More Like this

1. Becoming Somebody in STEM: Developing a cultural of criticality in the space between person and history

   Calabrese Barton, A. ( April 2018 , American Education Research Association)

   Autumn, a young white woman growing up in multi-generational poverty in an economically challenged midwestern city, has authored a STEM-empowered life against the dominant sociohistorical narrative in American society. She has attended public schools that served significant populations living in high poverty – overcrowded classrooms, high teacher turnovers, out-of-field teaching, and limited STEM resources. Yet, she has authored herself into STEM despite these mitigating circumstances. Autumn is currently a high school senior with aspirations to become an engineer or a hairdresser working in an eco-salon. She spends part of her time after school in a makerspace housed in her local community center, building things to solve other people’s problems. She reminds us that her out-of-school efforts to participate in STEM exist worlds away from schooling. However, she takes the optimistic view that if she could tell teachers about her out-of-school STEM experiences, her teachers might be better able to help her and her peers serve the interests and needs of her community, as well as see Autumn’s potential in STEM. Now a rising 12th grader, we have followed Autumn since 5th grade through school and afterschool. As she grew older, she became more interested in helping us to document and tell her story. She is an author on this paper. Also during this time, Autumn has shifted from wanting to be a hair designer (5th grade) to wanting to own a “green” (environmentally friendly) hair salon (8th grade), to considering engineering as a possible career (10th grade). Autumn has struggled with being labeled “a girl in the background” and someone whom her mother described in 7th grade as “if she would just get Ds, I would be happy.” However, over the past 6 years, Autumn has engaged in an ever-increasing range of STEM-rich actions and relationships, including building a Little Free STEM Library, leading workshops on energy efficiency, making educational movies for her community to teach about green energy, and writing for her afterschool STEM club blog. We are interested in Autumn’s engagement with others in these activities over time and space, and how they shape her own and her community’s engagement in STEM. The overarching question that guides this manuscript is: What are the interactional forces that operating across space and time that influence Autumn’s becoming in STEM, as a white girl, growing up in multi-generational poverty in a Midwestern city. Using Holland & Lave’s (2009) two forms of history –“history in person” and “history in institutionalized struggles”–we examined several pivotal events, and the micro-dynamics at play, identified by Autumn with respect to becoming in STEM. We sought to make sense of the ways in which Autumn’s STEM experiences were carried out in local practice but also enacted against the broader background of cultural/historical narratives. In this process we traced Autumn’s core commitments-in-practice in STEM & Community. We also examined how these core commitments-in-practice led, at times, to conflict and new forms of “contentious local practice” (LCP) as these commitments-in-practice pushed back against particular local, historical and sociocultural contexts. 

   more » « less
2. Becoming Somebody in STEM: Developing a cultural of criticality in the space between person and history

   Calabrese Barton, Angela ( April 2018 , American Education Research Association)

   Autumn, a young white woman growing up in multi-generational poverty in an economically challenged midwestern city, has authored a STEM-empowered life against the dominant sociohistorical narrative in American society. She has attended public schools that served significant populations living in high poverty – overcrowded classrooms, high teacher turnovers, out-of-field teaching, and limited STEM resources. Yet, she has authored herself into STEM despite these mitigating circumstances. Autumn is currently a high school senior with aspirations to become an engineer or a hairdresser working in an eco-salon. She spends part of her time after school in a makerspace housed in her local community center, building things to solve other people’s problems. She reminds us that her out-of-school efforts to participate in STEM exist worlds away from schooling. However, she takes the optimistic view that if she could tell teachers about her out-of-school STEM experiences, her teachers might be better able to help her and her peers serve the interests and needs of her community, as well as see Autumn’s potential in STEM. Now a rising 12th grader, we have followed Autumn since 5th grade through school and afterschool. As she grew older, she became more interested in helping us to document and tell her story. She is an author on this paper. Also during this time, Autumn has shifted from wanting to be a hair designer (5th grade) to wanting to own a “green” (environmentally friendly) hair salon (8th grade), to considering engineering as a possible career (10th grade). Autumn has struggled with being labeled “a girl in the background” and someone whom her mother described in 7th grade as “if she would just get Ds, I would be happy.” However, over the past 6 years, Autumn has engaged in an ever-increasing range of STEM-rich actions and relationships, including building a Little Free STEM Library, leading workshops on energy efficiency, making educational movies for her community to teach about green energy, and writing for her afterschool STEM club blog. We are interested in Autumn’s engagement with others in these activities over time and space, and how they shape her own and her community’s engagement in STEM. The overarching question that guides this manuscript is: What are the micro-dynamics that produce and challenge inequalities in Fall’s becoming in STEM, as a white girl, growing up in multi-generational poverty in a Midwestern city. Using Holland & Lave’s (2009) two forms of history –“history in person” and “history in institutionalized struggles”–we examined several pivotal events, and the micro-dynamics at play, identified by Autumn with respect to becoming in STEM. We sought to make sense of the ways in which Autumn’s STEM experiences were carried out in local practice but also enacted against the broader background of cultural/historical narratives. In this process we traced Autumn’s core commitments-in-practice in STEM & Community. We also examined how these core commitments-in-practice led, at times, to conflict and new forms of “contentious local practice” (LCP) as these commitments-in-practice pushed back against particular local, historical and sociocultural contexts. 

