Enhanced Colonoscopy

PI: Yang Cai ycai@cmu.edu, CyLab
Visual Intelligence Studio
College of Engineering

Website: http://www.andrew.cmu.edu/user/ycai/index.html

Executive Summary: Colonoscopies should have the potential to dramatically reduce colon cancer rates and associated deaths. However, observed reductions in cancer rates do not fit with expectations and people who receive colonoscopies may still get colon cancer. Studies show that there are high polyp miss rates. The objective of this project is to develop video analytics software for standardizing the colonoscopy procedures and aid practitioners with real-time feedback during the procedure. Our approach includes digitizing, storing, and streaming video output from colonoscopies and developing Quality of Exam (QoE) matrices and algorithms, as well as the visualization algorithms to highlight polyps for improving detection during procedures. Dr. Yang Cai expects results will result in improvement in the exam quality, resulting in fewer expenses put toward colon cancer treatments. The tools developed from this project will potentially be deployed through a healthcare network. The technology could also be used to assist lower-level providers trained in colonoscopy.

Next Generation ECG and Machine Learning Software

PI: David Stager, NREC cop@cmu.edu
Eric Meyhofer, NREC meyhofer@rec.ri.cmu.edu
Neil Stegal, NREC nfs@cmu.edu
School of Computer Science

Website: Stager: http://www.ri.cmu.edu/person.html?person_id=296
Meyhofer: http://www.ri.cmu.edu/person.html?person_id=1637
Stegall: http://www.ri.cmu.edu/person.html?person_id=1848

Executive Summary: David Stager and Eric Meyhofer of National Robotics Engineering Center (NREC/CMU) will collaborate with Dr. Emerson Liu and Dr. Amit Thosani of Allegheny General Hospital (AGH) to develop the NextGen ECG: a portable ECG platform which employs high-resolution/high fidelity signal acquisition and modern signal processing techniques to better identify abnormal cardiac substrates, quantify presence of cardiac electrical instability, classify arrhythmias and risk-stratify for SCD. Dr. Stager’s group has enumerated various improvements into separate areas of development to include 1) extending the frequency band to include higher frequency data logging and analysis 2) increasing the number of electrodes for simultaneous beat-to-beat data sampling and 3) applying machine learning algorithms for automated detection and diagnosis of heart problems.

Anomaly and Pattern Detection in Medical Claims Data

PI: Jeff Schneider jeff.schneider@cs.cmu.edu
Robotics/School of Computer Science

Website: http://www.cs.cmu.edu/~schneide/

Executive Summary: Medical claims data is filled with mistakes, both intentional and unintentional, as well as opportunities to improve cost and reimbursement both by recognizing trends and causing changes. Dr. Schneider proposes to develop and apply machine learning algorithms to identify these phenomena. The algorithms will need to be statistically efficient because the number of hypotheses tested will be large but the tolerance for false positives will be limited. They will also need to be computationally efficient in order to sift through billions of records.

The research will proceed in three stages. First, methods of identifying anomalous mistakes and trends will be developed. Second, methods of assessing the utility of real and hypothetical changes will be developed to aid the prioritization of discoveries. Finally, supervised learning methods will be developed that take feedback from the user in order to learn and flag re-occurring patterns.

Dr. Schneider and his team will build on past expertise and experience in developing and deploying learning algorithms in disease surveillance and fleet health monitoring. If successful this effort will lead directly to commercialization in claims monitoring and can be extended to analysis of medical treatment practices.

Development of anti-biofilm surface coatings based on long-wavelength photosensitizers

PI: Aaron Mitchell, Biological Sciences, apm1@cmu.edu
Luisa Hiller, Biological Sciences, lhiller@andrew.cmu.edu
Frederick Lanni, Biological Sciences, lanni@andrew.cmu.edu

Website: Mitchell: http://www.cmu.edu/bio/faculty/mitchell.html
Hiller: http://www.cmu.edu/bio/faculty/hiller.html
Lanni: http://www.cmu.edu/bio/faculty/lanni.html

Executive Summary: Microbial biofilm growth on implanted medical hardware is a source of persistent infection in postoperative patients. For many reasons, biofilm cells are not cleared by the immune system or systemic antibiotics in these patients. This project focuses on suppression of the initial step of biofilm formation as it would occur in the post-op patient – the growth of adherent microbes on the surfaces of newly implanted medical hardware. The team’s approach is to develop a surface coating for implantable objects that contains a long-wavelength photosensitizer dye in non-releasable form. The photosensitizer is activated by light to generate toxic oxygen derivatives, and a single coating may be activated repeatedly by exposure to light. Dr. Mitchell and his team focus on photosensitizers that are excited by far-red light, because far-red light is capable of penetrating skin and musculoskeletal tissue. In practice in an immediate post-op setting, transdermal illumination will be used to activate the dye coating. However, because reactive oxygen generated by the photosensitizer is short-lived, toxic effects will be restricted to the interface region between hardware and tissue, precisely where biofilm initiation occurs. The proposed work will test the key features of photosensitized-surface biofilm suppression in culture conditions and in a biofilm infection model.

