More Like this

1. Differentiation Between Humic and Non-Humic Substances Using Alkaline Extraction and Ultraviolet Spectroscopy

   https://doi.org/10.1093/jaoacint/qsad001  

   Mayhew, Lawrence ; Singh, Amit Pratap ; Li, Peng ; Perdue, E. Michael ( January 2023 , Journal of AOAC INTERNATIONAL)

   Although humic substances are the principal ingredients in processed humic products, there has been no practical way to determine if a material is humified, allowing fake products to be used by farmers instead of genuine humic substances.

   To develop a test method using conventional laboratory techniques to determine if a material is humified.

   A neutralized extract is prepared using the standardized extraction protocols specified in ISO 19822:2018(E). A portion of the extract is used to determine the concentration of dissolved organic matter on an ash-free basis. A portion of the remaining neutralized extract is diluted to a concentration of 30 mg/kg of dissolved organic matter and transferred to a quartz UV cuvette for ultraviolet-visible (UV-Vis) spectroscopy. UV-Vis absorbance is recorded over a wavelength range of 220–500 nm at 5 nm intervals. The absorbance data are normalized by conversion to scaled absorbance, which is compared to a reference scaled absorbance spectral curve for humic substances to determine if the tested material is humic or non-humic.

   This method was able to differentiate legitimate humic substances from non-humic adulterants in a multiple-laboratory validation study (P ≤ 0.05).

   This method can differentiate humic from non-humic substances in materials intended to be used as ingredients in commercial humic productsmore »or for research.

   This method uses common laboratory procedures and equipment.

   « less
2. Inorganic salts and organic matter effects on nanorod, nanowire, and nanoplate MoO ₃ aggregation, dissolution, and photocatalysis

   https://doi.org/10.1039/D0EN00708K  

   Fanourakis, Sofia K. ; Peña-Bahamonde, Janire ; Rodrigues, Debora F. ( December 2020 , Environmental Science: Nano)

   Use of visible light photocatalytic nanomaterials in water treatment can be promising in treating contaminants. However, little research has been conducted examining the effects of more complex chemistries in the nanomaterial's performance. In this work, the effects of inorganic salts (NaCl and CaCl 2 ) and natural organic matter (NOM) such as humic acid (HA) and extracellular polymeric substances (EPS) on nanoparticle aggregation, dissolution, and ultimately on the photocatalytic properties of molybdenum trioxide (MoO 3 ), i.e. nanorods, nanowires, and nanoplates were examined. In the presence of NaCl, nanorod, nanowire, and nanoplate MoO 3 had similar critical coagulation concentrations, while the nanorods showed higher instability in CaCl 2 . Overall, the presence of inorganic salts caused high colloidal instability in the MoO 3 nanostructures in terms of aggregation behavior, but greatly aided in the reduction of MoO 3 dissolution. NOM presence decreased aggregation rates, albeit dissolution was not similarly affected in all three structures. Only the dissolution of the nanowire structures was reduced in the presence of HA or EPS. Furthermore, the photocatalytic activity of the nanowires and nanoplates was overall reduced when inorganic salts or natural organic matter were present. While the presence of natural organic matter alone didmore »reduce photocatalytic effectiveness of the nanorod MoO 3 , the presence of salts seemed to negate the effects from the organic compounds. Furthermore, the presence of CaCl 2 resulted in a highly enhanced photocatalytic activity regardless of the presence of natural organic matter. The structural and chemical differences of the nanomaterials played a significant role in their aggregation, dissolution, and ability to photocatalytically degrade methylene blue in solution. This study demonstrates that a better understanding of water chemistry effects on nanomaterials is essential prior to their intended applications.« less
3. Proceedings of the 11th International Conference on Porous Metals and Metallic Foams (MetFoam 2019)

   https://doi.org/10.1007/978-3-030-42798-6  

   A Abaidi, Abou Horeira ( January 2019 , Proceedings of the 11th International Conference on Porous Metals and Metallic Foams (MetFoam 2019))

