Improved Glioma Grading Using Deep Convolutional Neural Networks.

BACKGROUND AND PURPOSE: Accurate determination of glioma grade leads to improved treatment planning. The criterion standard for glioma grading is invasive tissue sampling. Recently, radiomic features have shown excellent potential in glioma-grade prediction. These features may not fully exploit the underlying information in MR images. The objective of this study was to investigate the performance of features learned by a convolutional neural network compared with standard radiomic features for grade prediction. MATERIALS AND METHODS: A total of 237 patients with gliomas were included in this study. All images were resampled, registered, skull-stripped, and segmented to extract the tumors. The learned features from the trained convolutional neural network were used for grade prediction. The performance of the proposed method was compared with standard machine learning approaches, support vector machine, random forests, and gradient boosting trained with radiomic features. RESULTS: The experimental results demonstrate that using learned features extracted from the convolutional neural network achieves an average accuracy of 87%, outperforming the methods considering radiomic features alone. The top-performing machine learning model is gradient boosting with an average accuracy of 64%. Thus, there is a 23% improvement in accuracy, and it is an efficient technique for grade prediction. CONCLUSIONS: Convolutional neural networks are able to learn discriminating features automatically, and these features provide added value for grading gliomas. The proposed framework may provide substantial improvement in glioma-grade prediction; however, further validation is needed.

The influence of a bupivacaine and fentanyl epidural infusion after epidural fentanyl in patients allowed to ambulate in early labor.

UNLABELLED: Epidural fentanyl after a lidocaine and epinephrine test dose provides adequate analgesia and allows for ambulation during early labor. This study was designed to determine the influence of an epidural infusion of bupivacaine plus fentanyl administered after initiation of epidural labor analgesia with fentanyl. Specifically, we evaluated whether there is an increase in motor block or an increased time to request for further analgesic medication. Fifty-one laboring primigravid women at <5 cm cervical dilation who requested epidural analgesia were enrolled. After a 3-mL epidural test dose of 1.5% lidocaine with epinephrine (5 microg/mL), patients received fentanyl 100 microg via the epidural catheter. They then randomly received either an infusion (10 mL/h) of 0.0625% bupivacaine with fentanyl (3 microg/mL) or an infusion of preservative-free saline. After the administration of the initial analgesic, pain scores and side effects were recorded for each patient at 10, 20, and 30 min, every 30 min thereafter, and at the time of request for additional analgesic medication, by an observer blinded to the technique used. There were no demographic differences between the two groups. The mean duration of analgesia (time from initial dose to request for additional analgesia) was increased in the group that received a continuous infusion of bupivacaine and fentanyl compared with the Saline group (198 +/- 86 vs 145 +/- 50 min; P < 0.009). Side effects were similar between the two groups. No patient in either group experienced any detectable motor block. Fourteen patients chose to ambulate in the Saline group, and 12 patients chose to ambulate in the Infusion group. In early laboring patients, a continuous infusion of 0.0625% bupivacaine infusion with fentanyl (3 microg/mL) prolonged the duration until top-up was required, after epidural fentanyl 100 microg after a lidocaine and epinephrine test dose, and did not cause any clinically detectable motor block. IMPLICATIONS: A 0.0625% bupivacaine and fentanyl (3 microg/mL) infusion, when added to epidural fentanyl (100 microg), prolongs the analgesic duration without increasing motor block in women in early labor.

Mixture of experts for classification of gender, ethnic origin, and pose of human faces.

In this paper we describe the application of mixtures of experts on gender and ethnic classification of human faces, and pose classification, and show their feasibility on the FERET database of facial images. The FERET database allows us to demonstrate performance on hundreds or thousands of images. The mixture of experts is implemented using the "divide and conquer" modularity principle with respect to the granularity and/or the locality of information. The mixture of experts consists of ensembles of radial basis functions (RBFs). Inductive decision trees (DTs) and support vector machines (SVMs) implement the "gating network" components for deciding which of the experts should be used to determine the classification output and to restrict the support of the input space. Both the ensemble of RBF's (ERBF) and SVM use the RBF kernel ("expert") for gating the inputs. Our experimental results yield an average accuracy rate of 96% on gender classification and 92% on ethnic classification using the ERBF/DT approach from frontal face images, while the SVM yield 100% on pose classification.

Controlled delivery systems for proteins using polyanhydride microspheres.

A method to provide near-constant sustained release of high molecular weight, water-soluble proteins from polyanhydride microspheres is described. The polyanhydrides used were poly(fatty acid dimer) (PFAD), poly(sebacic acid) (PSA), and their copolymers [P(FAD-SA)]. P(FAD-SA) microspheres containing proteins of different molecular sizes--lysozyme, trypsin, heparinase, ovalbumin, albumin, and immunoglobulin--were prepared by a solvent evaporation method using a double emulsion. The microspheres containing proteins were spherical, with diameters of 50-125 microns, and encapsulated more than 80% of the protein, irrespective of the protein used. Enzymatic activity studies showed that encapsulation of enzymes inside polyanhydride microspheres can protect them from activity loss. When not placed inside polyanhydride microspheres, trypsin lost 80% of its activity in solution at 37 degrees C at pH 7.4 in 12 hr, whereas inside the polyanhydride microspheres the activity loss was less than 10% under these conditions. About 47% of the enzymatic activity of heparinase encapsulated in the microspheres was lost at 37 degrees C in 24 hr, while in solution it lost over 90% of its activity. The protein-loaded microspheres displayed near-zero-order erosion kinetics over 5 days as judged by the release of sebacic acid (SA) from the microspheres. The microspheres degraded to form SA and FAD monomers. All proteins were released at a near-constant rate without any large initial burst, irrespective of polymer molecular weight and protein loading. The period of protein release was longer than that of SA and continued protein release was observed even after the microsphere matrix had completely degraded. Differential scanning calorimetric studies demonstrated an interaction between protein and the FAD monomers produced with microsphere degradation. It is likely that the protein interaction with FAD monomers permits formation of water-insoluble protein aggregates which slowly dissolve and diffuse out of the matrix, leading to delayed protein release. For trypsin-loaded microspheres, trypsin lost 40% of its activity during microsphere preparation. Activity studies demonstrated that the sonication process was primarily responsible for activity loss. A reduction in the period of ultrasound exposure decreased the loss of protein activity to around 20%.