RESUMO
BACKGROUND: MRI-Linac systems enable daily diffusion-weighed imaging (DWI) MRI scans for assessing glioblastoma tumor changes with radiotherapy treatment. PURPOSE: Our study assessed the image quality of echoplanar imaging (EPI)-DWI scans compared with turbo spin echo (TSE)-DWI scans at 0.35 Tesla (T) and compared the apparent diffusion coefficient (ADC) values and distortion of EPI-DWI on 0.35 T MRI-Linac compared to high-field diagnostic MRI scanners. METHODS: The calibrated National Institute of Standards and Technology (NIST)/Quantitative Imaging Biomarkers Alliance (QIBA) Diffusion Phantom was scanned on a 0.35 T MRI-Linac, and 1.5 T and 3 T MRI with EPI-DWI. Five patients were scanned on a 0.35 T MRI-Linac with a TSE-DWI sequence, and five other patients were scanned with EPI-DWI on a 0.35 T MRI-Linac and a 3 T MRI. The quality of images was compared between the TSE-DWI and EPI-DWI on the 0.35 T MRI-Linac assessing signal-to-noise ratios and presence of artifacts. EPI-DWI ADC values and distortion magnitude were measured and compared between 0.35 T MRI-Linac and high-field MRI for both phantom and patient studies. RESULTS: The average ADC differences between EPI-DWI acquired on the 0.35 T MRI-Linac, 1.5 T and 3 T MRI scanners and published references in the phantom study were 1.7%, 0.4% and 1.0%, respectively. Comparing the ADC values based on EPI-DWI in glioblastoma tumors, there was a 3.36% difference between 0.35 and 3 T measurements. Susceptibility-induced distortions in the EPI-DWI phantoms were 0.46 ± 1.51 mm for 0.35 MRI-Linac, 0.98 ± 0.51 mm for 1.5 T MRI and 1.14 ± 1.88 mm for 3 T MRI; for patients -0.47 ± 0.78 mm for 0.35 T and 1.73 ± 2.11 mm for 3 T MRIs. The mean deformable registration distortion for a phantom was 1.1 ± 0.22 mm, 3.5 ± 0.39 mm and 4.7 ± 0.37 mm for the 0.35 T MRI-Linac, 1.5 T MRI, and 3 T MRI scanners, respectively; for patients this distortion was -0.46 ± 0.57 mm for 0.35 T and 4.2 ± 0.41 mm for 3 T. EPI-DWI 0.35 T MRI-Linac images showed higher SNR and lack of artifacts compared with TSE-DWI, especially at higher b-values up to 1000 s/mm2. CONCLUSION: EPI-DWI on a 0.35 T MRI-Linac showed superior image quality compared with TSE-DWI, minor and less distortions than high-field diagnostic scanners, and comparable ADC values in phantoms and glioblastoma tumors. EPI-DWI should be investigated on the 0.35 T MRI-Linac for prediction of early response in patients with glioblastoma.
RESUMO
Deep learning (DL) is poised to redefine the way medical images are processed and analyzed. Convolutional neural networks (CNNs), a specific type of DL architecture, are exceptional for high-throughput processing, allowing for the effective extraction of relevant diagnostic patterns from large volumes of complex visual data. This technology has garnered substantial interest in the field of neuro-oncology as a promising tool to enhance medical imaging throughput and analysis. A multitude of methods harnessing MRI-based CNNs have been proposed for brain tumor segmentation, classification, and prognosis prediction. They are often applied to gliomas, the most common primary brain cancer, to classify subtypes with the goal of guiding therapy decisions. Additionally, the difficulty of repeating brain biopsies to evaluate treatment response in the setting of often confusing imaging findings provides a unique niche for CNNs to help distinguish the treatment response to gliomas. For example, glioblastoma, the most aggressive type of brain cancer, can grow due to poor treatment response, can appear to grow acutely due to treatment-related inflammation as the tumor dies (pseudo-progression), or falsely appear to be regrowing after treatment as a result of brain damage from radiation (radiation necrosis). CNNs are being applied to separate this diagnostic dilemma. This review provides a detailed synthesis of recent DL methods and applications for intratumor segmentation, glioma classification, and prognosis prediction. Furthermore, this review discusses the future direction of MRI-based CNN in the field of neuro-oncology and challenges in model interpretability, data availability, and computation efficiency.