Leveraging Multi-Task Learning to Cope With Poor and Missing Labels of Mammograms

In breast cancer screening, binary classification of mammograms is a common task aiming to determine whether a case is malignant or benign. A Computer-Aided Diagnosis (CADx) system based on a trainable classifier requires clean data and labels coming from a confirmed diagnosis. Unfortunately, such labels are not easy to obtain in clinical practice, since the histopathological reports of biopsy may not be available alongside mammograms, while normal cases may not have an explicit follow-up confirmation. Such ambiguities result either in reducing the number of samples eligible for training or in a label uncertainty that may decrease the performances. In this work, we maximize the number of samples for training relying on multi-task learning. We design a deep-neural-network-based classifier yielding multiple outputs in one forward pass. The predicted classes include binary malignancy, cancer probability estimation, breast density, and image laterality. Since few samples have all classes available and confirmed, we propose to introduce the uncertainty related to the classes as a per-sample weight during training. Such weighting prevents updating the network's parameters when training on uncertain or missing labels. We evaluate our approach on the public INBreast and private datasets, showing statistically significant improvements compared to baseline and independent state-of-the-art approaches. Moreover, we use mammograms from Susan G. Komen Tissue Bank for fine-tuning, further demonstrating the ability to improve the performances in our multi-task learning setup from raw clinical data. We achieved the binary classification performance of AUC = 80.46 on our private dataset and AUC = 85.23 on the INBreast dataset.

Implantation of bovine hydroxyapatite and secretome with different oxygen concentration may improve massive bone defect regeneration: An experimental study on animal model

The most widely used biomaterials in the treatment of massive bone defects are allograft bone or metal implants. The current problem is that the availability of allographs is limited and metal implants are very expensive. Mass production of secretome can make bone reconstruction of massive bone defects using a scaffold more effective and efficient. This study aims to prove bone regeneration in massive bone defects using bovine hydroxyapatite reconstruction with normoxic and hypoxic secretome conditions using collagen type 1 (COL1), alkaline phosphate (ALP), osteonectin (ON), and osteopontin (OPN) parameters. This is an in vivo study using male New Zealand white rabbits aged 6–9 months. The research was carried out at the Biomaterials Center—Tissue Bank, Dr. Soetomo Hospital for the manufacturer of bovine hydroxyapatite (BHA) and secretome BM-MSC culture under normoxic and hypoxic conditions, and UNAIR Tropical Disease Institute for implantation in experimental animals. Data analysis was carried out with the one-way ANOVA statistical test and continued with the Post Hoc test LSD statistical test to determine whether or not there were significant differences between groups. There were significant differences between hypoxic to normoxic group and hypoxic to BHA group at day-30 observation using ALP, COL 1, ON, and OPN parameters. Meanwhile, there is only osteonectin parameter has significant difference at day-30 observation. At day-60 observation, only OPN parameter has significant differences between hypoxic to normoxic and hypoxic to BHA group. Between day-30 and day-60 observation, BHA and normoxic groups have a significant difference at all parameters, but in hypoxic group, there are only difference at ALP, COL 1, and ON parameters. Hypoxic condition BM-MSC secretome with BHA composite is superior and could be an option for treating bone defect.

Assessment of Cognitive Symptoms in Brain Bank-Registered Control Subjects: Feasibility and Utility of a Telephone-Based Screening

Background: For neuroscience research, the study of brain tissue of neurologically unimpaired subjects is crucial to interpret findings in neurodegenerative diseases. Sub-optimal neurological follow-up and the presence of neuropathological lesions in supposedly asymptomatic subjects casts doubt as to whether these subjects present an undetected underlying neurodegenerative disease or are resilient to neurodegeneration. Objective: We aimed to assess whether the control donors registered in the Neurological Tissue Bank-Hospital Clínic-IDIBAPS (NTB-HCI) are still free of cognitive symptoms at follow-up and to evaluate the feasibility and utility of a telephone-based screening. Methods: All control subjects older than 65 years registered at the NTB-HCI database were selected for the study. After a structured telephone interview, those subjects already diagnosed with a neurological disease were excluded. Then, a cognitive screening was performed, including the telephone version of the Mini-Mental State Examination (t-MMSE) and the eight-item interview (AD-8) to the subject and to one informant. Results: In total, 73.8% of the registered donors collaborated in the study. Only 21.4% had at least one of the three cognitive screening tools impaired, and 2.7% had a profile highly suggestive of cognitive impairment. AD-8i correlated moderately with t-MMSE. Conclusion: Telephone-based neurologic screening in control donors is feasible and was within the normal range in most of the subjects in our cohort. Albeit, the involvement of neurologists and periodic neurological screenings are desirable in a control subjects brain donor program, AD8-i could be used to screen the control’s neurological status in the absence of accurate clinical data at the time of the death.