   MetFoam 2019 is the 11th edition of the biannual International Conference on Porous Metals and Metallic Foams; it was heled in Dearborn, Michigan, USA between August 20 and 23, 2019. Previous editions were held in various cities around the world starting in Bremen, Germany in 1999. This conference is the largest cross-disciplinary, international technical meeting focused exclusively on the production, properties, and applications of these lightweight multifunctional porous materials. In Dearborn, a total of 115 presentations were delivered at the conference by participants from 17 countries. Many participants travelled long distances to present and attend the conference. Students delivered 40% of the presentations. Emerging areas such as additive manufacturing and freeze casting, as well as nanoporous materials, cellular materials, and metallic metamaterials were covered by some of the presenters. The proceedings volume of MetFoam 2019 is the culmination of several months of work, which included the preparation of the papers by the authors, editing, and reviewing. The papers collected in this volume provide a representative snapshot of the research activity in the field. It is my sincere hope that this proceedings volume will remain a valuable record of MetFoam 2019, and that it will serve as a reference for researchersmore »interested in porous metals and metallic foams. I would like to thank the reviewers and the session chairs for volunteering their precious time and effort. The International Scientific Board and the Steering Committee of MetFoam 2019 provided indispensable input and guidance throughout the planning of the conference. Thanks are also due to Kelcy Wagner and Trudi Dunlap from our organization partner The Minerals, Metals & Materials Society (TMS) for their assistance in the production of the proceedings volume, as well as for their vital hard work during the organization phase of the conference. Certainly, a special thanks go to our sponsors and exhibitors for making the event possible.« less
4. 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 notmore »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.« less
5. Scientific Cooperation: Supporting Circumpolar Permafrost Monitoring and Data Sharing

   https://doi.org/10.3390/land10060590  

   Bouffard, Troy J. ; Uryupova, Ekaterina ; Dodds, Klaus ; Romanovsky, Vladimir E. ; Bennett, Alec P. ; Streletskiy, Dmitry ( June 2021 , Land)

   null (Ed.)

   While the world continues to work toward an understanding and projections of climate change impacts, the Arctic increasingly becomes a critical component as a bellwether region. Scientific cooperation is a well-supported narrative and theme in general, but in reality, presents many challenges and counter-productive difficulties. Moreover, data sharing specifically represents one of the more critical cooperation requirements, as part of the “scientific method [which] allows for verification of results and extending research from prior results”. One of the important pieces of the climate change puzzle is permafrost. In general, observational data on permafrost characteristics are limited. Currently, most permafrost data remain fragmented and restricted to national authorities, including scientific institutes. The preponderance of permafrost data is not available openly—important datasets reside in various government or university labs, where they remain largely unknown or where access restrictions prevent effective use. Although highly authoritative, separate data efforts involving creation and management result in a very incomplete picture of the state of permafrost as well as what to possibly anticipate. While nations maintain excellent individual permafrost research programs, a lack of shared research—especially data—significantly reduces effectiveness of understanding permafrost overall. Different nations resource and employ various approaches to studying permafrost, including the growingmore »complexity of scientific modeling. Some are more effective than others and some achieve different purposes than others. Whereas it is not possible for a nation to effectively conduct the variety of modeling and research needed to comprehensively understand impacts to permafrost, a global community can. In some ways, separate scientific communities are not necessarily concerned about sharing data—their work is secured. However, decision and policy makers, especially on the international stage, struggle to understand how best to anticipate and prepare for changes, and thus support for scientific recommendations during policy development. To date, there is a lack of research exploring the need to share circumpolar permafrost data. This article will explore the global data systems on permafrost, which remain sporadic, rarely updated, and with almost nothing about the subsea permafrost publicly available. The authors suggest that the global permafrost monitoring system should be real time (within technical and reasonable possibility), often updated and with open access to the data (general way of representing data required). Additionally, it will require robust co-ordination in terms of accessibility, funding, and protocols to avoid either duplication and/or information sharing. Following a brief background, this article will offer three supporting themes, (1) the current state of permafrost data, (2) rationale and methods to share data, and (3) implications for global and national interests.« less