Search

BACKGROUND AND PURPOSE: The World Health Organization has recently placed new emphasis on the integration of genetic information for gliomas. While tissue sampling remains the criterion standard, noninvasive imaging techniques may provide complimentary insight into clinically relevant genetic mutations. Our aim was to train a convolutional neural network to independently predict underlying molecular genetic mutation status in gliomas with high accuracy and identify the most predictive imaging features for each mutation. MATERIALS AND METHODS: MR imaging data and molecular information were retrospectively obtained from The Cancer Imaging Archives for 259 patients with either low- or high-grade gliomas. A convolutional neural network was trained to classify isocitrate dehydrogenase 1 (IDH1) mutation status, 1p/19q codeletion, and O6-methylguanine-DNA methyltransferase (MGMT) promotor methylation status. Principal component analysis of the final convolutional neural network layer was used to extract the key imaging features critical for successful classification. RESULTS: Classification had high accuracy: IDH1 mutation status, 94%; 1p/19q codeletion, 92%; and MGMT promotor methylation status, 83%. Each genetic category was also associated with distinctive imaging features such as definition of tumor margins, T1 and FLAIR suppression, extent of edema, extent of necrosis, and textural features. CONCLUSIONS: Our results indicate that for The Cancer Imaging Archives dataset, machine-learning approaches allow classification of individual genetic mutations of both low- and high-grade gliomas. We show that relevant MR imaging features acquired from an added dimensionality-reduction technique demonstrate that neural networks are capable of learning key imaging components without prior feature selection or human-directed training.

The present effort has led to the development of a numerical procedure of the stage flow field based upon solution of the Navier-Stokes equations. The specific case considered is the basic case in which two-dimensional flow and equal rotor and stator pitch are assumed. The procedure developed is capable of obtaining periodic solutions for a grid of 7500 points within ten CPU minutes of Cray run time. This represents a very efficient technique which will permit runs of this type on a regular basis, and which will allow extension to either three dimensions or to stages in which rotor and stator have unequal pitch and the multiple stages. The results obtained show the qualitatively expected features. Periodicity was obtained essentially within five cycles. The major effect of the interaction to the specific case considered appeared in the leading edge region of the downstream rotor blade. Significant perturbation pressures were noted, Delta Cp approx. = 0.4, which has major implications regarding unsteady loading and unsteady fluid structure interaction. The concentration of the interaction effect in the leading edge has particular significance in regard to unsteady pitching moment. Although less significant, unsteady effects were noted over the aft portion of the trailing blade, as well as over the entire leading blade.

An important problem in axial flow turbomachinery is that of flow through a turbine or compressor stage in which interaction occurs between the rotor and the stator. Although several experimental efforts have focused upon this problem, relatively little analysis has been applied to this problem. The present effort applies a Navier-Stokes analysis to the rotor/stator problem. The effort consists of several tasks. These include assessment of an existing Navier-Stokes analysis for time-dependent flows, exploration of appropriate boundary conditions, development of an appropriate coordinate system and a demonstration calculation. The present annual report discusses the time-dependent assessment, exploration of boundary conditions and the coordinate system problem.