Breast cancer detection with upstream data fusion, machine learning, and automated registration: initial results

Purpose

To develop an artificial intelligence algorithm for the detection of breast cancer by combining upstream data fusion (UDF), machine learning (ML), and automated registration, using digital breast tomosynthesis (DBT) and breast ultrasound (US).

Approach

Our retrospective study included examinations from 875 women obtained between April 2013 and January 2019. Included patients had a DBT mammogram, breast US, and biopsy proven breast lesion. Images were annotated by a breast imaging radiologist. An AI algorithm was developed based on ML for image candidate detections and UDF for fused detections. After exclusions, images from 150 patients were evaluated. Ninety-five cases were used for training and validation of ML. Fifty-five cases were included in the UDF test set. UDF performance was evaluated with a free-response receiver operating characteristic (FROC) curve.

Results

Forty percent of cases evaluated with UDF (22/55) yielded true ML detections in all three images (craniocaudal DBT, mediolateral oblique DBT, and US). Of these, 20/22 (90.9%) produced a UDF fused detection that contained and classified the lesion correctly. FROC analysis for these cases showed 90% sensitivity at 0.3 false positives per case. In contrast, ML yielded an average of 8.0 false alarms per case.

Conclusions

An AI algorithm combining UDF, ML, and automated registration was developed and applied to test cases, showing that UDF can yield fused detections and decrease false alarms when applied to breast cancer detection. Improvement of ML detection is needed to realize the full benefit of UDF.

Towards CNN-Based Registration of Craniocaudal and Mediolateral Oblique 2-D X-ray Mammographic Images

We investigate methodologies for the automated registration of pairs of 2-D X-ray mammographic images, taken from the two standard mammographic angles. We present two exploratory techniques, based on Convolutional Neural Networks, to examine their potential for co-registration of findings on the two standard mammographic views. To test algorithm performance, our analysis uses a synthetic, surrogate data set for performing controlled experiments, as well as real 2-D X-ray mammogram imagery. The preliminary results are promising, and provide insights into how the proposed techniques may support multi-view X-ray mammography image registration currently and as technology evolves in the future.

Synthesis of 4,5-Diazaspiro[2.3]hexanes and 1,2-Diazaspiro[3.3]heptanes as Hexahydropyridazine Analogues

4,5-Diazaspiro[2.3]hexanes are made by dihalocarbene addition across the exocyclic double bond of readily accessible 3-alkylidene-1,2-diazetidines. Using difluorocarbene, generated from TMSCF₃/NaI, these spirocycles were produced in yields up to 97% by stereospecific addition across the alkene. Lower yields (up to 64%) were observed using more reactive dichlorocarbene, due to competitive insertion of the carbene into the N-N bond. Larger 1,2-diazaspiro[3.3]heptanes are produced by [2 + 2] cycloaddition of 3-alkylidene-1,2-diazetidines with tetracyanoethylene (TCNE) in up to 99% yield.

Identifying Inhibitors of Inflammation: A Novel High-Throughput MALDI-TOF Screening Assay for Salt-Inducible Kinases (SIKs)

Matrix-assisted laser desorption/ionization time-of-flight (MALDI TOF) mass spectrometry has become a promising alternative for high-throughput drug discovery as new instruments offer high speed, flexibility and sensitivity, and the ability to measure physiological substrates label free. Here we developed and applied high-throughput MALDI TOF mass spectrometry to identify inhibitors of the salt-inducible kinase (SIK) family, which are interesting drug targets in the field of inflammatory disease as they control production of the anti-inflammatory cytokine interleukin-10 (IL-10) in macrophages. Using peptide substrates in in vitro kinase assays, we can show that hit identification of the MALDI TOF kinase assay correlates with indirect ADP-Hunter kinase assays. Moreover, we can show that both techniques generate comparable IC50 data for a number of hit compounds and known inhibitors of SIK kinases. We further take these inhibitors to a fluorescence-based cellular assay using the SIK activity-dependent translocation of CRTC3 into the nucleus, thereby providing a complete assay pipeline for the identification of SIK kinase inhibitors in vitro and in cells. Our data demonstrate that MALDI TOF mass spectrometry is fully applicable to high-throughput kinase screening, providing label-free data comparable to that of current high-throughput fluorescence assays.