   more » « less
3. The Temple University Digital Pathology Corpus: The Breast Tissue Subset

   https://doi.org/10.1109/SPMB52430.2021.9672275  

   Wevodau, Z. ; Doshna, B. ; Jhala, N. ; Akhtar, I. ; Obeid, I. ; Picone, J. ( December 2021 , 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
4. “A Makerspace Is More Than Just a Room Full of Tools”: What Learning Looks Like for Female Students in Makerspaces

   https://doi.org/10.1115/DETC2018-86276  

   Tomko, Megan ; Schwartz, Amanda ; Newstetter, Wendy ; Alemán, Melissa ; Nagel, Robert ; Linsey, Julie ( August 2018 , : ASME 2018 International Design Engineering Technical Conferences and Computers and Information in Engineering Conference)

   Postulating that the act of making stimulates learning, a widespread effort prompted the integration of makerspaces on college campuses. From community colleges to research-based higher education institutions, large investments were and still are being made to advance the making spirit and encourage non-traditional learning in academic settings. While optimistic that students are taking advantage of the makerspace resources and are in fact learning from their experiences, there needs to be a more direct effort to understand the learning, if any, that is occurring in the makerspace. The makerspace is labeled as an open, learning environment where students are able to design, create, innovate, and collaborate [1, 2]. In response, we investigate the claims of this statement through the research question: how is learning experienced by female students in an academic makerspace? Female students in STEM, especially those engaged in makerspaces, have unique and uncharacteristic experiences that can lend way to various learning and pedagogical implications. The purpose of this paper is to highlight our methodological process for incorporating in-depth phenomenologically based interviewing and for utilizing open and axial coding methods to establish grounded theory. We interview five female students through purposeful maximum variation sampling and snowball sampling. Through a rigorous and iterative data analysis process of the ten-percent of the overall, we created a preliminary coding scheme that articulates how learning is occurring, what design skills are being learned, and what life skills are being learned. These preliminary findings show that not only are these female students learning by doing and learning how to problem solve in design, but they are also overcoming fears, developing patience, and communicating ideas in these design-oriented makerspaces. 

   more » « less
5. “I Understand Their Frustrations a Little Bit Better”: Elementary Teachers’ Affective Stances in Engineering in an Online Learning Program

   https://doi.org/10.18260/1-2--33969  

   Portsmore, Merredith ; Watkins, Jessica ; Swanson, Rebecca ( January 2020 , ASEE annual conference)

   As K-12 engineering education becomes more ubiquitous in the U.S, increased attention has been paid to preparing the heterogeneous group of in-service teachers who have taken on the challenge of teaching engineering. Standards have emerged for professional development along with research on teacher learning in engineering that call for teachers to facilitate and support engineering learning environments. Given that many teachers may not have experienced engineering practice calls have been made to engage teaches K-12 teachers in the “doing” of engineering as part of their preparation. However, there is a need for research studying more specific nature of the “doing” and the instructional implications for engaging teachers in “doing” engineering. In general, to date, limited time and constrained resources necessitate that many professional development programs for K-12 teachers to engage participants in the same engineering activities they will enact with their students. While this approach supports teachers’ familiarity with curriculum and ability to anticipate students’ ideas, there is reason to believe that these experiences may not be authentic enough to support teachers in developing a rich understanding of the “doing” of engineering. K-12 teachers are often familiar with the materials and curricular solutions, given their experiences as adults, which means that engaging in the same tasks as their students may not be challenging enough to develop their understandings about engineering. This can then be consequential for their pedagogy: In our prior work, we found that teachers’ linear conceptions of the engineering design process can limit them from recognizing and supporting student engagement in productive design practices. Research on the development of engineering design practices with adults in undergraduate and professional engineering settings has shown significant differences in how adults approach and understand problems. Therefore, we conjectured that engaging teachers in more rigorous engineering challenges designed for adult engineering novices would more readily support their developing rich understandings of the ways in which professional engineers move through the design process. We term this approach meaningful engineering for teachers, and it is informed by work in science education that highlights the importance of learning environments creating a need for learners to develop and engage in disciplinary practices. We explored this approach to teachers’ professional learning experiences in doing engineering in an online graduate program for in-service teachers in engineering education at Tufts University entitled the Teacher Engineering Education Program (teep.tufts.edu). In this exploratory study, we asked: 1. How did teachers respond to engaging in meaningful engineering for teachers in the TEEP program? 2. What did teachers identify as important things they learned about engineering content and pedagogy? This paper focuses on one theme that emerged from teachers’ reflections. Our analysis found that teachers reported that meaningful engineering supported their development of epistemic empathy (“the act of understanding and appreciating someone's cognitive and emotional experience within an epistemic activity”) as a result of their own affective experiences in doing engineering that required significant iteration as well as using novel robotic materials. We consider how epistemic empathy may be an important aspect of teacher learning in K-12 engineering education and the potential implications for designing engineering teacher education. 

   more » « less