**3. Utility of AI in radiology**

AI is increasingly being used in radiology to improve diagnostic accuracy, efficiency, and decision-making. Some of the most common applications of AI in radiology include image analysis, computer-aided diagnosis (CADe), image segmentation, automated image interpretation, and automated reporting [4–6].

AI-based systems can be trained to detect and identify specific structures or abnormalities on medical images such as tumors, blood vessels, or organ abnormalities [5, 7, 8]. This can improve diagnostic accuracy and efficiency by highlighting potential abnormalities that may have been missed by radiologists [9]. Additionally, AI-based systems can be used to assist radiologists in the diagnostic process by classifying different types of tumors or identifying specific patterns on medical images, which can help radiologists make more accurate, specific diagnoses, and guide treatment decisions [4, 5, 8]. AI-based systems can also be used to generate automated reports and summaries that include relevant information and analysis, which can save time, reduce the workload and errors caused by manual reporting, and improve communication with other healthcare providers [4, 6, 10].

AI applications have the protentional to be used in radiology for detection and characterization in many body systems [8, 9]. Recent advances in AI for thoracic applications have focused on using deep learning techniques to assist with a lung cancer diagnosis and pulmonary nodule detection on CT scans. In abdominal and pelvic applications, AI has been used to assist with liver lesion analysis and the detection of abnormalities on CT and MRI scans. General lesion analysis using AI typically involves training a model on a large dataset of images to identify and classify various types of lesions [11]. This can include detecting and characterizing tumors, identifying and measuring anatomic structures, and determining the presence of certain disease states.

### **3.1 How is AI utilized in radiology?**

AI in radiology utilizes the expertise of experienced radiologists to supply predefined criteria for properly programming the algorithm [4, 8]. Radiologists with specialized knowledge in chest, abdominal, or musculoskeletal radiology can offer the essential insight and direction required to train AI algorithms to identify and locate specific structures or anomalies on medical images. This involves providing the algorithm with a set of "ground truth" or baseline images that have been annotated by radiologists to indicate the presence and location of specific structures or abnormalities. The algorithm can then learn to identify these structures or abnormalities based on the patterns and features that are present in the images.

*Unlocking the Potential of Artificial Intelligence (AI) for Healthcare DOI: http://dx.doi.org/10.5772/intechopen.111489*

AI algorithms can also learn from a large volume of data with supervised or unsupervised strategies. Supervised learning is when the algorithm is provided with labeled data, where each image is associated with a specific diagnosis or label [4, 12]. This allows the algorithm to learn from the data and make predictions about new images. Unsupervised learning is when the algorithm is provided with unlabeled data, where the algorithm must learn to identify patterns and features in the data without any prior knowledge [4, 12]. This can be useful for identifying new or previously unknown patterns or features in the data. The algorithm can also be trained to extract information *via* patterns and share deep insights that can be used to improve diagnostic accuracy, efficiency, and decision-making [4].