Antibody and T-cell responses associated with experimental human malaria infection or vaccination show limited relationships

This study examined specific antibody and T-cell responses associated with experimental malaria infection or malaria vaccination, in malaria-naive human volunteers within phase I/IIa vaccine trials, with a view to investigating inter-relationships between these types of response. Malaria infection was via five bites of Plasmodium falciparum-infected mosquitoes, with individuals reaching patent infection by 11-12 days, having harboured four or five blood-stage cycles before drug clearance. Infection elicited a robust antibody response against merozoite surface protein-119 , correlating with parasite load. Classical class switching was seen from an early IgM to an IgG1-dominant response of increasing affinity. Malaria-specific T-cell responses were detected in the form of interferon-γ and interleukin-4 (IL-4) ELIspot, but their magnitude did not correlate with the magnitude of antibody or its avidity, or with parasite load. Different individuals who were immunized with a virosome vaccine comprising influenza antigens combined with P. falciparum antigens, demonstrated pre-existing interferon-γ, IL-2 and IL-5 ELIspot responses against the influenza antigens, and showed boosting of anti-influenza T-cell responses only for IL-5. The large IgG1-dominated anti-parasite responses showed limited correlation with T-cell responses for magnitude or avidity, both parameters being only negatively correlated for IL-5 secretion versus anti-apical membrane antigen-1 antibody titres. Overall, these findings suggest that cognate T-cell responses across a range of magnitudes contribute towards driving potentially effective antibody responses in infection-induced and vaccine-induced immunity against malaria, and their existence during immunization is beneficial, but magnitudes are mostly not inter-related.

The discovery of potent, orally bioavailable pyrimidine-5-carbonitrile-6-alkyl CXCR2 receptor antagonists

A hit-to-lead optimisation programme was carried out on the Novartis archive screening hit, pyrimidine 2-((2,6-dichlorobenzyl)thio)-5-isocyano-6-phenylpyrimidin-4-ol 4, resulting in the discovery of CXCR2 receptor antagonist 2-((2,3-difluorobenzyl)thio)-6-(2-(hydroxymethyl)cyclopropyl)-5-isocyanopyrimidin-4-ol 24. The SAR was investigated by systematic variation of the aromatic group at c-6, the linker between c-2 and the halogenated ring, and the c-5 nitrile moiety.

The discovery of potent, orally bioavailable pyrazolo and triazolopyrimidine CXCR2 receptor antagonists

A hit-to-lead optimisation programme was carried out on the Novartis archive screening hit, pyrazolopyrimidine 2-methyl-5-((phenylthio)methyl)pyrazolo[1,5-a]pyrimidin-7-ol 1, resulting in the discovery of CXCR2 receptor antagonist 2-benzyl-5-(((2,3-difluorophenyl)thio)methyl)-[1,2,4]triazolo[1,5-a]pyrimidin-7-ol 14. The SAR was investigated by systematic variation of the pendant thiol, alkyl and pyrimidinol groups. Replacement of the pyrazolopyrimidine core with a triazolo alternative led to a dual series of antagonists with favourable biological and pharmacokinetic properties.

