Prior radiomics studies have focused on two-class brain tumor classification, which limits generalizability. The performance of radiomics in differentiating the three most common malignant brain tumors (glioblastoma (GBM), primary central nervous system lymphoma (PCNSL), and metastatic disease) is assessed; factors affecting the model performance and usefulness of a single sequence versus multiparametric MRI (MP-MRI) remain largely unaddressed. This retrospective study included 253 patients (120 metastatic (lung and brain), 40 PCNSL, and 93 GBM). Radiomic features were extracted for whole a tumor mask (enhancing plus necrotic) and an edema mask (first pipeline), as well as for separate enhancing and necrotic and edema masks (second pipeline). Model performance was evaluated using MP-MRI, individual sequences, and the T1 contrast enhanced (T1-CE) sequence without the edema mask across 45 model/feature selection combinations. The second pipeline showed significantly high performance across all combinations (Brier score: 0.311–0.325). GBRM fit using the full feature set from the T1-CE sequence was the best model. The majority of the top models were built using a full feature set and inbuilt feature selection. No significant difference was seen between the top-performing models for MP-MRI (AUC 0.910) and T1-CE sequence with (AUC 0.908) and without edema masks (AUC 0.894). T1-CE is the single best sequence with comparable performance to that of multiparametric MRI (MP-MRI). Model performance varies based on tumor subregion and the combination of model/feature selection methods.

BACKGROUND AND PURPOSE: To evaluate the radiomics-based model performance for differentiation between glioblastoma (GB) and brain metastases (BM) using magnetization prepared rapid gradient echo (MPRAGE) and volumetric interpolated breath-hold examination (VIBE) T1-contrast enhanced sequences. MATERIALS AND METHODS: T1-CE MPRAGE and VIBE sequences acquired in 108 patients (31 GBs and 77 BM) during the same MRI session were retrospectively evaluated. Post standardized image pre-processing and segmentation, radiomics features were extracted from necrotic and enhancing tumor components. Pearson correlation analysis of radiomics features from tumor subcomponents was also performed. A total of 90 machine learning (ML) pipelines were evaluated using a five-fold cross validation. Performance was measured by mean AUC-ROC, Log-loss and Brier scores. RESULTS: A feature-wise comparison showed that the radiomic features between sequences were strongly correlated, with the highest correlation for shape-based features. The mean AUC across the top-ten pipelines ranged between 0.851-0.890 with T1-CE MPRAGE and between 0.869-0.907 with T1-CE VIBE sequence. Top performing models for the MPRAGE sequence commonly used support vector machines, while those for VIBE sequence used either support vector machines or random forest. Common feature reduction methods for top-performing models included linear combination filter and least absolute shrinkage and selection operator (LASSO) for both sequences. For the same ML-feature reduction pipeline, model performances were comparable (AUC-ROC difference range: [-0.078, 0.046]). CONCLUSIONS: Radiomic features derived from T1-CE MPRAGE and VIBE sequences are strongly correlated and may have similar overall classification performance for differentiating GB from BM. ABBREVIATIONS: BM: Brain metastases, GB: glioblastoma, T1-CE: T1 contrast enhanced sequence, MPRAGE: magnetization prepared rapid gradient echo, ML: machine learning, RF: random forest, VIBE: volumetric interpolated breath-hold examination.

Quantum image processing is a growing field attracting attention from both the quantum computing and image processing communities. We propose a novel method in combining a graph-theoretic approach for optimal surface segmentation and hybrid quantum-classical optimization of the problem-directed graph. The surface segmentation is modeled classically as a graph partitioning problem in which a smoothness constraint is imposed to control surface variation for realistic segmentation. Specifically, segmentation refers to a source set identified by a minimum s-t cut that divides graph nodes into the source (s) and sink (t) sets. The resulting surface consists of graph nodes located on the boundary between the source and the sink. Characteristics of the problem-specific graph, including its directed edges, connectivity, and edge capacities, are embedded in a quadratic objective function whose minimum value corresponds to the ground state energy of an equivalent Ising Hamiltonian. This work explores the use of quantum processors in image segmentation problems, which has important applications in medical image analysis. Here, we present a theoretical basis for the quantum implementation of LOGISMOS and the results of a simulation study on simple images. Quantum Approximate Optimization Algorithm (QAOA) approach was utilized to conduct two simulation studies whose objective was to determine the ground state energies and identify bitstring solutions that encode the optimal segmentation of objective functions. The objective function encodes tasks associated with surface segmentation in 2-D and 3-D images while incorporating a smoothness constraint. In this work, we demonstrate that the proposed approach can solve the geometric-constrained surface segmentation problem optimally with the capability of locating multiple minimum points corresponding to the globally minimal solution.

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) uses radioactive contrast agents as a tracer to provide tumor morphology and contrast kinetics information of tumor regions, which are crucial in breast cancer diagnosis and treatment. The effectiveness of the imaging modality relies on its capacity to acquire dynamic data at a sufficient sampling rate to gain the desired temporal resolution. This can be achieved by sparsely sampling the k-space data and applying advanced image reconstruction method that exploits compressed sensing, such as the recently proposed GenSeT (Generalized Series with Temporal constraint) method. Due to the highly nonlinear nature of compressed-sensing-based approach, computational complexity of the reconstruction algorithm is a practical challenge, especially for large breast DCE-MRI datasets. In this study, the GenSeT algorithm was implemented in GPU using CUDA platform to significantly reduce reconstruction time, yielding a much more practical solution. Experimental results showed that for a breast DCE-MRI data, the proposed GPU-based GenSeT implementation achieved approximately 48 times faster in the reconstruction time as compared to the CPU approach, without sacrificing the image quality. Although this work focuses on accelerating image reconstruction for sparsely-sampled breast DCE-MRI, the proposed GPU-based algorithm can be easily applied for sparsely sampled DCE-MRI of other organs.