Glioma Adaptive Radiotherapy With Development of an Artificial Intelligence Workflow

Glioblastoma multiforme (GBM) represents grade 4 diffuse gliomas accounting for the most common primary malignant central nervous system (CNS) tumors in adults . GBM is treated with radiotherapy (RT) and concurrent chemotherapy following maximal safe resection, with a median survival of approximately 15-18 months . GBM harbors significant intratumoral heterogeneity with areas of multiclonal and hypoxic areas rendering higher chances of disease relapse following standard RT .

Similarly, distinct compartments can be well appreciated on magnetic resonance imaging (MRI): enhancing tumor core (TC) with central necrotic areas and the peritumoral region (PTR), which consists of microscopic tumor infiltration and vasogenic edema . Similar to regions of radioresistant areas within the TC, the microscopic disease in the PTR plays a vital role in disease relapse . Other grade 2 and 3 diffuse gliomas include isocitrate dehydrogenase (IDH) mutant astrocytoma and oligodendroglioma . In the recent World Health Organization (WHO) classification of CNS tumors, molecular information is combined with histopathological information for integrated classification. IDH-wildtype tumors are further molecularly characterized and considered as GBM since the prognosis is shown to be dismal. Oligodendrogliomas are confirmed based on the presence of deletion of 1p19q chromosomal arms. Grade 2/3 diffuse gliomas are typically seen as tumors with T2-weighted hyperintense tumors. The treatment is similar to GBM, with maximal safe resection followed by radiation and concurrent and adjuvant chemotherapy.

Radiotherapy for Diffuse Gliomas Radiotherapy (RT) forms an integral role in the multimodality management of diffuse gliomas . Radiation is indicated in low-grade gliomas with high-risk features or high-grade gliomas following maximal safe resection . The radiation (RT) in diffuse gliomas in GBM is delivered using conformal techniques to the residual disease and cavity, called the gross tumor volume (GTV). The surrounding area is included in the clinical target volume (CTV) to treat areas of microscopic disease. For GBM, an expansion of 1.5 -2 cm is done from the GTV, which is identified as an enhancing area on T1c sequences to include areas of PTR (T2w hyperintensity) and edited from anatomical barriers like meninges and dural reflections . For IDH-mutant gliomas, the residual tumor and cavity (identified as T2w hyperintensity region) are included as GTV and further expansion of 5-10 mm is done to be included as CTV . The standard practice involves delivering a uniform dose of radiation to the planning target volume (PTV), which encompasses an isotropic margin expansion surrounding the CTV to account for set-up uncertainties. In GBM and IDHmutant high-grade astrocytoma, the total dose of 59.4-60 Gy is delivered over 6-7 weeks with 1.8-2 Gy per fraction. A relatively lower dose of radiation, in the range of 54.0-59.4 Gy, is delivered over 6-7 weeks using 1.8-2 Gy per fraction for oligodendrogliomas. As per the current paradigm, radiation is planned on computed tomography (CT) for dose computation and MRI for visualization of target volumes and organs at risk (OAR), done once before treatment, based on which fractionated radiation is delivered over 6- 7 weeks. Recent evidence with MRI undertaken during the course of treatment has demonstrated the changes in dynamics of the residual disease, surgical cavity, and also the OARs in a proportion of patients, suggesting that treatment is delivered based on imaging at a single time-point can lead to inaccuracies .Therefore, adaptive radiotherapy (ART) to modify radiation plans based on the spatial changes of the target volume and OAR is likely to improve the accuracy of RT treatments. Also, serial imaging during treatment can be used to identify areas of tumor or PTR showing refractory disease or vasogenic edema, with provisions for biological modifications of RT doses . The use of conformal radiation techniques like intensity-modulated radiotherapy (IMRT), volumetric modulated arc therapy (VMAT) can enable delivery of differential radiation doses precisely to different areas of the target volume, known as dose painting .Positron Emission Tomography (PET) Functional imaging with positron emission tomography (PET) has attained wide popularity in oncology in disease staging, identifying hypoxic areas, and guiding radiation planning . For gliomas, amino acid PET like O-(2-[18F] fluoroethyl) -L-tyrosine (FET) or Fluorodopa (F-DOPA) has been proven effective with areas of a higher tumor to white matter ratio, suggestive areas of active disease (19) . The use of PET scans during treatment can help identify areas refractory to RT, reflected by higher uptake of the radioisotope. Higher doses to such regions provide a window for biological adaption and can potentially improve control rates. Similarly, quantitative analysis of imaging (more popularly known as radiomics) can help differentiate areas of microscopic tumor from vasogenic edema in the PTR, which otherwise appears similar to conventional imaging . Artificial Intelligence The role of artificial intelligence in oncology is increasingly recognized, ranging from optimization of healthcare resource utilization and decision-making to quantitative image analysis for prognostication and the potential ability to serve as a noninvasive biomarker . The practice of contemporary radiation oncology heavily relies on the interaction of humans and machines in almost every treatment planning process, including contouring of target volumes, OAR, treatment-planning processes, and during treatment delivery. Creating an artificial intelligence framework in radiation oncology promises to improve workflow efficiency and accuracy and enable treatment planning and delivery rapidly and efficiently. The use of adaptive radiotherapy will be further facilitated using machine learning algorithms with appropriate identification of patients to be benefitted from volumetric or biological adaptation, autosegmentation of target/OAR, automated treatment planning, and biological modification based on spatial and temporal changes of quantitative imaging parameters. Standard institutional practice The standard institutional practice includes a dose of 59.4 Gy in 33 fractions over 6.5 weeks for patients with glioblastoma and 55.8 Gy in 31 fractions over 6 weeks for patients with oligodendroglioma. Concurrent temozolomide is used for all patients undergoing radiation at the dose of 75 mg/m 2 of body surface area during the course of radiation with weekly monitoring on blood counts. All radiation treatments are planned based on single time CT and MRI scan without any scheduled interval scans during radiation, and no adaptation is done. Adjuvant chemotherapy with temozolomide is started after 4 weeks of radiation completion at dose of 150 mg/m 2 for five days and repeated on monthly basis and dose escalated to 200 mg/m 2 if tolerating well and normal blood counts. As standard practice 6 and 12 cycles of temozolomide are scheduled for GBM and IDH-mutant glioma (astrocytoma and oligodendroglioma) respectively. After treatment completion clinical follow-up is scheduled every 3-6 months in the first 2 years and thereafter every 6-12 months for all the patients. Surveillance imaging is scheduled every 6-12 months in the first 5 years and thereafter on annual basis or interval imaging undertaken as prompted clinically.