More Like this

1. 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
2. Enhancing understanding of the hydrological cycle via pairing of process‐oriented and isotope ratio tracers

   https://doi.org/10.5061/dryad.8w9ghx3mr  

   Fiorella, Richard ; Siler, Nicholas ; Nusbaumer, Jesse ; Noone, David ( January 2021 , Dryad)

   This dataset contains monthly average output files from the iCAM6 simulations used in the manuscript "Enhancing understanding of the hydrological cycle via pairing of process-oriented and isotope ratio tracers," in review at the Journal of Advances in Modeling Earth Systems. A file corresponding to each of the tagged and isotopic variables used in this manuscript is included. Files are at 0.9° latitude x 1.25° longitude, and are in NetCDF format. Data from two simulations are included: 1) a simulation where the atmospheric model was "nudged" to ERA5 wind and surface pressure fields, by adding an additional tendency (see section 3.1 of associated manuscript), and 2) a simulation where the atmospheric state was allowed to freely evolve, using only boundary conditions imposed at the surface and top of atmosphere. Specific information about each of the variables provided is located in the "usage notes" section below. Associated article abstract: The hydrologic cycle couples the Earth's energy and carbon budgets through evaporation, moisture transport, and precipitation. Despite a wealth of observations and models, fundamental limitations remain in our capacity to deduce even the most basic properties of the hydrological cycle, including the spatial pattern of the residence time (RT) of water in the atmosphere and the mean distance traveled from evaporation sources to precipitation sinks. Meanwhile, geochemical tracers such as stable water isotope ratios provide a tool to probe hydrological processes, yet their interpretation remains equivocal despite several decades of use. As a result, there is a need for new mechanistic tools that link variations in water isotope ratios to underlying hydrological processes. Here we present a new suite of “process-oriented tags,” which we use to explicitly trace hydrological processes within the isotopically enabled Community Atmosphere Model, version 6 (iCAM6). Using these tags, we test the hypotheses that precipitation isotope ratios respond to parcel rainout, variations in atmospheric RT, and preserve information regarding meteorological conditions during evaporation. We present results for a historical simulation from 1980 to 2004, forced with winds from the ERA5 reanalysis. We find strong evidence that precipitation isotope ratios record information about atmospheric rainout and meteorological conditions during evaporation, but little evidence that precipitation isotope ratios vary with water vapor RT. These new tracer methods will enable more robust linkages between observations of isotope ratios in the modern hydrologic cycle or proxies of past terrestrial environments and the environmental processes underlying these observations.   Details about the simulation setup can be found in section 3 of the associated open-source manuscript, "Enhancing understanding of the hydrological cycle via pairing of process‐oriented and isotope ratio tracers." In brief, we conducted two simulations of the atmosphere from 1980-2004 using the isotope-enabled version of the Community Atmosphere Model 6 (iCAM6) at 0.9x1.25° horizontal resolution, and with 30 vertical hybrid layers spanning from the surface to ~3 hPa. In the first simulation, wind and surface pressure fields were "nudged" toward the ERA5 reanalysis dataset by adding a nudging tendency, preventing the model from diverging from observed/reanalysis wind fields. In the second simulation, no additional nudging tendency was included, and the model was allowed to evolve 'freely' with only boundary conditions provided at the top (e.g., incoming solar radiation) and bottom (e.g., observed sea surface temperatures) of the model. In addition to the isotopic variables, our simulation included a suite of 'process-oriented tracers,' which we describe in section 2 of the manuscript. These variables are meant to track a property of water associated with evaporation, condensation, or atmospheric transport. Metadata are provided about each of the files below; moreover, since the attached files are NetCDF data - this information is also provided with the data files. NetCDF metadata can be accessed using standard tools (e.g., ncdump). Each file has 4 variables: the tagged quantity, and the associated coordinate variables (time, latitude, longitude). The latter three are identical across all files, only the tagged quantity changes. Twelve files are provided for the nudged simulation, and an additional three are provided for the free simulations: Nudged simulation files iCAM6_nudged_1980-2004_mon_RHevap: Mass-weighted mean evaporation source property: RH (%) with respect to surface temperature. iCAM6_nudged_1980-2004_mon_Tevap: Mass-weighted mean evaporation source property: surface temperature in Kelvin iCAM6_nudged_1980-2004_mon_Tcond: Mass-weighted mean condensation property: temperature (K) iCAM6_nudged_1980-2004_mon_columnQ: Total (vertically integrated) precipitable water (kg/m2).  Not a tagged quantity, but necessary to calculate depletion times in section 4.3 (e.g., Fig. 11 and 12). iCAM6_nudged_1980-2004_mon_d18O: Precipitation d18O (‰ VSMOW) iCAM6_nudged_1980-2004_mon_d18Oevap_0: Mass-weighted mean evaporation source property - d18O of the evaporative flux (e.g., the 'initial' isotope ratio prior to condensation), (‰ VSMOW) iCAM6_nudged_1980-2004_mon_dxs: Precipitation deuterium excess (‰ VSMOW) - note that precipitation d2H can be calculated from this file and the precipitation d18O as d2H = d-excess - 8*d18O. iCAM6_nudged_1980-2004_mon_dexevap_0: Mass-weighted mean evaporation source property - deuterium excess of the evaporative flux iCAM6_nudged_1980-2004_mon_lnf: Integrated property - ln(f) calculated from the constant-fractionation d18O tracer (see section 3.2). iCAM6_nudged_1980-2004_mon_precip: Total precipitation rate in m/s. Note there is an error in the metadata in this file - it is total precipitation, not just convective precipitation. iCAM6_nudged_1980-2004_mon_residencetime: Mean atmospheric water residence time (in days). iCAM6_nudged_1980-2004_mon_transportdistance: Mean atmospheric water transport distance (in km). Free simulation files iCAM6_free_1980-2004_mon_d18O: Precipitation d18O (‰ VSMOW) iCAM6_free_1980-2004_mon_dxs: Precipitation deuterium excess (‰ VSMOW) - note that precipitation d2H can be calculated from this file and the precipitation d18O as d2H = d-excess - 8*d18O. iCAM6_free_1980-2004_mon_precip: Total precipitation rate in m/s. Note there is an error in the metadata in this file - it is total precipitation, not just convective precipitation. 

