skip to main content

Explore Research Products in the NSF-PAR

Title: Methods and Outcomes of the NSF Project on Synthesizing Environments for Digitally-Mediated Team Learning
This poster paper describes the authors’ single-year National Science Foundation (NSF) project DRL-1825007 titled, “DCL: Synthesis and Design Workshop on Digitally-Mediated Team Learning” which has been conducted as one of nine awards within NSF-18-017: Principles for the Design of Digital STEM Learning Environments. Beginning in September 2018, the project conducted the activities herein to deliver a three-day workshop on Digitally-Mediated Team Learning (DMTL) to convene, invigorate, and task interdisciplinary science and engineering researchers, developers, and educators to coalesce the leading strategies for digital team learning. The deliverable of the workshop is a White Paper composed to identify one-year, three-year, and five-year research and practice roadmaps for highly-adaptable environments for computer-supported collaborative learning within STEM curricula. As subject to the chronology of events, highlights of the White Paper’s outcomes will be showcased within the poster itself.  more » « less
Award ID(s):
1825007
NSF-PAR ID:
10092034
Author(s) / Creator(s):
; ; ; ; ; ; ; ; ; ; ; ; ; ;
Date Published:
Journal Name:
2019 ASEE annual conference & exposition proceedings
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. ( , Rapid Community Report)
    The purpose of the Digitally-Mediated Team Learning Workshop (sponsored by the National Science Foundation through a Dear Colleague Letter [NSF 18-017] via grant 1825007) was to ascertain the current state of the field and future research approaches for DMTL delivered through synchronous modalities in STEM classrooms for students in upper elementary grades through college. The overarching question for the workshop was: “How can we advance effective and scalable digital environments for synchronous team-based learning involving problem-solving and design activities within STEM classrooms for all learners?” The workshop explored the state of the field and future directions of DMTL through its four tracks: (a) student-facing and instructor-facing tools, (b) learning analytics, (c) pedagogical and andragogical strategies, and (d) inclusivity. 
    more » « less
  2. ; ; ; ; ; ( , Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB))
    Obeid, I. (Ed.)
    The Neural Engineering Data Consortium (NEDC) is developing the Temple University Digital Pathology Corpus (TUDP), an open source database of high-resolution images from scanned pathology samples [1], as part of its National Science Foundation-funded Major Research Instrumentation grant titled “MRI: High Performance Digital Pathology Using Big Data and Machine Learning” [2]. The long-term goal of this project is to release one million images. We have currently scanned over 100,000 images and are in the process of annotating breast tissue data for our first official corpus release, v1.0.0. This release contains 3,505 annotated images of breast tissue including 74 patients with cancerous diagnoses (out of a total of 296 patients). In this poster, we will present an analysis of this corpus and discuss the challenges we have faced in efficiently producing high quality annotations of breast tissue. It is well known that state of the art algorithms in machine learning require vast amounts of data. Fields such as speech recognition [3], image recognition [4] and text processing [5] are able to deliver impressive performance with complex deep learning models because they have developed large corpora to support training of extremely high-dimensional models (e.g., billions of parameters). Other fields that do not have access to such data resources must rely on techniques in which existing models can be adapted to new datasets [6]. A preliminary version of this breast corpus release was tested in a pilot study using a baseline machine learning system, ResNet18 [7], that leverages several open-source Python tools. The pilot corpus was divided into three sets: train, development, and evaluation. Portions of these slides were manually annotated [1] using the nine labels in Table 1 [8] to identify five to ten examples of pathological features on each slide. Not every pathological feature is annotated, meaning excluded areas can include focuses particular to these labels that are not used for training. A summary of the number of patches within each label is given in Table 2. To maintain a balanced training set, 1,000 patches of each label were used to train the machine learning