Applying Grounded Theory to Understand Asian Women’s Motivations to Donate Healthy Breast Tissue

Background: The Komen Tissue Bank (KTB) is a clinical trial and the only global biorepository that collects and stores healthy breast tissue to be used as controls in breast cancer (BC) research. Due to a variety of barriers, there is a lack of participation by racial and ethnic minority women in tissue donation. In order to increase this participation, it is necessary to understand why or why not these populations choose to participate in clinical trials such as the KTB. This study used grounded theory methodology to explore the motivations behind Asian women’s decisions to donate their breast tissue. Methods: Guided by grounded theory, we conducted interviews with previous breast tissue donors who self-identified as Asian (n=20). We then transcribed and coded the interviews to discover common attitudes and motivations for participating in breast tissue donation. Findings: Preliminary findings were obtained from 11 interviews. We identified three common themes that influenced these women’s donations: altruistic behavior, comfort with science, and Asian identity. Identified sub-themes include factors such as personal ties to BC and background in research and clinical trials. It is of note that over half of the women expressed Asian identity and comfort with science as important factors, and all mentioned altruistic tendencies, either towards family or towards research and others. Conclusion and Future Work: We identified common factors for donating healthy breast tissue from using grounded theory to interview previous donors of Asian descent. We will transcribe and code 9 more interviews, as well as use those interviews to confirm theoretical saturation. The findings from this study will be used in the future to inform a framework for developing recruitment strategies to increase overall participation of historically excluded individuals in the KTB. Future work will include exploring the motivations of Latinas regarding donating their healthy breast tissue.

Analysis of Cryopreservation Protocols and Their Harmful Effects on the Endothelial Integrity of Human Corneas

Corneal cryopreservation can partially solve the worldwide concern regarding donor cornea shortage for keratoplasties. In this study, human corneas were cryopreserved using two standard cryopreservation protocols that are employed in the Tissue Bank of the Teresa Herrera Hospital (Spain) to store corneas for tectonic keratoplasties (TK protocol) and aortic valves (AV protocol), and two vitrification protocols, VS55 and DP6. Endothelial viability and general corneal state were evaluated to determine the protocol that provides the best results. The potential corneal cryopreservation protocol was studied in detail taking into consideration some cryopreservation-related variables and the endothelial integrity and stroma arrangement of the resulting cryopreserved corneas. TK corneas showed mostly viable endothelial cells, while the others showed few (AV) or none (DP6 and VS55). The corneal structure was well maintained in TK and AV corneas. TK corneas showed endothelial acellular areas surrounded by injured cells and a normal-like stromal fiber arrangement. Cryoprotectant solutions of the TK protocol presented an increasing osmolality and a physiological pH value. Cooling temperature rate of TK protocol was of 1 °C/min to −40 °C and 3 °C/min to −120 °C, and almost all of dimethyl sulfoxide left the tissue after washing. Future studies should be done changing cryopreservation-related variables of the TK protocol to store corneas of optical grade.

Immune Profiling of Responses to Influenza Vaccination in Patients with Myeloproliferative Neoplasms