ROS evaluation for a series of CNTs and their derivatives using an ESR method with DMPO

Carbon nanotubes (CNTs) are important materials in advanced industries. It is a concern that pulmonary exposure to CNTs may induce carcinogenic responses. It has been recently reported that CNTs scavenge ROS though non-carbon fibers generate ROS. A comprehensive evaluation of ROS scavenging using various kinds of CNTs has not been demonstrated well. The present work specifically investigates ROS scavenging capabilities with a series of CNTs and their derivatives that were physically treated, and with the number of commercially available CNTs. CNT concentrations were controlled at 0.2 through 0.6 wt%. The ROS scavenging rate was measured by ESR with DMPO. Interestingly, the ROS scavenging rate was not only influenced by physical treatments, but was also dependent on individual manufacturing methods. Ratio of CNTs to DMPO/ hydrogen peroxide is a key parameter to obtain appropriate ROS quenching results for comparison of CNTs. The present results suggest that dangling bonds are not a sole factor for scavenging, and electron transfer on the CNT surface is not clearly determined to be the sole mechanism to explain ROS scavenging.

Comparison of modeling methods to determine liver-to-blood inocula and parasite multiplication rates during controlled human malaria infection

Controlled human malaria infection is used to measure efficacy of candidate malaria vaccines before field studies are undertaken. Mathematical modeling using data from quantitative polymerase chain reaction (qPCR) parasitemia monitoring can discriminate between vaccine effects on the parasite's liver and blood stages. Uncertainty regarding the most appropriate modeling method hinders interpretation of such trials. We used qPCR data from 267 Plasmodium falciparum infections to compare linear, sine-wave, and normal-cumulative-density-function models. We find that the parameters estimated by these models are closely correlated, and their predictive accuracy for omitted data points was similar. We propose that future studies include the linear model.

Effect of vaccine dose on the safety and immunogenicity of a candidate TB vaccine, MVA85A, in BCG vaccinated UK adults

Purpose

A non-randomised, open-label, Phase I safety and immunogenicity dose-finding study to assess the safety and immunogenicity of the candidate TB vaccine Modified Vaccinia virus Ankara expressing Antigen 85A (MVA85A) from Mycobacterium tuberculosis (MTB) in healthy adult volunteers previously vaccinated with BCG.

Methods

Healthy BCG-vaccinated volunteers were vaccinated with either 1 × 10⁷ or 1 × 10⁸ PFU of MVA85A. All adverse events were documented and antigen specific T cell responses were measured using an ex vivo IFN-γ ELISPOT assay. Safety and immunogenicity were compared between the 2 dose groups and with a previous trial in which a dose of 5 × 10⁷ PFU MVA85A had been administered.

Results

There were no serious adverse events recorded following administration of either 1 × 10⁷ or 1 × 10⁸ PFU of MVA85A. Systemic adverse events were more frequently reported following administration of 1 × 10⁸ PFU of MVA85A when compared to either 5 × 10⁷ or 1 × 10⁷ PFU of MVA85A but were mild or moderate in severity and resolved completely within 7 days of immunisation. Antigen specific T cell responses as measured by the IFN-γ ELISPOT were significantly higher following immunisation in adults receiving 1 × 10⁸ PFU compared to the 5 × 10⁷ and 1 × 10⁷ doses. Additionally, a broader range of Ag85A epitopes are detected following 1 × 10⁸ PFU of MVA85A.

Conclusion

A higher dose of 1 × 10⁸ PFU of MVA85A is well-tolerated, increases the frequency of IFN-γ secreting T cells detected following immunisation and broadens the range of Ag85A epitopes detected.

Impact on malaria parasite multiplication rates in infected volunteers of the protein-in-adjuvant vaccine AMA1-C1/Alhydrogel+CPG 7909

Background

Inhibition of parasite growth is a major objective of blood-stage malaria vaccines. The in vitro assay of parasite growth inhibitory activity (GIA) is widely used as a surrogate marker for malaria vaccine efficacy in the down-selection of candidate blood-stage vaccines. Here we report the first study to examine the relationship between in vivo Plasmodium falciparum growth rates and in vitro GIA in humans experimentally infected with blood-stage malaria.