model. Throughout all sets, only annotated patches were involved in model development. The performance of this model in identifying all the patches in the evaluation set can be seen in the confusion matrix of classification accuracy in Table 3. The highest performing labels were background, 97% correct identification, and artifact, 76% correct identification. A correlation exists between labels with more than 6,000 development patches and accurate performance on the evaluation set. Additionally, these results indicated a need to further refine the annotation of invasive ductal carcinoma (“indc”), inflammation (“infl”), nonneoplastic features (“nneo”), normal (“norm”) and suspicious (“susp”). This pilot experiment motivated changes to the corpus that will be discussed in detail in this poster presentation. To increase the accuracy of the machine learning model, we modified how we addressed underperforming labels. One common source of error arose with how non-background labels were converted into patches. Large areas of background within other labels were isolated within a patch resulting in connective tissue misrepresenting a non-background label. In response, the annotation overlay margins were revised to exclude benign connective tissue in non-background labels. Corresponding patient reports and supporting immunohistochemical stains further guided annotation reviews. The microscopic diagnoses given by the primary pathologist in these reports detail the pathological findings within each tissue site, but not within each specific slide. The microscopic diagnoses informed revisions specifically targeting annotated regions classified as cancerous, ensuring that the labels “indc” and “dcis” were used only in situations where a micropathologist diagnosed it as such. Further differentiation of cancerous and precancerous labels, as well as the location of their focus on a slide, could be accomplished with supplemental immunohistochemically (IHC) stained slides. When distinguishing whether a focus is a nonneoplastic feature versus a cancerous growth, pathologists employ antigen targeting stains to the tissue in question to confirm the diagnosis. For example, a nonneoplastic feature of usual ductal hyperplasia will display diffuse staining for cytokeratin 5 (CK5) and no diffuse staining for estrogen receptor (ER), while a cancerous growth of ductal carcinoma in situ will have negative or focally positive staining for CK5 and diffuse staining for ER [9]. Many tissue samples contain cancerous and non-cancerous features with morphological overlaps that cause variability between annotators. The informative fields IHC slides provide could play an integral role in machine model pathology diagnostics. Following the revisions made on all the annotations, a second experiment was run using ResNet18. Compared to the pilot study, an increase of model prediction accuracy was seen for the labels indc, infl, nneo, norm, and null. This increase is correlated with an increase in annotated area and annotation accuracy. Model performance in identifying the suspicious label decreased by 25% due to the decrease of 57% in the total annotated area described by this label. A summary of the model performance is given in Table 4, which shows the new prediction accuracy and the absolute change in error rate compared to Table 3. The breast tissue subset we are developing includes 3,505 annotated breast pathology slides from 296 patients. The average size of a scanned SVS file is 363 MB. The annotations are stored in an XML format. A CSV version of the annotation file is also available which provides a flat, or simple, annotation that is easy for machine learning researchers to access and interface to their systems. Each patient is identified by an anonymized medical reference number. Within each patient’s directory, one or more sessions are identified, also anonymized to the first of the month in which the sample was taken. These sessions are broken into groupings of tissue taken on that date (in this case, breast tissue). A deidentified patient report stored as a flat text file is also available. Within these slides there are a total of 16,971 total annotated regions with an average of 4.84 annotations per slide. Among those annotations, 8,035 are non-cancerous (normal, background, null, and artifact,) 6,222 are carcinogenic signs (inflammation, nonneoplastic and suspicious,) and 2,714 are cancerous labels (ductal carcinoma in situ and invasive ductal carcinoma in situ.) The individual patients are split up into three sets: train, development, and evaluation. Of the 74 cancerous patients, 20 were