   more » « less
3. Relating referential clarity and phonetic clarity in infant‐directed speech

   https://doi.org/10.1111/desc.13442  

   Beech, Caroline ; Swingley, Daniel ( August 2023 , Developmental Science)

   Psycholinguistic research on children's early language environments has revealed many potential challenges for language acquisition. One is that in many cases, referents of linguistic expressions are hard to identify without prior knowledge of the language. Likewise, the speech signal itself varies substantially in clarity, with some productions being very clear, and others being phonetically reduced, even to the point of uninterpretability. In this study, we sought to better characterize the language‐learning environment of American English‐learning toddlers by testing how well phonetic clarity and referential clarity align in infant‐directed speech. Using an existing Human Simulation Paradigm (HSP) corpus with referential transparency measurements and adding new measures of phonetic clarity, we found that the phonetic clarity of words’ first mentions significantly predicted referential clarity (how easy it was to guess the intended referent from visual information alone) at that moment. Thus, when parents’ speech was especially clear, the referential semantics were also clearer. This suggests that young children could use the phonetics of speech to identify globally valuable instances that support better referential hypotheses, by homing in on clearer instances and filtering out less‐clear ones. Such multimodal “gems” offer special opportunities for early word learning.

   In parent‐infant interaction, parents’ referential intentions are sometimes clear and sometimes unclear; likewise, parents’ pronunciation is sometimes clear and sometimes quite difficult to understand.

   We find that clearer referential instances go along with clearer phonetic instances, more so than expected by chance.

   Thus, there are globally valuable instances (“gems”) from which children could learn about words’ pronunciations and words’ meanings at the same time.

   Homing in on clear phonetic instances and filtering out less‐clear ones would help children identify these multimodal “gems” during word learning.

    

   more » « less
4. Speech sounds in larger inventories are not (necessarily) less variable

   https://doi.org/10.1121/10.0014912  

   Hauser, Ivy ( November 2022 , The Journal of the Acoustical Society of America)

   This paper reconsiders a classic claim about phonetic variability—that speech sounds in larger phonemic inventories should exhibit less within-category variability in production. Although this hypothesis is intuitive, existing literature provides limited unqualified support for the claim, further complicated by the fact that null results (like those failing to find a difference in variability between languages) often go unpublished. Even so, existing work suggests that factors contributing to extent of variability are multifaceted. While phonological contrast may affect variability patterns, inventory size alone is not a reliable predictor of variability differences. This paper reviews relevant findings in the literature, presents an additional case study, and argues for more nuanced alternatives to account for cross-linguistic differences in extent of phonetic variability.

    

   more » « less
5. Does Photomineralization of Dissolved Organics Matter in Temperate Rivers?

   https://doi.org/10.1029/2021JG006402  

   Maavara, Taylor ; Logozzo, Laura ; Stubbins, Aron ; Aho, Kelly ; Brinkerhoff, Craig ; Hosen, Jacob ; Raymond, Peter ( July 2021 , Journal of Geophysical Research: Biogeosciences)

   Sunlight can oxidize dissolved organic carbon (DOC) to dissolved inorganic carbon (DIC) in freshwaters. The importance of complete photooxidation, or photomineralization, as a sink for DOC remains unclear in temperate rivers, as most estimates are restricted to lakes, high latitude rivers, and coastal river plumes. In this study, we construct a model representing over 75,000 river reaches in the Connecticut River Watershed (CRW), USA, to calculate spectrally resolved photomineralization. We test the hypothesis that photomineralization is a negligible DOC sink across all reaches and flow conditions relative to DOC fluxes. Our model quantifies reaction rates and transport drivers within the river reaches for the ranges of flow conditions, incoming solar irradiance, and canopy cover shading observed throughout the year. Our model predicts average daily areal photomineralization rates ranging from 1.16 mg‐C m⁻² day⁻¹in low flow river reaches in the winter, to 18.33 mg‐C m⁻² day⁻¹in high flow river reaches during the summer. Even for high photomineralization fluxes, corresponding photomineralization uptake velocities are typically at least an order of magnitude smaller than those reported for other instream processes. We calculate DOC elimination by photomineralization relative to DOC fluxes through individual stream reaches as well as the entire riverine portion of the CRW. We find that relative photomineralization fluxes are highest in summer drought conditions in low order streams. In median flows and mean light intensities, for an average watershed travel distance, 3%–5% of the DOC fluxes are eliminated, indicating that photomineralization is a minor DOC sink in temperate rivers.

    

   more » « less