Page 19 - ARNM-3-2

Advances in Radiotherapy
& Nuclear Medicine Role of PET/CT in exploring tumor heterogeneity

(ROI) in medical images. Ensuring the robustness and Machine learning algorithms and AI, when applied to
reproducibility of these features is paramount, as variations these radiomic features, can identify complex patterns and
in imaging protocols or segmentation methods can affect associations that may predict treatment outcomes or guide
the results. After feature extraction, optimal features are personalized therapy decisions. For example, CNNs have
selected, and AI-based models are incorporated to analyze demonstrated the ability to automatically segment tumors
the data. These models can predict clinical outcomes, such and classify them based on molecular subtypes with high
as treatment response or survival, based on the extracted accuracy. In addition, AI-driven evaluation of PET/CT
radiomic features. For instance, AI-driven radiomics scans combined with clinical information can generate
models have demonstrated promise in predicting responses predictive analytics forecasting treatment outcomes and
to immunotherapy and chemotherapy in patients with patient prognosis.
non-small cell lung cancer (NSCLC) and breast cancer. 60 These advanced techniques are paving the way for more

8.3. Applications in tumor heterogeneity precise and personalized cancer management, potentially
allowing for early identification of treatment resistance
Radiomics plays a significant role in characterizing and enabling adaptive treatment strategies. However, it is
tumor heterogeneity, which is a major challenge in cancer important to note that while these methods are promising,
management. By analyzing spatial and temporal variations they still require extensive validation in large, multi-center
in tracer uptake, radiomics can identify subregions within studies before widespread clinical implementation. The
tumors that exhibit different biological behaviors. This integration of AI and deep learning with conventional
information is crucial for personalized treatment planning, PET/CT analysis represents a major step toward unlocking
as it allows clinicians to target aggressive tumor regions the full diagnostic and predictive power of molecular
more effectively. For example, studies have demonstrated imaging in oncology. 61-63
that radiomic features derived from F-FDG PET/CT
18
scans can predict intratumoral heterogeneity and guide 10. Challenges and limitations of PET/CT
radiotherapy planning by identifying regions with high Future developments in PET/CT technology and the
metabolic activity. 14,59
expanding knowledge of tumor heterogeneity will
8.4. Future directions of radiomics undoubtedly improve the diagnosis, restaging, and response
prediction of new radiopharmaceuticals. However, the
The integration of radiomics with AI and machine application of PET/CT with new radiopharmaceuticals
learning is expected to further enhance the diagnostic in everyday clinical practice is also constrained by many
and predictive capabilities of PET/CT imaging. Advanced challenges:
algorithms, such as convolutional neural networks (i) Cost and accessibility: PET/CT scans are expensive,
(CNNs), can automatically segment tumors and classify and the cost can be prohibitive for many patients,
them based on molecular subtypes with high accuracy. especially in low-resource settings. The high cost
In addition, radiomics combined with multi-tracer of radiopharmaceuticals, coupled with the need
PET imaging (e.g., 18 F-FDG and 68 Ga-PSMA) can for specialized equipment and facilities, limits the
provide a more comprehensive understanding of tumor widespread availability of this technology. Addressing
heterogeneity, enabling more precise and personalized disparities in access to advanced PET imaging
cancer management. 59 technologies is crucial to ensure equitable healthcare
delivery.
13,15
9. AI and machine learning applications (ii) Time and workflow: PET/CT scans require significant
The field of PET/CT imaging is rapidly evolving, with preparation, imaging, and interpretation time.
radiomics and AI emerging as powerful tools for The synthesis of radiopharmaceuticals, patient
enhancing tumor characterization and treatment planning. preparation, and the scanning process itself can
Incorporating radiomics with AI can develop models that be time-consuming, which may delay treatment
will lead to detailed analysis of medical images. Radiomics decisions. In addition, the need for multiple scans to
offers a means to capture tumor heterogeneity beyond what assess tumor heterogeneity further complicates the
is visually apparent by the high-throughput extraction workflow. 13,17
of quantitative features from medical images. Texture (iii) Radiation exposure: PET/CT scans involve exposure
analysis, a key component of radiomics, can quantify to ionizing radiation, which raises concerns, especially
spatial variations in tracer uptake, potentially revealing for patients requiring frequent follow-up scans.
subtle patterns indicative of tumor aggressiveness or While the radiation dose is generally considered safe,
treatment resistance. cumulative exposure over time can increase the risk

Volume 3 Issue 2 (2025) 11 doi: 10.36922/ARNM025040005