Postdoctoral Fellow in Wound Assessment using Deep Learning on Smartphone Images (wpi)

Job Description

JOB TITLE

Postdoctoral Fellow in Wound Assessment using Deep Learning on Smartphone Images

LOCATION

Worcester

DEPARTMENT NAME

Computer Science NFR - JM

DIVISION NAME

Worcester Polytechnic Institute - WPI

JOB DESCRIPTION SUMMARY

Postdoctoral Fellow for 18 month appointment for NIH research

JOB DESCRIPTION

We seek to hire a post-doctoral researcher to join our NIH-funded Smartphone Wound Image Analysis and Decision Support (SmartWAnDS) research team, which includes faculty and several graduate students. The SmartWAnDS project is researching and developing techniques, machine learning models and a framework to automatically assess the healing progress of various types of wounds outside the clinic from a smartphone image. This postdoc position focuses on two key issues:

1. Performance optimization of deployed deep learning models for assessing wounds from photographs: The performance of machine learning models often decreases when they are deployed live, and the performance numbers achieved are much lower than those observed during model development on static datasets and evaluation using cross-validation.

2. Detecting infected wounds by jointly analyzing a thermal image and photograph using deep learning: Over 50% of Diabetic Foot Ulcers (DFUs) become infected, of which 20% lead to amputations.

JOB IS FROM: postdocjobs.oneVIEW

The ideal candidate will have a PhD in computer science or a closely related field, with a demonstrated background in image analysis, computer vision and deep learning. Example topics of research tasks will fall into two broad areas include but not limited to:

Performance Optimization of Deployed Machine Learning Wound Assessment System

• Image noise reduction, artefact rejection, image pre-processing and enhancement
• On-device low quality image rejection by using deep learning models to predict wrong or low certainty assessments from measurable image parameter (E.g., image contrast, brightness, wound coverage of image pixels)
• Domain adaptation of deep learning wound image assessment models to optimize live performance.

Research and Development of Deep Learning Models that detect infected wounds by jointly analyzing thermal images and photographs.

• Disentangled joint representations of photographs and thermal wound images.
• Convolutional Neural Networks (CNN) architectures for joint analyses of photographs and thermal images to predict wound infection.
• Generative Adversarial Networks (GANs) and diffusion models for data augmentation of pairs of wound photographs and thermal images
• Few shot techniques for joint analyses of photographs and thermal images

An interest in interdisciplinary research in digital health, and a proven track record of high-quality research and academic excellence are expected. Experience using deep learning to analyze images is desirable. Excellent written and verbal communication in English is required.

Responsibilities and Position Details

In addition to producing original research that will result in high quality publications on the above topics, the postdoctoral scholar will also assist in managing the project, directing the research of graduate students, coordinating research across the team, and preparing project deliverables for the sponsor. 

Research funding for the post-doctoral position is available for 12 months but an extension is possible.

FLSA STATUS

United States of America (Exempt)

WPI is an Equal Opportunity Employer that actively seeks to increase the diversity of its workplace. All qualified candidates will receive consideration for employment without regard to race, color, age, religion, sex, sexual orientation, gender identity, national origin, veteran status, or disability. It seeks individuals with​ diverse backgrounds and experiences who will contribute to a culture of creativity, collaboration, inclusion, problem solving, innovation, high performance, and change making. It is committed to maintaining a campus environment free of harassment and discrimination.