allotted for both the development and evaluation sets, while the remain 34 were allotted for train. The remaining 222 patients were split up to preserve the overall distribution of labels within the corpus. This was done in hope of creating control sets for comparable studies. Overall, the development and evaluation sets each have 80 patients, while the training set has 136 patients. In a related component of this project, slides from the Fox Chase Cancer Center (FCCC) Biosample Repository (https://www.foxchase.org/research/facilities/genetic-research-facilities/biosample-repository -facility) are being digitized in addition to slides provided by Temple University Hospital. This data includes 18 different types of tissue including approximately 38.5% urinary tissue and 16.5% gynecological tissue. These slides and the metadata provided with them are already anonymized and include diagnoses in a spreadsheet with sample and patient ID. We plan to release over 13,000 unannotated slides from the FCCC Corpus simultaneously with v1.0.0 of TUDP. Details of this release will also be discussed in this poster. Few digitally annotated databases of pathology samples like TUDP exist due to the extensive data collection and processing required. The breast corpus subset should be released by November 2021. By December 2021 we should also release the unannotated FCCC data. We are currently annotating urinary tract data as well. We expect to release about 5,600 processed TUH slides in this subset. We have an additional 53,000 unprocessed TUH slides digitized. Corpora of this size will stimulate the development of a new generation of deep learning technology. In clinical settings where resources are limited, an assistive diagnoses model could support pathologists’ workload and even help prioritize suspected cancerous cases. ACKNOWLEDGMENTS This material is supported by the National Science Foundation under grants nos. CNS-1726188 and 1925494. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation. REFERENCES [1] N. Shawki et al., “The Temple University Digital Pathology Corpus,” in Signal Processing in Medicine and Biology: Emerging Trends in Research and Applications, 1st ed., I. Obeid, I. Selesnick, and J. Picone, Eds. New York City, New York, USA: Springer, 2020, pp. 67 104. https://www.springer.com/gp/book/9783030368432. [2] J. Picone, T. Farkas, I. Obeid, and Y. Persidsky, “MRI: High Performance Digital Pathology Using Big Data and Machine Learning.” Major Research Instrumentation (MRI), Division of Computer and Network Systems, Award No. 1726188, January 1, 2018 – December 31, 2021. https://www. isip.piconepress.com/projects/nsf_dpath/. [3] A. Gulati et al., “Conformer: Convolution-augmented Transformer for Speech Recognition,” in Proceedings of the Annual Conference of the International Speech Communication Association (INTERSPEECH), 2020, pp. 5036-5040. https://doi.org/10.21437/interspeech.2020-3015. [4] C.-J. Wu et al., “Machine Learning at Facebook: Understanding Inference at the Edge,” in Proceedings of the IEEE International Symposium on High Performance Computer Architecture (HPCA), 2019, pp. 331–344. https://ieeexplore.ieee.org/document/8675201. [5] I. Caswell and B. Liang, “Recent Advances in Google Translate,” Google AI Blog: The latest from Google Research, 2020. [Online]. Available: https://ai.googleblog.com/2020/06/recent-advances-in-google-translate.html. [Accessed: 01-Aug-2021]. [6] V. Khalkhali, N. Shawki, V. Shah, M. Golmohammadi, I. Obeid, and J. Picone, “Low Latency Real-Time Seizure Detection Using Transfer Deep Learning,” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2021, pp. 1 7. https://www.isip. piconepress.com/publications/conference_proceedings/2021/ieee_spmb/eeg_transfer_learning/. [7] J. Picone, T. Farkas, I. Obeid, and Y. Persidsky, “MRI: High Performance Digital Pathology Using Big Data and Machine Learning,” Philadelphia, Pennsylvania, USA, 2020. https://www.isip.piconepress.com/publications/reports/2020/nsf/mri_dpath/. [8] I. Hunt, S. Husain, J. Simons, I. Obeid, and J. Picone, “Recent Advances in the Temple University Digital Pathology Corpus,” in Proceedings of the IEEE Signal Processing in Medicine and Biology Symposium (SPMB), 2019, pp. 1–4. https://ieeexplore.ieee.org/document/9037859. [9] A. P. Martinez, C. Cohen, K. Z. Hanley, and X. (Bill) Li, “Estrogen Receptor and Cytokeratin 5 Are Reliable Markers to Separate Usual Ductal Hyperplasia From Atypical Ductal Hyperplasia and Low-Grade Ductal Carcinoma In Situ,” Arch. Pathol. Lab. Med., vol. 140, no. 7, pp. 686–689, Apr. 2016. https://doi.org/10.5858/arpa.2015-0238-OA. 