Myeloproliferative neoplasms (MPNs) are clonal stem cell neoplasms that include polycythemia vera (PV), essential thrombocythemia (ET), and myelofibrosis (MF). MPN patients are immunocompromised due to immune dysregulation from clonal hematopoiesis, heightened inflammation, and immunosuppressive treatments. Infection is a leading cause of morbidity and mortality in MPNs, and as a result annual influenza vaccination is a critical component of care. However, efficacy of the influenza vaccine in MPN patients is unclear. This study sought to identify differences between MPN patients and healthy donors in their immune response to influenza vaccination and subsequent in vitro stimulation. The Massachusetts General Hospital (MGH) maintains a tissue bank of peripheral blood and bone marrow samples collected in consented patients with hematologic malignancy, including MPNs. The tissue bank was interrogated for MPN patients with peripheral blood samples banked 1-6 months after yearly trivalent/quadrivalent influenza vaccination. A total of 13 MPN patients were included, with a median age of 70 (32-86). Five patients had MF, 3 patients had PV, 4 patients had ET, and 1 patient had pre-fibrotic MF. Five patients were on treatment with ruxolitinib, 4 patients were on hydroxyurea (HU), and 4 patients were without treatment. Peripheral blood mononuclear cells (PBMCs) from 11 healthy donors 3-6 months post-vaccination, with a median age of 50 (28-56), were sourced commercially as controls. Cryopreserved samples were thawed and PBMC samples divided equally prior to incubation for 18 hours with peptides derived from Influenza A or sham. Subsequently samples were labeled with a 41-antibody panel and analyzed by mass cytometry. Manual gating together with FlowSOM clustering followed by EdgeR using T-test and FDR to correct for multiple comparisons were used to identify differentially abundant immune cell clusters between cohorts and treatment conditions. Without stimulation, MPN patients were decreased in B cell and memory CD4+ T cell populations as compared to healthy donors (p<0.05), in agreement with a previous report. MPN patients also exhibited an increase in a subset of mature B cells (p<0.05) and lineage negative CD123+ cells (p<0.05) as compared to controls. Stimulation included in this study aimed to expand on these distinctions and identified that baseline differences were retained with additional increases in CD4+ and CD8+ T cell responses in healthy donors as compared to MPN patients. Conversely, stimulation increased the frequency of inflammatory monocytic cells in MPN patients compared to healthy controls (p<0.05). In healthy donors stimulation induced several alterations in, most notably an increase in CD4+ effector memory cells (p<0.03), as compared to matched unstimulated controls. In contrast, comparison of pre- and post-stimulation within the MPN cohort revealed no significant differences in immune cell populations apart from a decrease in naïve CD4+ T cells (p<0.05). Patients receiving HU were enriched for naïve CD4+ and effector CD8+ T cells as compared to patients receiving ruxolitinib (p<0.05); however, these differences were diminished following stimulation. Within MPN patients, pre-stimulation samples in MF patients were characterized by increased abundance of activated monocyte/macrophage clusters and a lineage-negative CD123+ population; there were no significant differences in post-stimulation samples between MF and ET/PV/pre-fibrotic MF patients. We evaluated immunologic responses to influenza vaccination, with and without peptide stimulation, in MPN patients and compared to healthy controls. Overall, MPN patients demonstrated less robust B- and T-cell responses and a muted response to stimulation as compared to healthy controls, which may indicate reduced cellular responses following influenza vaccination. Ongoing analysis will interrogate specific immune populations and subsets in detail to refine current observations and will evaluate associations between immunologic cell clusters with serologic responses and additional patient characteristics. The results of our study highlight important alterations in the immune response to influenza vaccination and muted response to stimulation in MPN patients and will inform a future prospective study of vaccine responsiveness and immune dysfunction in patients with MPNs. Figure 1 Figure 1. Disclosures Reeves: PR Fludigm: Honoraria; Immunoscape: Consultancy. Hobbs: Merck: Research Funding; Incyte Corporation: Research Funding; AbbVie.: Consultancy; Novartis: Consultancy; Celgene/Bristol Myers Squibb: Consultancy; Constellation Pharmaceuticals: Consultancy, Research Funding; Bayer: Research Funding.

High Plasma Apolipoprotein(a) Concentration and Low Plasmin Tpa Enzymatic Activity in Hospitalized Patients with COVID-19

