systems marked an important transition toward automated image analysis. Early systems relied on manually designed features and predefined rules to identify suspicious regions. Although these systems contributed to the development of the field, their ability to handle anatomical complexity and substantial variation between images was limited in some applications. Subsequently, machine learning enabled the development of models that recognize statistical relationships between image features and diagnostic classifications. Deep learning then introduced further advances by allowing neural networks to learn complex image representations directly from training data [1]. Convolutional neural networks (CNNs) became an important technique in medical image analysis because of their ability to extract spatial features and recognize anatomical patterns.