    more » « less
  3. ; ; ; ; ( , ASEE Conferences)
    nterest in science, technology, engineering, and mathematics (STEM) begins as early as elementary and middle school. As youth enter adolescence, they begin to shape their personal identities and start making decisions about who they are and could be in the future. Students form their career aspirations and interests related to STEM in elementary school, long before they choose STEM coursework in high school or college. Much of the literature examines either science or STEM identity and career aspirations without separating out individual sub-disciplines. Therefore, the purpose of this paper is to describe the development of a survey instrument to specifically measure engineering identity and career aspirations in adolescents and preadolescents. When possible, we utilized existing measures of STEM identity and career aspirations, adapting them when necessary to the elementary school level and to fit the engineering context. The instrument was developed within the context of a multi-year, NSF-funded research project examining the dynamics between undergraduate outreach providers and elementary students to understand the impact of the program on students’ engineering identity and career aspirations. Three phases of survey development were conducted that involved 492 elementary students from diverse communities in the United States. Three sets of items were developed and/or adapted throughout the four phases. The first set of items assessed Engineering Identity. Recent research suggests that identity consists of three components: recognition, interest, and performance/competence. Items assessing each of these constructs were included in the survey. The second and third sets of items reflected Career Interests and Aspirations. Because elementary and middle school students often have a limited or nascent awareness of what engineers do or misconceptions about what a job in science or engineering entails, it is problematic to measure their engineering identity or career aspirations by directly asking them whether they want to be a scientist/engineer or by using a checklist of broad career categories. Therefore, similar to other researchers, the second set of items assessed the types of activities that students are interested in doing as part of a future career, including both non-STEM and STEM (general and engineering-specific) activities. These items were created by the research team or adapted from activity lists used in existing research. The third set of items drew from career counseling measures relying on Holland’s Career Codes. We adapted the format of these instruments by asking students to choose the activity they liked the most from a list of six activities that reflected each of the codes rather than responding to their interest about each activity. Preliminary findings for each set of items will be discussed. Results from the survey contribute to our understanding of engineering identities and career aspirations in preadolescent and adolescent youth. However, our instrument has the potential for broader application in non-engineering STEM environments (e.g., computer science) with minor wording changes to reflect the relevant science subject area. More research is needed in determining its usefulness in this capacity. 
    more » « less
  4. ; ; ; ; ( , ASEE Peer)
    The STEM Excellence through Engagement in Collaboration, Research, and Scholarship (SEECRS) project at Whatcom Community College is a five-year program aiming to support academically talented students with demonstrated financial need in biology, chemistry, geology, computer science, engineering, and physics. This project is funded by an NSF S-STEM (Scholarships in Science, Technology, Engineering, and Mathematics) grant awarded in January 2017. Through an inclusive and long-range effort, the college identified a strong need for financial and comprehensive supports for STEM students. This project will offer financial, academic, and professional support to three two-year cohorts of students. The SEECRS project aims to utilize a STEM-specific guided pathways approach to strengthen recruitment, retention, and matriculation of STEM students at the community college level. Scholarship recipients will be supported through participation in the SEECRS Scholars Academy, a multi-pronged approach to student support combining elements of community building, faculty mentorship, targeted advising activities, authentic science practice, and social activities. Students are introduced to disciplines of interest through opportunities to engage in course-based undergraduate research experiences (CUREs) in Biology, Chemistry and Engineering courses, funded summer research opportunities, and seminars presented by STEM professionals. Communities of practice will be nurtured through the introduction of cohort building and faculty mentorship. Cohort development starts with a required two-credit course for all scholars that emphasizes STEM identity development, specifically focusing on identifying and coping with the ways non-dominant individuals (racial/ethnic minorities, non-male gender, lower socioeconomic status, first-generation, 2-year community college vs. 4-year institutions) are made to feel as outsiders in STEM. Each SEECRS scholar is paired with a faculty mentor who engages in ongoing mentor training. The project evaluation will determine the efficacy of the project activities in achieving their intended outcomes. Specifically, we will collect data to answer the research question: To what extent can a guided pathways approach provide a coordinated and supported STEM experience at Whatcom Community College that: (1) increases student success, and (2) positively shifts students’ STEM self-identity? The evaluation will employ a quasi-experimental research design, specifically a pretest-posttest design with a matched comparison group. Our first cohort of 14 students was selected over two application rounds (winter and summer 2017). We awarded ten full scholarships and four half-scholarships based on financial need data. Cohort demographics of note compared to institutional percentages are: females (64% vs. 57%), Hispanic (14% vs. 17%), African American (7% vs. 2%), white (79% vs. 66%), first generation college bound (43% vs. 37%). The cohort is comprised of six students interested in engineering, six in biology, and one each in geology and environmental sciences. With increased communication between the project team, our Financial Aid office, Entry and Advising, high school outreach, and the Title III grant-funded Achieve, Inspire, Motivate (AIM) Program, as well as a longer advertising time, we anticipate significantly enhancing our applicant pool for the next cohort. The results and lessons learned from our first year of implementation will be presented. 
