Gemstone Team Research

Permanent URI for this collectionhttp://hdl.handle.net/1903/9070

The Gemstone Program at the University of Maryland is a unique multidisciplinary four-year research program for selected undergraduate honors students of all majors. Under guidance of faculty mentors and Gemstone staff, teams of students design, direct and conduct significant research, often but not exclusively exploring the interdependence of science and technology with society. Gemstone students are members of a living-learning community comprised of fellow students, faculty and staff who work together to enrich the undergraduate experience. This community challenges and supports the students in the development of their research, teamwork, communication and leadership skills. In the fourth year, each team of students presents its research in the form of a thesis to experts, and the students complete the program with a citation and a tangible sense of accomplishment.

Browse

Filters

20241

Settings

Start date: 2020Subject: search.filters.subject.detection

Search Results

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    Enhancement of Detection and Diagnosis of Non-Small Cell Lung Cancer Through The Improvement of Machine Learning and AI Models
    (2024) Beshaw, Yael; Cancro, George; Chang, Darren; Fomengia, Jayda; Mehta, Vanshika; Vedantham, Arjun; Yaragudipati, Ritvik; Feizi, Soheil
    Due to low survival rates and an unparalleled burden of non-small cell lung cancer on underserved communities, there is great urgency for innovative and accessible methods that will improve healthcare access for lung cancer patients. To combat this inequity, Team DOC aims to develop an AI model that is able to not only improve lung cancer diagnoses but also predict the progression of non-small cell lung cancer. We intend to evaluate the performance of a convolutional neural network on the LIDC-IDRI dataset and retrain the final layers of the model to improve its performance on the same dataset. Repeating this process on different model architectures allows us to determine which model performs optimally, providing a foundation to develop an end-to-end explainable AI workflow that can extract clinically relevant predictions of cancer progression for further analysis. Throughout our training process, we resolve to address the accuracy and potential for bias. Additionally, we are carrying out a survey among underserved populations and communities to discern the need for our improved cancer detection model. We hope that our model will be able to be implemented in communities with lack of access to healthcare systems to bridge the gap between underprivileged communities and unbiased care.