Methods

In this phase I/IIa open-label clinical trial five healthy malaria-naive volunteers were immunised with AMA1/C1-Alhydrogel+CPG 7909, and together with three unvaccinated controls were challenged by intravenous inoculation of P. falciparum infected erythrocytes.

Results

A significant correlation was observed between parasite multiplication rate in 48 hours (PMR) and both vaccine-induced growth-inhibitory activity (Pearson r = −0.93 [95% CI: −1.0, −0.27] P = 0.02) and AMA1 antibody titres in the vaccine group (Pearson r = −0.93 [95% CI: −0.99, −0.25] P = 0.02). However immunisation failed to reduce overall mean PMR in the vaccine group in comparison to the controls (vaccinee 16 fold [95% CI: 12, 22], control 17 fold [CI: 0, 65] P = 0.70). Therefore no impact on pre-patent period was observed (vaccine group median 8.5 days [range 7.5–9], control group median 9 days [range 7–9]).

Conclusions

Despite the first observation in human experimental malaria infection of a significant association between vaccine-induced in vitro growth inhibitory activity and in vivo parasite multiplication rate, this did not translate into any observable clinically relevant vaccine effect in this small group of volunteers.

Trial Registration

ClinicalTrials.gov [NCT00984763]

Multi-walled carbon nanotube-induced gene expression in the mouse lung: association with lung pathology

Due to the fibrous shape and durability of multi-walled carbon nanotubes (MWCNT), concerns regarding their potential for producing environmental and human health risks, including carcinogenesis, have been raised. This study sought to investigate how previously identified lung cancer prognostic biomarkers and the related cancer signaling pathways are affected in the mouse lung following pharyngeal aspiration of well-dispersed MWCNT. A total of 63 identified lung cancer prognostic biomarker genes and major signaling biomarker genes were analyzed in mouse lungs (n=80) exposed to 0, 10, 20, 40, or 80μg of MWCNT by pharyngeal aspiration at 7 and 56days post-exposure using quantitative PCR assays. At 7 and 56days post-exposure, a set of 7 genes and a set of 11 genes, respectively, showed differential expression in the lungs of mice exposed to MWCNT vs. the control group. Additionally, these significant genes could separate the control group from the treated group over the time series in a hierarchical gene clustering analysis. Furthermore, 4 genes from these two sets of significant genes, coiled-coil domain containing-99 (Ccdc99), muscle segment homeobox gene-2 (Msx2), nitric oxide synthase-2 (Nos2), and wingless-type inhibitory factor-1 (Wif1), showed significant mRNA expression perturbations at both time points. It was also found that the expression changes of these 4 overlapping genes at 7days post-exposure were attenuated at 56days post-exposure. Ingenuity Pathway Analysis (IPA) found that several carcinogenic-related signaling pathways and carcinogenesis itself were associated with both the 7 and 11 gene signatures. Taken together, this study identifies that MWCNT exposure affects a subset of lung cancer biomarkers in mouse lungs.

A human Phase I/IIa malaria challenge trial of a polyprotein malaria vaccine

We examined the safety, immunogenicity and efficacy of a prime-boost vaccination regime involving two poxvirus malaria subunit vaccines, FP9-PP and MVA-PP, expressing the same polyprotein consisting of six pre-erythrocytic antigens from Plasmodium falciparum.

Following safety assessment of single doses, 15 volunteers received a heterologous prime-boost vaccination regime and underwent malaria sporozoite challenge. The vaccines were safe but interferon-γ ELISPOT responses were low compared to other poxvirus vectors, despite targeting multiple antigens. There was no vaccine efficacy as measured by delay in time to parasitaemia. A number of possible explanations are discussed, including the very large insert size of the polyprotein transgene.