Thrombotic and thromboembolic complications in patients diagnosed with coronavirus disease 2019 (COVID-19) are emerging as important sequelae that contribute to mortality, including disseminated intravascular coagulation, pulmonary embolism, deep vein thrombosis, ischemic stroke, and myocardial infraction. Reported incidence of thrombotic and thromboembolic complications in moderate/severe COVID-19 patients is from 21% to 49%, while even higher incidence in non-surviving COVID-19 patients. However, the underlying mechanism between thrombosis and COVID-19 is still unclear. Tissue-type plasminogen activator (tPA) plays an important role on initiating fibrinolysis by converting zymogen plasminogen to plasmin, a serine protease that degrades the fibrin clot, and therefore preventing excessive pathological blood clots. A homologous protein to plasminogen is apolipoprotein(a) [apo(a)], a major component of lipoprotein(a). The apo(a) inhibits fibrinolysis and exacerbates thrombosis through blocking the conversion from Glu-plasminogen to Lys-plasminogen, which has a higher binding affinity to fibrin and is a better substrate to tPA. The population distribution of plasma apo(a) level is positively skewed (most values are clustered around the left tail of the distribution close to zero), and the plasma apo(a) level in most people is less than 300 μg/mL. High plasma concentration of apo(a) (> 300 μg/mL), or genetic variants of LPA, the gene that encodes for apo(a), correlates with thrombotic cardiovascular risk and thromboembolic risk in many population-based clinical or genetic studies. To investigate the potential correlation between infection of SARS-COV-2 and thrombosis, we tested de-identified plasma samples collected from hospitalized patients with or without positivity of SARS-CoV-2 testing results and COVID-19 diagnosis (ICD10CM:U07.1) through the COVD-19 Tissue Bank at the Medical College of Wisconsin. The tPA enzymatic activity was measured by the release of p-nitroaniline chromophore from a plasmin-specific synthetic substrate with exogenous human plasminogen, with the intensity of color proportional to tPA activity. The apo(a) concentration is measured by ELISA capturing total apo(a) antigen. Our results show that the SARS-CoV-2-positive inpatients have higher plasma tPA concentration than the SARS-CoV-2 negative inpatients (6.0 versus 3.0 ng/mL, p<0.05), while plasma tPA enzymatic activity is lower in SARS-CoV-2-positive inpatients than the SARS-CoV-2 negative inpatients (15.2 versus 25.5 ΔA/min/mL/10 4, p<0.0001) (Figure A). The plasma apo(a) concentration is significantly higher in SARS-CoV-2-positive inpatients than in the plasma from SARS-CoV-2-negative inpatients (the median of the two groups are 114.8 versus 34.4 μg/mL, p<0.05) (Figure B). Among the 20 hospitalized patients with COVID-19, 13 survived. The 13 survived patients have one additional plasma sample collected after recovering from COVID-19 (date range between the two blood collections of onset and after recovery is from 19 to 87 days, the mean duration is 42 days). After recovery, 11 out of 13 surviving patients have increased plasma tPA enzymatic activity (the mean value at onset versus recovery is 5.2 versus 7.1 ΔA/min/mL/10 4, p<0.05) (Figure C). Consistently, 11 out of 13 surviving patients have decreased plasma apo(a) concentration compared to the plasma collected during the onset of COVID-19 from the same individuals (the median values of the onset and recovery are 141.1 versus 106.5 μg/mL, p<0.001) (Figure D). In summary, our study shows lower tPA enzymatic activity and higher apo(a) concentration in SARS-CoV-2-positive hospitalized patients compared to SARS-CoV-2-negative hospitalized patients. Among the survived patients, the reduction of apo(a) concentration after recovering from COVID-19 is accordance with the increase of tPA enzymatic activity. Considering the role of apo(a) in inhibiting fibrinolysis through limiting tPA-mediated plasminogen to plasmin conversion, the alteration in apo(a) concentration provide a possible explanation of change of tPA activity in patients with severe COVID-19. Figure 1 Figure 1. Disclosures Baumann Kreuziger: CSL Behring: Consultancy; Quercegen Pharmaceuticals: Consultancy; Vaccine Injury Compensation Program: Consultancy.

ESTABLISHMENT OF A SPECIMEN/TISSUE BANK AND ASSOCIATED DNA REFERENCE DATA FOR eDNA ANALYSIS v1

This protocol is intended to provide guidelines on the curation and establishment of a specimen/tissue bank and associated DNA sequence data to be used as reference material/data for subsequent environmental DNA (eDNA) analysis, with particular emphasis on marine non-indigenous and invasive species.

Chronic Porphyromonas gingivalis lipopolysaccharide induces adverse myocardial infarction wound healing through activation of CD8+ T-cells

Oral and gum health have long been associated with incidence and outcomes of cardiovascular disease. Periodontal disease increases myocardial infarction (MI) mortality by seven-fold through mechanisms that are not fully understood. The goal of this study was to evaluate whether lipopolysaccharide (LPS) from a periodontal pathogen accelerates inflammation post-MI through memory T-cell activation. We compared 4 groups (no MI, chronic LPS, day 1 post-MI, and day 1 post-MI with chronic LPS (LPS+MI); n=68 mice) using the mouse heart attack research tool 1.0 database and tissue bank coupled with new analyses and experiments. LPS+MI increased total CD8+ T-cells in the left ventricle versus the other groups (p<0.05 versus all). Memory CD8+ T-cells (CD44+CD27+) were 10-fold greater in LPS+MI compared to MI alone (p=0.02). Interleukin (IL)-4 stimulated splenic CD8+ T-cells away from an effector phenotype and towards a memory phenotype, inducing secretion of factors associated with the Wnt/β-catenin signaling that promoted monocyte migration and decreased viability. To dissect the effect of CD8+ T-cells post-MI, we administered a major histocompatibility complex-I blocking antibody starting 7 days before MI, which prevented effector CD8+ T-cell activation without affecting the memory response. The reduction in effector cells diminished infarct wall thinning but had no effect on macrophage numbers or MertK expression. LPS+MI+IgG attenuated macrophages within the infarct without effecting CD8+ T-cells suggesting these two processes were independent. Overall, our data indicate that effector and memory CD8+ T-cells at post-MI day 1 are amplified by chronic LPS to potentially promote infarct wall thinning.