    more » « less
  5. ; ; ; ; ( , ASEE annual conference & exposition proceedings)
    The STEM Excellence through Engagement in Collaboration, Research, and Scholarship (SEECRS) project at Whatcom Community College is a five-year program aiming to support academically talented students with demonstrated financial need in biology, chemistry, geology, computer science, engineering, and physics. This project is funded by an NSF S-STEM (Scholarships in Science, Technology, Engineering, and Mathematics) grant awarded in January 2017. Through an inclusive and long-range effort, the college identified a strong need for financial and comprehensive supports for STEM students. This project will offer financial, academic, and professional support to three two-year cohorts of students. The SEECRS project aims to utilize a STEM-specific guided pathways approach to strengthen recruitment, retention, and matriculation of STEM students at the community college level. Scholarship recipients will be supported through participation in the SEECRS Scholars Academy, a multi-pronged approach to student support combining elements of community building, faculty mentorship, targeted advising activities, authentic science practice, and social activities. Students are introduced to disciplines of interest through opportunities to engage in course-based undergraduate research experiences (CUREs) in Biology, Chemistry and Engineering courses, funded summer research opportunities, and seminars presented by STEM professionals. Communities of practice will be nurtured through the introduction of cohort building and faculty mentorship. Cohort development starts with a required two-credit course for all scholars that emphasizes STEM identity development, specifically focusing on identifying and coping with the ways non-dominant individuals (racial/ethnic minorities, non-male gender, lower socioeconomic status, first-generation, 2-year community college vs. 4-year institutions) are made to feel as outsiders in STEM. Each SEECRS scholar is paired with a faculty mentor who engages in ongoing mentor training. The project evaluation will determine the efficacy of the project activities in achieving their intended outcomes. Specifically, we will collect data to answer the research question: To what extent can a guided pathways approach provide a coordinated and supported STEM experience at Whatcom Community College that: (1) increases student success, and (2) positively shifts students’ STEM self-identity? The evaluation will employ a quasi-experimental research design, specifically a pretest-posttest design with a matched comparison group. Our first cohort of 14 students was selected over two application rounds (winter and summer 2017). We awarded ten full scholarships and four half-scholarships based on financial need data. Cohort demographics of note compared to institutional percentages are: females (64% vs. 57%), Hispanic (14% vs. 17%), African American (7% vs. 2%), white (79% vs. 66%), first generation college bound (43% vs. 37%). The cohort is comprised of six students interested in engineering, six in biology, and one each in geology and environmental sciences. With increased communication between the project team, our Financial Aid office, Entry and Advising, high school outreach, and the Title III grant-funded Achieve, Inspire, Motivate (AIM) Program, as well as a longer advertising time, we anticipate significantly enhancing our applicant pool for the next cohort. The results and lessons learned from our first year of implementation will be presented. 
    more » « less
  • Citation Formats
  • MLA
  • APA
  • Chicago
  • BibTeX
  • Save / Share this Record
  • Send to Email