Comparison of sub-chronic metabolic effects of stable forms of naturally occurring GIP(1-30) and GIP(1-42) in high-fat fed mice

Glucose-dependent insulinotropic polypeptide (GIP) is a 42 amino acid hormone secreted from intestinal K-cells, which exhibits a number of actions including stimulation of insulin release. A truncated form, GIP(1-30), has recently been demonstrated in intestine and islet α-cells. To evaluate the potential physiological significance of this naturally occurring form of GIP, the present study has examined and compared the bioactivity of enzymatically stabilised forms, [D-Ala²]GIP(1-30) and [D-Ala²]GIP(1-42), in high-fat fed mice. Twice-daily injection of GIP peptides for 42 days had no significant effect on food intake or body weight. However, non-fasting glucose levels were significantly lowered, and insulin levels were elevated in both treatment groups compared to saline controls. The glycaemic response to i.p. glucose was correspondingly improved (P < 0.05) in [D-Ala²]GIP(1-30)- and [D-Ala²]GIP(1-42)-treated mice. Furthermore, glucose-stimulated plasma insulin levels were significantly elevated in both treatment groups compared to control mice. Insulin sensitivity was not significantly different between any of the groups. Similarly, plasma lipid profile, O₂ consumption, CO₂ production, respiratory exchange ratio, and energy expenditure were not altered by 42 days twice-daily treatment with [D-Ala²]GIP(1-30) or [D-Ala²]GIP(1-42). In contrast, ambulatory activity was significantly (P < 0.05) elevated during the light phase in both GIP treatment groups compared to saline controls. The results reveal that sustained GIP receptor activation exerts a spectrum of beneficial metabolic effects in high-fat fed mice. However, no differences were discernable between the biological actions of the enzyme-resistant analogues of the naturally occurring forms, GIP(1-30) and GIP(1-42).

Area Selective Atomic Layer Deposition by Soft Lithography

Area selective HfO2 thin film growth through atomic layer deposition (ALD) has been achieved on octadecyltrichlorosilane (ODTS) patterned Si substrates. Patterned hydrophobic self-assembled monolayers (SAMs) were first transferred to Si substrates by micro-contact printing. Using hafnium-tetrachloride or tetrakis(dimethylamido) hafnium(IV) and water as ALD precursors, amorphous HfO2 layers were then grown selectively on the SAM-free regions of the surface where native hydroxyl groups nucleate growth from the vapor phase. The HfO2 pattern was readily observed through scanning electron microscopy and scanning Auger imaging, demonstrating that soft lithography is a simple and promising method to achieve area selective ALD. To evaluate the selectivity, the resolution of the soft lithography based method was compared with that of area selective ALD of HfO2 by selective surface modification of patterned silicon oxide obtained using long-time SAM exposure. It was found that the selective surface modification showed much higher spatial resolution and selectivity, an observation consistent with previous studies indicating that highly ordered and densely packed ODTS films were important to achieve complete deactivation.

Four weeks administration of Liraglutide improves memory and learning as well as glycaemic control in mice with high fat dietary-induced obesity and insulin resistance

AIM

Liraglutide is a long-acting glucagon-like peptide-1 (GLP-1) mimetic which is a treatment option for type 2 diabetes. GLP-1 peptides, including Liraglutide, cross the blood-brain barrier and may additionally act to improve brain function. The present study tested the hypothesis that, in addition to its antihyperglycaemic actions, peripheral administration of Liraglutide exerts positive actions on cognitive function in mice with high fat dietary-induced obesity and insulin resistance.

METHODS

Young Swiss TO mice maintained on high fat diet for 20 weeks received twice-daily injections of Liraglutide (200 µg/kg bw; sc) or saline vehicle over 28 days. An additional group of mice on standard diet received twice-daily saline injections. Energy intake, bodyweight, non-fasting plasma glucose and insulin concentrations were monitored at regular intervals. Glucose tolerance, open field assessment, object recognition testing and electrophysiological long-term potentiation (LTP) were performed at termination of the study.

