Editorial: Advancing cancer imaging technologies: bridging the gap from research to clinical practice

Imaging plays a crucial role in the management of cancer, from screening and diagnosis through staging, treatment planning, and follow-up to assess response. The imaging landscape is rapidly evolving, with functional imaging and deep learning now being incorporated into clinical practice. With advances in imaging, it is equally essential that these be appropriately applied in clinical practice. In the era of precision medicine, these developments have the potential to revolutionize treatment. Accurate diagnosis, characterization, and monitoring of response are core tenets of cancer management. Imaging relies on visual assessment, which can be enhanced substantially by computational analysis and multilayered data. Artificial intelligence can interpret qualitative data in detail to predict tumour genotype and biological course from radiographic phenotype and clinical outcome. 1,2 It can identify patterns which the human eye may not capture. Based on these patterns, it not only detects lesions early but also provides insight into the subtype, aggressiveness, mutational status and stage. 3,4 However, it is limited by the absence of a standardized protocol and clear guidelines. It is crucial to develop an algorithm using balanced, well-labelled data, and to validate it for reproducibility and generalizability to improve clinical applicability. 5 Additionally, ethical considerations must be clearly defined to foster collaboration and research. 6 This issue presents articles on the role of artificial intelligence in oncology and the expanding use of imaging in routine clinical practice.Functional imaging has garnered interest due to its non-invasive, quantitative nature for studying tumour physiology, molecular processes, tumour molecules, and metabolites. It can complement structural imaging to improve detection and diagnosis and to assess response. It surpasses traditional imaging by deciphering tumour gene expression and tracking cells and therapeutic drugs. 7 Functional imaging has also gained importance as it can visualize, characterize, and quantify biological processes at the cellular, subcellular, and molecular levels.Molecular functional imaging helps in the visualisation and quantification of visualize and quantify biochemical and physiological processes during tumorigenesis. Radiogeomics, the integration of multifunctional imaging with genomics, further helps elucidate tumour genetics. 8,9 This is particularly relevant in the era of targeted therapy, where the molecular imaging-based probes can target cell signalling pathways. 10 It is a crucial tool for studying newer receptor-targeted and gene therapies, as well as for cancer stem cell research, which requires the development of biotechnology, information technology, and basic translational science to achieve optimal outcomes. 11 The collection covers articles on various aspects of functional and molecular imaging and their application in clinical practice.Radiomics can extract high-dimensional quantitative features from routine imaging, which not only helps in diagnosis and risk stratification. 12 It also supports molecular characterization and outcome prediction. It has been explored across various tumour subsites, including pediatric brain tumours, to identify tumour subtypes, grading, molecular subgroups, and mutations. 13,14 Deep-learning-based radiogenomic (DLR) models and radiomic signatures have been used to identify the EGFR mutation status of non-small cell lung cancer from CT imaging. 15 These developments in imaging have the potential to boost clinical decision-making in the era of precision medicine.The collection has articles on radiomics in oncology and its clinical implications. Despite these strides in development in artificial intelligence and radiomics, there is a need to overcome translational barriers to bridge the gap from research to clinical implementation. There are several practical barriers, such as sustainability and system maintenance requiring updates and recalibrations, a lack of standardization and external validation, particularly in algorithms developed on a single institutional retrospective dataset, and regulatory challenges with authorities still defining the approval process for adaptive algorithms. 12,16,17 The issue is further compounded by a lack of robust data on cost-effectiveness and the need for high initial investments in digital infrastructure, personnel training, and information technology setup. 12 16 In addition, there are ethical concerns about data privacy as AI use expands, necessitating advances in data encryption. The models and algorithms are developed on the training dataset.The training dataset may not cover the entire spectrum of the patient population due to underrepresentation of racial, ethnic, or socioeconomic groups, and hence may not be suitable for this cohort of patients. There may be reluctance among clinicians to use AI in clinical practice, as it operates like a black box: it does not provide a rationale for its predictions. 16, 17 12 There is a need to overcome these limitations to translate the development into the day-today clinical practice.Overall, tThe research topic presents a rich collection of imaging articles that cover conventional and advanced imaging techniques and the evolving role of deep learning in cancer radiology. The collection highlights the multitude of imaging protocols with the potential to impact the clinical practice of cancer management. The section below provides brief glimpses of these articles. Pituitary neuroendocrine tumours are common intracranial tumours, and a small percentage of these tumours may be invasive and infiltrate the surrounding structures. It is crucial to identify the invasion to provide appropriate counselling and surgical planning. Yang et al analyzed data from 133 patients who underwent surgery and had histopathology, and who also underwent multiparameter magnetic resonance imaging (mpMRI) preoperatively. The authors selected 10 radiomics features from 306 primitive features and evaluated the predictive performance of single-and combined-radiomics models for cavernous sinus invasion.Ultrasound, Magnetic Resonance Imaging, and Contrast-Enhanced Mammography (CEM) with histopathological confirmation via nipple-areolar complex (NAC) punch biopsies in 2 cases of Paget's disease. The authors emphasized the diagnostic ability of CEM in this scenario.Alzheimer's disease is a neurodegenerative disorder that leads to dementia. It is essential to identify it before the onset of frank symptoms, when the irreversible changes have already Algorithms have been developed to detect lung nodules in computed tomography scans. 18 It is crucial to distinguish malignant from benign lung nodules to ensure appropriate management.Low-dose CT (LDCT) is used for screening in asymptomatic individuals to detect lung nodules.However, it has a limitation in differentiating malignant from benign nodules. Chen et al developed a diagnostic model based on the general condition of the patient, MP outputs from the AI system, the Mayo Clinic model, and serum biomarkers. Data from 260 participants were analyzed, of whom 72 had benign lung nodules, and 188 had early malignant lung nodules.The authors developed a nomogram and conducted a decision curve analysis to construct the model, using 60% of the cohort for training and 40% for validation.Literature suggests that tumour angiogenesis is an independent prognostic factor for cervical cancer that impacts recurrence, disease-free survival and overall survival. Microvessel density assessment by immunohistochemistry is the gold standard for evaluating tumour angiogenesis.It has limitations, including its invasive nature and tumour heterogeneity, which affect characterization and, in cases treated non-surgically. Zhu et al compared quantified superb microvascular imaging and quantified contrast-enhanced ultrasonography for assessing vascularization in operable versus non-operable uterine cervical cancer.Overall, this collection is an enriching mix of articles on advancing cancer imaging technologies, which have strong clinical implications and the potential to impact decisions in cancer management. This collection features diverse articles showcasing cutting-edge cancer imaging technologies and research with direct relevance to everyday clinical practice.The technological advancement is the future of medicine. It can play a pivotal role in diagnostic imaging and radiomics, treatment planning and assessing treatment response. It can analyze vast patient data and support predictive analytics, which has the potential to change the face of oncology in the era of personalized and precision medicine. AI can also help expand the horizon of drug development by integrating multifunctional imaging with genomics to elucidate tumour genetics in the era of targeted therapy, where molecular imaging-based probes can target cell signaling pathways. This can have a profound effect in oncology. However, efforts need to focus on overcoming translational limitations and challenges to harness the full potential of these technological advances.