We will develop an application emulator that will reproduce application characteristics found in many defense and high-end civilian applications that involve sensor data analysis, sensor data fusion and real time sensor data processing. We are focusing on emulating application scenarios that will be of practical relevance in a 5 to 15 year time frame.
One of the motivating applications (the Virtual Microscope) under development at Johns Hopkins involves development of software that makes it possible to carry out a realistic digital emulation of a high power light microscope. Raw data is captured by scanning collections of full microscope slides under high power. For each microscope slide, high power images are captured at multiple focal planes. Once the image is captured, this application will make it possible to emulate the behavior of a microscope by allowing users to continuously move the stage and to simulate changing magnification and focus. The Virtual Microscope is being designed so that it will be possible to achieve an interactive level of response as the same dataset is simultaneously explored by multiple users. In the ongoing Johns Hopkins application effort, the Virtual Microscope will be coupled to computation modules that carry out: (1) three dimensional image reconstruction from data found in multiple focal planes and on multiple microscope slides, (2) image registration and compositing that takes into account data obtained using various special stains that reveal the presence or absence of biochemical markers, (3) image segmentation and pattern recognition to better characterize known malignancies, and (4) applications that aid the pathologist in screening for possible malignancy.
In the proposed effort, we will develop an applications emulator that prototypes the performance characteristics associated with the current application suite along with the performance characteristics that will be associated with more ambitious future projects. The future projects will involve integration of additional sensor modalities along with associative retrieval of sensor data obtained from related cases. The sensor modalities that would be involved in a sensor data fusion effort include (1) different radiological imaging modalities such as CT, MRI and PET, (2) electron microscopy, (3) confocal microscopy, and (4) conventional high power light microscopy.
In the development of our application emulator, we will make the realistic assumption that sensor data is stored in multiple data repositories. Each data repository will consist of a multiprocessor architecture along with secondary and tertiary storage. The application emulator will carry out varying quantities and patterns of computation (computational patterns will be abstracted from our application codes). The processing carried out at a particular repository will result in a variable degree of data size reduction. Sensor data fusion will be emulated by prototyping processing algorithms that take as inputs processed data sets produced by several repositories.
The application emulator will emulate this ambitious sensor data fusion application suite in a parameterized fashion. Parameter adjustment will make it possible to use the emulator for various application scenarios. The behavior we emulate will include computation, secondary storage accesses, tertiary storage accesses, remote object invocations, and program migration between processing nodes.