A Wearable System For Home-monitoring of Chronic Movement Disorders: Cost-effective Solution to Frequent Clinic Visits

PI: Jessica K. Hodgins, jkh@cs.cmu.edu
Fernando de la Torre, ftorre@cs.cmu.edu
Computer Science

Website: Hodgins: http://www.cs.cmu.edu/~jkh/
De La Torre: http://www.cs.cmu.edu/~ftorre/index.html

Executive Summary: There has been a large volume of research directed towards improving and building wearable technologies for assessing movement disorders. Most of them involve body-worn motion sensors like accelerometers. However, these cyber-physical systems are generally blind in the sense that they lack awareness of the context in which they are being used. In some cases, the motion profile of ordinary daily activities might match to those of major clinical symptoms (e.g, motion of hands while brushing teeth is similar to some involuntary movements in Parkinson’s Disease). An understanding of the semantic and functional context (what was the patient doing and what object he was interacting with) is required to differentiate between the two. The goal of Dr. Hodgins’ work is to couple ongoing advancements in wearable systems and statistical machine learning with state-of-the-art context based reasoning approaches. By advancing these areas of research and leveraging their complementary strengths, the team anticipates being able to construct robust systems for monitoring of patients with a variety of conditions. They believe that understanding context will provide the necessary clues for analyzing motion data and greatly simplify the task of reconstructing from sparse and noisy signals from sensors such as accelerometers. Conversely, those noisy signals will make reconstruction of the context easier.

This project is an attempt to marry computer vision and audio processing with learning-based human motion sensing systems. It aims improves the state-of-the-art of home monitoring systems and proposes to utilize multiple types of sensors (audio, video and motion) for context-aware processing. Dr. Hodgins’ team proposes to integrate wearable IMU based systems together with context reasoning algorithms to make them more effective for clinical purposes. The monitoring system, consisting of audio-visual and motion sensors will utilize novel modeling and machine learning algorithms for segmentation and activity classification/analysis. In order to address the uncontrolled nature of the daily living environments, they plan to develop novel weakly supervised learning algorithms that will learn visuomotor patterns from insufficient ground truth information.

Novel Therapy Targeting Food/Water Poisoning

PI: Adam Linstedt, linstedt@andrew.cmu.edu
Biological Sciences

Website: http://www.cmu.edu/bio/faculty/linstedt.html

Executive Summary: Enteric disease & diarrhea are globally significant health problems. A major fraction of this disease comes from food and waterborne disease caused by contaminating bacteria producing proteinaceous toxins. Indeed, Shiga Toxin (STx) producing bacteria of the Shigella genus and enterohemorrhagic E. coli (EHEC) species infect over 150 million individuals each year, cause over a million deaths and represent a global public health problem. The economic impact is also significant given that even the 100,000 EHEC cases diagnosed annually in the United States cost over $500 million. A recent outbreak of 3816 cases (including 54 deaths) in Germany originated from contaminated sprouts and resulted in 845 cases of costly and deadly hemolytic–uremic syndrome. Even a single bacterium infecting an individual can cause the disease. Currently, there is no definitive medical treatment for STx infections. The only recourse is rehydration therapy to support patients as they suffer chronic and bloody diarrhea often leading to fatal kidney failure. Antidotes for the toxin are not available and treatment of the bacteria with antibiotics is contraindicated because it is known to increase STx release and the development of lethal kidney disease.

Dr. Linstedt’s group will work to provide a therapy to combat Shiga toxicosis using the well-known metal ion manganese (Mn) either alone or in conjunction with antibiotics. Manganese is an essential nutrient and the fourth most widely available metal on earth. Its toxicology is well studied and, as a nutritional supplement, it is already approved for oral and intravenous use. The low cost and wide availability of Mn makes it amenable for use in developing countries where >95% of STx infections occur. Prolonged over-exposure to Mn can also cause health problems, but, as described in the next paragraph, we have shown that acute, non-toxic doses of Mn yield complete protection in cellular and mouse models against lethal STx challenges.

2013 Seed Funded Projects

Retinal Prosthesis for the Blind: Preclinical Electrical and Thermal Modeling

PI: Shawn K. Kelly, Institute for Complex Engineered Systems, skkelly@andrew.cmu.edu
Website: http://www.ices.cmu.edu/shawn-kelly.asp

Therapeutic Polymer Dressings for Chronic Wound Healing

PI: Kathryn Whitehead, Chemical Engineering & Biomedical Engineering, kawhite@cmu.edu
Website: http://www.cmu.edu/cheme/people/faculty/kathryn-a-whitehead.html

Drug Eluting Coils for Improved Treatment of Brain Aneurysms

PI: Christopher J Bettinger, Biomedical Engineering, cbetting@andrew.cmu.edu
Website: http://www.materials.cmu.edu/people/bettinger.html

Office-centered Three-dimensional Ultrasonic Imaging of Rotator Cuff Tears

PI: Kenji Shimada, Mechanical Engineering, shimada@cmu.edu
Website: http://www.andrew.cmu.edu/user/shimada/

Conformable Ultrasound Transducer for Diagnosis of Deep Vein Thrombosis

PI: David Greve, Electrical and Computer Engineering, dg07@andrew.cmu.edu
Website: http://users.ece.cmu.edu/~dwg/

Monitoring and Coaching to Promote Proper Inhaler Technique

PI: Alexander Hauptmann, Computer Science Department and Language Technologies Institute, alex@cs.cmu.edu
Website: http://www.cs.cmu.edu/~alex/

System for Semi-automated Minimally Invasive Myocardial Gene Transfer

PI: Cameron Riviere, Robotics Institute, camr@ri.cmu.edu
Website: http://www.ri.cmu.edu/person.html?person_id=248

Development of a rapid diagnostic tool for the detection and characterization of infections during surgery

PI: Marcel Bruchez, Biological Sciences, bruchez@cmu.edu
Website: http://www.cmu.edu/bio/faculty/bruchez.html

Contact Us

  Phone: 412-268-3978
  Email: lwinter@andrew.cmu.edu