Deep Learning Machine

YUN ZOU

* ********* *****, ** ***** MO, *****

******@*****.*** 314-***-****

SUMMARY

A proficient deep learning researcher with a strong passion for developing innovative applications. 5+ years academic experience as a researcher and engineer for signal processing, computer vision and deep learning models. Experienced in biomedical image reconstruction, image segmentation, and disease classification.

SKILLS AND INTERESTS

Programming Languages: Python, MATLAB, C, C++

• Frameworks: Pytorch, Tensorflow, OpenCV, Scipy

• Techniques: Neuron Radiance Field (NeRF), YOLO, DeepSort, ViT, TransUnet, Image processing algorithms, COMSOL, Reinforcement Learning, Object detection, Object tracking, Image segmentation

Tools: Linux, Anaconda, MySQL, Git, Bash, Google Colab EDUCATION

Washington University in St Louis, St Louis, MO Sep. 2019 – Present Major: Ph.D. in Biomedical Engineering, GPA: 3.9 / 4 Minor: Machine Learning and Computer Science, GPA: 3.9 / 4 Tsinghua University, Beijing, China Aug. 2016 – Jul. 2019 Master of Science in Engineer Physics, GPA: 3.9 / 4 Tsinghua University, Beijing, China Aug. 2012 – Jul. 2016 Bachelor of Science in Engineer Physics, GPA: 3.8 / 4 RESEARCH EXPERIENCE

Ultrasound and Optical Imaging Lab, Washington University in St Louis Research assistant, Advisor: Prof. Quing Zhu Aug. 2019 – Present

Deep learning model for DOT reconstruction: 1. Designed an auto-encoder-like deep learning model for Diffuse Optical Tomography (DOT) reconstruction. 2. Combined physical information losses (Born Constraint and Anatomical Constraint) into the model. 3. Achieved a 36% increase in contrast between benign and malignant breast lesions.

US-enhanced Unet model for qPAT reconstruction: 1. Developed an ultrasound-enhanced Unet model for quantitative photoacoustic tomography (PAT) reconstruction. 2. Extracted US images features by ResNet-18 model. 3. Obtained an AUC of 0.94 and an accuracy of 0.89 for ovarian cancer diagnosis.

PA-NeRF model for 3D synthesis: 1. Implemented a Photoacoustic Neural Radiance Field (PA-NeRF) model on 3D PAT reconstruction from sparse B-scan data. 2.Outputed PA images and US images for accurate rendering 3D volume.

Dual-input-transformer model for classification Treatment response: 1. Constructed dual input transformer model (DiT) for Pathological Complete Response (PCR) to Neoadjuvant Chemotherapy in Breast Cancer. 2.Explored different input setups to choose the best classification model. X-Ray Imaging & Computed Tomography Lab, University of Wisconsin Madison Research Intern, Advisor: Prof. Guanghong Chen Jul. 2018 – Sep. 2018

Dimension reduction for Breast Cancer images: 1. Compared Boltzmann machine, auto-encoder, and PCA for dimension reduction on medical images to achieve unsupervised learning. 2. Based on reduced dimensional CT images to cluster different categories of breast cancer. SELECTED HONORS AND AWARDS

WashU Biomedical Engineering Outstanding Research Award 2023

Top Downloaded Papers in Biomedical Optics Express 2022

Second Award, Student scholarship, Tsinghua University 2018 MAIN PUBLICATIONS

Ultrasound and diffuse optical tomography-transformer model for assessing pathological complete response to neoadjuvant chemotherapy in breast cancer.

Journal of Biomedical Optics 29.7 (2024) First author https://doi.org/10.1117/1.JBO.29.7.076007 PA-NeRF, a neural radiance field model for 3D photoacoustic tomography reconstruction from limited Bscan data

Biomedical Optics Express 15.3 (2024) First author https://doi.org/10.1364/BOE.511807 Machine learning model with physical constraints for diffuse optical tomography

Biomedical Optics Express 12.9 (2021) First author doi: https://doi.org/10.1364/BOE.432786 Ultrasound-enhanced Unet model for quantitative photoacoustic tomography of ovarian lesions

Photoacoustics 28 (2022): 100420. First author doi: https://doi.org/10.1016/j.pacs.2022.100420

Contact this candidate