Jonathan Rohrer

Memory-efficient distribution of regular expressions for fast deep packet inspection

Current trends in network security force network intrusion detection systems (NIDS) to scan network traffic at wirespeed beyond 10 Gbps against increasingly complex patterns, often specified using regular expressions. As a result, dedicated regular-expression accelerators have recently received considerable attention. The storage efficiency of the compiled patterns is a key factor in the overall performance and critically depends on the distribution of the patterns to a limited number of parallel pattern-matching engines. In this work, we first present a formal definition and complexity analysis of the pattern distribution problem and then introduce optimal and heuristic methods to solve it. Our experiments with five sets of regular expressions from both public and proprietary NIDS result in an up to 8.8x better storage efficiency than the state of the art. The average improvement is 2.3x.

A Parallel GPU algorithm for mutual information based 3D nonrigid image registration

Many applications in biomedical image analysis require alignment or fusion of images acquired with different devices or at different times. Image registration geometrically aligns images allowing their fusion. Nonrigid techniques are usually required when the images contain anatomical structures of soft tissue. Nonrigid registration algorithms are very time consuming and can take hours for aligning a pair of 3D medical images on commodity workstation PCs. In this paper, we present parallel design and implementation of 3D non-rigid image registration for the Graphics Processing Units (GPUs). Existing GPU-based registration implementations are mainly limited to intra-modality registration problems. Our algorithm uses mutual information as the similarity metric and can process images of different modalities. The proposed design takes advantage of highly parallel and multi-threaded architecture of GPU containing large number of processing cores. The paper presents optimization techniques to effectively utilize high memory bandwidth provided by GPU using on-chip shared memory and co-operative memory update by multiple threads. Our results with optimized GPU implementation showed an average performance of 2.46 microseconds per voxel and achieved factor of 28 speedup over a CPU-based serial implementation. This improves the usability of nonrigid registration for some real world clinical applications and enables new ones, especially within intra-operative scenarios, where strict timing constraints apply.

Accelerating 3D nonrigid registration using the cell broadband engine processor

Registration or alignment of medical images in clinical applications requires cost-effective high-performance computing. In this paper, we present a parallel design and implementation of a mutualinformation-based multiresolution nonrigid registration algorithm that takes advantage of the Cell Broadband Engine® (Cell/B.E.) Architecture by exploiting the different levels of parallelism and optimization strategies. The new method was tested with a dualprocessor Cell/B.E. processor-based system. The experiments show an average performance of 1.09 µs per voxel and excellent scalability, demonstrating real-time or near-real-time performance for the computationally demanding task of nonrigid image registration.

Anatomical object recognition and labeling by atlas-based focused non-rigid registration and region-growing

Computer-assisted recognition of anatomical objects in medical images is at the center of many important clinical applications. Automatic extraction and recognition of human abdominal structures from CT images has been particularly challenging for medical imaging research and applications. Intra-patient and inter-patient spatial, morphological and intensity variability is typically significantly contributing to great difficulties in developing satisfactory automatic solutions to the problem. In this paper we report on a new approach, which treats recognition of anatomical objects from a given medical image as a task of non-rigid registration followed by segmentation. It uses the knowledge inferred from an atlas or model image to specify a sequence of smaller sub-image spaces or spatial contexts to register progressively the atlas image with the given patient image. The labels of the target objects in the atlas image are carried over to the patient image by the registration process representing the recognition result, which is further improved by a region-growing process. Preliminary experiments of artery blood vessel recognition and labeling with real patient data have demonstrated the potential of the method to be a viable alternative solution to the problem.

Exploring the design space of programmable regular expression matching accelerators

State-of-the-art regular expression (regex) accelerators combine parallel programmable state machines with cascaded, wide-issue instruction processors to improve the storage efficiency and the processing rates, while preserving the programmability. The pattern-matching engine (PME) included on the IBM PowerEN(TM) (Edge-of-Network) processor is one such design, and can be used as an architectural template for a broad design-space exploration. The regex compiler is a key component of such an exploration, involving sophisticated transformations to map large sets of complex regexs to the memory contents and the configuration registers of the accelerator hardware. The design space is explored by varying the main microarchitectural parameters, including the memory size, the number of parallel state machines, and the parameters of the instruction processor. While the design-space exploration confirms the main architectural choices of the PME, it also shows that further optimization is possible by eliminating the bottlenecks in the instruction dispatch mechanisms, which results in an up to 50% reduction in the storage requirements. The design-space exploration utilizes a parameterizable and synthesizable hardware model to evaluate the effects the microarchitectural choices have on the chip area and operating frequency. The synthesis results demonstrate the scalability of the optimization chosen and the need to incorporate these choices into future regex accelerator architectures.

Focused atlas-based image registration for recognition

Registration with an atlas is a method to recognize and label structures that is used especially for the analysis of medical images. Its application to images containing many objects, such as images of the human abdomen, may prove to be very challenging. The underlying idea of our approach is to follow a strategy that resembles commonsense problem solving - goal-directed, context-based and focused. In our method, the goal of recognizing a particular anatomical structure, such as the aorta, is achieved by following a plan, which is a sequence of image registration tasks. The plan is derived by decomposing the registration of the goal object into a series of object registration subtasks using the anatomy knowledge and the modality-specific image properties. Preliminary experiments to recognize the aorta in human abdominal CT images have demonstrated the potential of our method.