Voxeleron LLC is a software house specializing in algorithm development for computer vision applications.

• Lead the 3D medical image analysis research and development in the United States for Fujifilm’s Picture Archiving and Communication System (PACS). Worked together with consulting physician partners and marketing department to define the research agenda for Fujifilm’s U.S. R&D units in the CT and mammography applications space. Managed relationships with outside clinical research groups for data collection and algorithm validation.• Managed the algorithm development (research, schedules, and deliverables) for CT and mammography R&D resulting in several 3D image analysis projects. Algorithms are currently integrated into applications for Fujifilm’s PACS product/workstation.• Collaborated overseas with Fujifilm’s Tokyo labs to help allocate resources for development projects. Worked as a liaison between U.S. and Japanese operations to facilitate the design control and regulatory documentation of U.S. algorithms.• Designed and implemented a support vector machine system with an automated distributed, parallel-processing training method for general pattern analysis tasks. This system is being used in several products and has improved the accuracy of existing classification systems by up to 35 percent. • Designed and implemented a generalized statistical shape modeling system for atlas generation and model-based segmentation of arbitrary shapes. This system has been used in both the U.S. and Japan to aid segmentation algorithms for several different anatomical structures. • Served as technical lead for design and implementation of expert systems for feature-based pattern recognition of malignant tumors. Improved performance of existing system by more than 50 percent.• Designed and oversaw implementation of a distributed computer cluster for batch testing of algorithms using Condor. The cluster sped up regression testing by more than 10x and, since 2006, has performed more than 59 compute years of processing.

See above

• Built end-to-end, intensity-based 2D-3D registration system for minimally-invasive image-guided surgery and validated it to 1 mm accuracy on real, clinical data.• Developed attenuation-field rendering, an extension of light-field rendering from computer graphics applied to X-ray fluoroscopy. Achieved 20-30x speedup for rendering using attenuation fields vs. conventional techniques.

• Designed and implemented a 3D gesture tracking system for American Sign Language recognition.

• Designed and implemented a stereo-based, 3D head tracking system for use with a steerable microphone array.