Abstract
241243
Introduction: In the field of Molecular Imaging, there is more to it than what meets the eye. What the human eyes of Nuclear Physicians cannot perceive – Radiomics can. Radiomics provides insightful information through quantification of imaging features, which potentially corelate with clinical and genetic data.
We aim to summarise the current applications of Radiomics in Molecular Imaging, whilst highlighting and focusing on its future potential applications for the Nuclear Medicine Community.
Methods: In this study, we share about the concept behind Radiomics (an emerging field in Artificial Intelligence) and summarise the current clinical applications of Radiomics in Molecular Imaging from current literature.
The following key concepts behind Radiomics are considered:
• Radiomics’ strength of feature extraction
• Multi-layer considerations behind feature extraction: multiple orders ranging for first-order spatial analysis to fourth-order convolutional neural networks
• Various means of feature extraction, including advanced AI segmentation algorithms
The current clinical applications of Radiomics are summarised as such:
• Detecting patterns through statistical analysis
• Studying correlations between imaging patterns and disease behaviour
• Prognostication and tailoring management
We then highlight how the Nuclear Medicine Community can leverage on this emerging technology for future potential applications.
Results: There is great promise for the Nuclear Medicine community to harness the potential for Radiomics in molecular imaging services. The key potential applications are as summarized:
1) Detecting patterns through statistical analysis
The application of Radiomics largely involves various fields within Oncology, such as head and neck cancers, lung cancers, breast cancers, liver cancers and neuroendocrine tumors etc. The Positron Emission Tomography (PET) images of these cancers are studied with Radiomics, and various patterns are picked out to better characterize lesions to aid in diagnosis. There is great potential for this to be further developed to diagnose cancers and their subtypes with greater accuracy.
2) Studying correlations between imaging patterns and disease behaviour
Furthermore, Radiomics has allowed for improved analysis of treatment response of oncotherapy. In the field of Immunotherapy for lung cancer, the ability for Radiomics to predict treatment response has been studied. There remains good potential for this to be further expanded to various types of oncotherapy, and not only immunotherapy.
3) Prognostication and tailoring management
Perhaps one of the greatest achievement of Radiomics is not only its ability to aid in diagnosis and assessment of disease response, but also its ability to prognosticate and influence management.
In the context of lymphoma, Radiomics has helped in the assessment of tumor heterogeneity, which has assisted with prognostication (taking into account other factors such as the biological and metabolic status of the underlying disease).
In fact, Radiomics has not only prognosticated, but also enabled clinicians to tailor therapeutic options according to imaging features. One such example is its application in guiding radiation oncology therapy, simply from analysing pre-treatment imaging to decide on dose planning.
There is immense potential for Radiomics to prognosticate diseases with greater granularity, and guide clinical management with more confidence.
Conclusions: The field of Artificial Intelligence has revolutionized the interpretation of various imaging modalities. Of note, Radiomics has allowed for the discovery of novel processes through its unique ability to pick out patterns that better characterize disease processes, thereby influencing prognosis and management significantly. This has proven great benefit to modern day Molecular Imaging, and has great potential for the future of our Nuclear Medicine Community.