RESULTS

Liraglutide treatment resulted in significant time-dependent reduction in bodyweight and energy intake, whilst improving non-fasting glucose and normalizing glucose tolerance. Although Liraglutide did not alter general behaviour, treated mice exhibited marked increase in recognition index (RI) during object recognition testing, indicative of enhanced learning and memory ability. Furthermore, Liraglutide rescued the deleterious effects of high fat diet on hippocampal LTP of neurotransmission following both chronic and direct intracerebroventricular (icv) administration.

CONCLUSION

Liraglutide administered peripherally not only improves metabolic parameters but exerts additional beneficial effects on cognitive function and hippocampal synaptic plasticity. Whether therapy with GLP-1 mimetics has similar effects in humans with type 2 diabetes needs to be established.

Multiple functions of human T cells generated by experimental malaria challenge

Protective immunity generated following malaria infection may be comprised of Ab or T cells against malaria Ag of different stages; however, the short-lived immunity that is observed suggests deficiency in immune memory or regulatory activity. In this study, cellular immune responses were investigated in individuals receiving Plasmodium falciparum sporozoite challenge by the natural (mosquito bite) route as part of a malaria vaccine efficacy trial. Parasitemia, monitored by blood film microscopy and PCR, was subsequently cleared with drugs. All individuals demonstrated stable IFN-gamma, IL-2 and IL-4 ex vivo ELISPOT effector responses against P. falciparum-infected RBC (iRBC) Ag, 28 and 90 days after challenge. However, infected RBC-specific central memory responses, as measured by IFN-gamma cultured ELISPOT, were low and unstable over time, despite CD4(+) T cells being highly proliferative by CFSE dilution, and showed an inverse relationship to parasite density. In support of the observation of poor memory, co-culture experiments showed reduced responses to common recall Ag, indicating malaria-specific regulatory activity. This activity could not be accounted for by the expression of IL-10, TGF-beta, FOXP3 or CTLA-4, but proliferating T cells expressed high levels of CD95, indicating a pro-apoptotic phenotype. Lastly, there was an inverse relationship between FOXP3 expression, when measured 10 days after challenge, and ex vivo IFN-gamma measured more than 100 days later. This study shows that malaria infection elicits specific Th1 and Th2 effector cells, but concomitant weak central memory and regulatory activity, which may help to explain the short-lived immunity observed.

The relationship between human effector and memory T cells measured by ex vivo and cultured ELISPOT following recent and distal priming

Maintenance of T-cell responses is an essential feature in protection from many infectious diseases that must be harnessed in vaccination. The relationship between effector T-cell responses and more durable and highly proliferative T-cell memory, particularly in humans, is not well understood. In this study, effector T-cell responses were measured by overnight ex vivo interferon-gamma (IFN-gamma) enzyme-linked immunosorbent spot-forming cell assay (ELISPOT), whereas memory T cells were measured by 10-day culture followed by IFN-gamma ELISPOT (cultured ELISPOT). We observed a significant correlation between IFN-gamma responses to CD4-stimulatory, but not to CD8-stimulatory, recall antigens measured by these assays, suggesting a divergence in regulation. In vaccine trial participants who received a prime-boost vaccination regimen comprising malaria antigens delivered by poxviruses, there was a correlation between ex vivo and cultured responses on day 7, but not 3 months post-vaccination, with the ratio of cultured : ex vivo response increasing over time. To compare responses revealed by cultured ELISPOT in more detail, tetramers comprising viral recall antigens were used to ascribe effector-memory and central-memory T-cell phenotypes through CCR7 and CD62L costaining. For CD8(+) responses the effector phenotype decreased during the initial culture period and memory populations remained high within the resulting 20-fold to 50-fold increased IFN-gamma-secreting or tetramer(+) population. This was less marked for CD4(+) responses, which had higher starting memory phenotype. Depletion of these central-memory T-cell populations generally ablated responses in cultured ELISPOT and reduced ex vivo responses. This study highlights differences between CD4(+) and CD8(+) effector and memory T cells, and the more complex phenotype of CD4(+) T cells.