(1) Led a machine learning team including two scientists and one engineer. Developed a deep-learning hardware / software framework for breast lesion analytics with live video lesion detection, live video lesion segmentation, lesion correlation, malignancy classification, and lesion descriptor / BI-RADS reporting based on breast ultrasound / x-ray tomosynthesis images and patient information.

(1) Investigated and developed a deep-learning based classification software module for breast density assessment to meet states' regulatory requirements. Significantly improved the breast composition category classification accuracies compared to other methods. (2) Developed Hologic's FDA approved machine-learning based breast density assessment product, Quantra 2.2, which provides more consistent density scores than that of the methods based x-ray imaging chain model (Quantra 2.0-2.1). Supported the FDA 510(k) filing. (3) Developed and released the core image processing algorithms and software modules for biopsy / specimen image processing in Hologic's Trident HD / Brevera / Affirm stereotactic imaging products. Supported the FDA 510(k) filing. (4) Developed and released a scanned film image enhancement software module in Hologic's DigitalNow HD product. Supported the FDA 510(k) filing. (5) Developed a deep-learning based image denoising algorithm and a software module for x-ray FFDM and tomosynthesis imaging, reducing x-ray dose significantly while maintaining the clinical diagnosis performance and image quality.(6) Developed breast cancer risk assessment software module (Osprey).(7) Developed a comprehensive x-ray dose calculations software for various x-ray anode / filter combinations, 2D and tomosynthesis, and ACR / EUREF guidelines. (8) Worked closely with marketing, clinical application, regulatory, and IP legal teams.

(1) Investigated X-ray optimization, modeled x-ray target/filter spectra for tungsten system, evaluated dose and image quality to optimize x-ray techniques for different applications, and performed clinical studies of breast contrast imaging in USC and tungsten-anode imaging in UCLA, respectively. (2) Created a comprehensive x-ray flat-panel detector test protocol for R&D and manufacturing. Performed intensive performance evaluation for different x-ray detectors. (3) Prepared the materials in the imaging performance section for FDA PMA and 510(k) filing. (4) Worked closely with marketing, engineering, regulatory, and global technical support teams.

(1) Investigated CEDM methods of dynamic analysis and dual-energy imaging with x-ray contrast agent injection. Performed clinical study at USC Norris cancer center. Acted as project manager. (2) Developed image processing algorithms and software modules for acquisition workstation and softcopy workstation, and next generation mammography product. (3) Developed image calibration algorithms and software for X-ray a-Se flat panel detector. (4) Solved the complicated field problems with the radiologists, technologists, applications specialists, field service engineers, and medical physicists involved.

(1) Pioneered in cone-beam volume CT breast imaging. Performed theoretical and experimental study regarding system design, x-ray technique optimization, cone-beam CT reconstruction algorithm, requirements on 2D detector, and 3D dose estimation.(2) Investigated X-ray volume CT angiography based on CCD-II and flat panel detector (FPD). Performed FPD calibration, data acquisition, cone-beam CT reconstruction algorithms, system evaluation, 3D image rendering using Analyzer software, and animal study.(3) Investigated X-ray volume CT lung imaging for early lung cancer detection based on a spiral scanner. Performed the prototype system evaluation.

(1) Cone-beam volume CT: investigated reconstruction algorithms, designed and implemented an experimental x-ray CCD-II cone-beam imaging system at SJTU University Hospital. (2) Reverse geometry based 2D and 3D image reconstruction: Developed reconstruction algorithms and built an experimental system with crystal detectors. (3) Multi-camera computer vision: Investigated camera calibration methods, and designed two-camera based hardware system.(4) Taught Signal and Linear System and Digital Image Processing.

(1) Three-dimensional rendering for in vivo lung volume: Designed 3D surface rendering algorithm and performed clinical study with a physician to evaluate 3D CT data. (2) Direct volume reconstruction: Developed volume CT reconstruction algorithms and performed experimental studies on a prototype imaging system.(3) Microprocessor-based smart instruments R&D: Developed an LCD display and compact-lead based miniaturized ECG recorder/analyzer.(4) Taught Signal and Linear System and Digital Image Processing.

(1) Investigated computer assisted endoscopic ENT surgery (commercialized by PRAXIM Inc., Grenoble, France). (2) Developed image registration method for real-time localizer data/3D CT images. Worked closely with an ENT surgeon for clinical validation study at CHU University Hospital.(3) Applied hybrid image registration method for computer assisted medical intervention. Performed 3D image segmentation, modeling, registration, rendering, and 3D localizer calibration for computer-assisted surgery.