The current standard treatment protocol for glioblastoma is surgical resection, followed by 6 weeks of radiation therapy plus concomitant temozolomide chemotherapy (CCRT) and 6 cycles of adjuvant temozolomide chemotherapy. A significant challenge post-CCRT is the presence of radiation-induced side effects, such as pseudoprogression (PsP), generally defined radiologically as a new or enlarging area(s) of lesion(s) occurring early after the end of radiotherapy, which subsides or stabilizes without a change in therapy in the absence of true tumor growth (tumor progression, PD). Enlarged enhancing lesions on conventional MR images may represent PsP in up to 46.8%–64% of cases. The difficulty in distinguishing PD from PsP impedes clinical decision making in the treatment of patients, and numerous attempts have been made for discrimination with non-invasive imaging-based techniques. 

Anatomical MRI is the current standard for imaging evaluation of GBM for diagnosis and measurement of response in both clinical practice and clinical trials. Post-treatment response is then evaluated by criteria, the most commonly used examples of which being the Macdonald criteria (1990) and the RANO criteria (2010). 

Literature were collected through sources referenced by reviews, as well as PubMed and Google Scholar databases using combinations of search terms including “pseudoprogression,” “progression,” “glioblastoma,” “imaging,” “MRI,” “PET,” and “machine learning.”

This review included nine studies and one meta-analysis; two studies focused on conventional MRI techniques, three studies focused on advanced MRI techniques, two studies focused on amino acid PET imaging, and two studies focused on machine learning techniques. The meta analysis included thirty-five studies on various MRI techniques. Predictive components, imaging techniques, size, ground truth, and predictive ability were all evaluated for each of the nine studies and placed in a table. For the meta-analysis, studies were split into type, and size and results were put into a table. Results were varied; the meta analysis found that spectroscopy and perfusion MRI had the highest diagnostic accuracy. 

The review suggests significant potential in advanced MRI, PET imaging, and ML in developing techniques for distinguishing between PD and PsP, but does not define a singular best method.

Therefore, there is still a need for a clinically validated and accessible technique. The meta-analysis suggests large, multicenter, longitudinal prospective trials to help determine the best method of differentiation. Furthermore, imaging comes with many limitations, and thus alternative differentiators should be explored.