Editorial: Towards Precision Oncology: Assessing the Role of Radiomics and Artificial Intelligence

Salvatore Claudio Fanni^1*Damiano Caruso²Lorenzo Faggioni¹Emanuele Neri¹Dania Cioni¹

• ¹Department of Translational Research, Academic Radiology, University of Pisa, Pisa, Italy
• ²Universita degli Studi di Roma La Sapienza Dipartimento di Scienze Medico-Chirurgiche e Medicina Translazionale, Rome, Italy

Another paper investigating the potential role of radiomics in di erential diagnosis, in a di erent clinical scenario then lung cancer, is the one published by Hu et al. The authors investigated the potential role of a Rad-score to discriminate esophageal sarcomatoid carcinoma and esophageal squamous carcinoma, and found a 0.828 (95% CI 0.636-1.000) in the validation cohort. Radiomics may also provide assistance in the characterization of tumoral lesions, which is of paramount importance to move radiology forward to precision oncology. Peng et al. evaluated the performance of a clinical-radiomics model to predict HER2 status in urothelial bladder carcinoma from contrast-enhanced CT images. Such clinical-radiomics model achieved an AUC of 0.85, and decision curve analysis indicated that the clinical-radiomics model provided good clinical benefit. Zhang et al. extracted radiomics from ultrasound images to predict tumor infiltrating lymphocyte levels in breast cancer. In comparison to grayscale ultrasound model, and radiomics model, the nomogram integrating both demonstrated superior discriminative ability on both the training (AUC 0.88) and testing (AUC 0.82) set. Ye et al. moved the field a step forward, as they compared a conventional radiomics model and a tumor internal heterogeneity habitat model in predicting pathologic complete response to neoadjuvant immunochemotherapy in non-small cell lung cancer. As a result, the habitat model exhibited higher AUC values compared to the conventional radiomics model, achieving an AUC of 0.78 and 0.72, respectively. Moreover, although radiomic features are often described as imperceptible to the human eye, this does not imply that their conversion into colorimetric maps could not aid image interpretation and lesion detection. Hertel et al. used radiomics to improve detection of colorectal liver metastases on images and found that the feature map for firstorder RootMeanSquared was ranked superior in terms of very high visual contrast in 57.4% of cases, compared to 41.0% in standard reconstructions. This very heterogeneous research topic also included papers investigating di erent clinical settings, such as the one by Kruzhilov et al., investigating the role of AI in whole-body PET imaging denoising, and the on by Tang et al., exploring the transformative application of the metaverse in nuclear medicine. These studies demonstrate that the integration of radiomics and artificial intelligence represents a promising pathway to transform oncologic imaging into a quantitative and decision-support tool: from pulmonary nodules to molecular biomarkers, and toward new modalities for data visualization and optimization. However, to consolidate this transition into clinical practice, investment is needed in multicenter studies, standardized pipelines, model transparency, and interpretative interfaces. Such an approach could support the sustainable implementation of clinical AI and open new frontiers, including the educational and collaborative metaverse, in precision oncology.