Explainable image classification with evidence counterfactual

Pattern Analysis and Applications - The complexity of state-of-the-art modeling techniques for image classification impedes the ability to explain model predictions in an interpretable way. A...     

Biomarkers for intake of dairy fat and dairy products

The aim of this study was to correlate the fatty acids characteristic of dairy fat (14:0, 14:1, 15:0, 17:0 and 17:1) in adipose tissue, serum lipid fractions and the estimated intake of dairy fat, and to investigate whether they can be used as biomarkers for dairy fat/product intake. The highest correlations were observed between 14:0 in adipose tissue and the estimated intakes of 14:0 (r = 0.60) and dairy fat (r = 0.50), and between 15:0 in adipose tissue and the estimated intakes of 15:0 (r = 0.55) and dairy fat (r = 0.55). Among the fatty acids in serum lipid fractions, 15:0 in cholesteryl esters (CE) and triacylglycerols showed the highest correlation to both 15:0 in adipose tissue, intakes of 15:0, dairy fat and dairy products. The results from our study suggest that the contents of fatty acids characteristic of dairy fat (14:0, 14:1, 15:0, 17:0, 17:1) in adipose tissue might be used as markers not only for the intake of dairy fat, but also of dairy products (times/day). When adipose tissue specimens are not available, 15:0 in serum CE appears to be the best marker for dairy fat intake. It was also observed that fatty acids from dairy fat are incorporated differently into serum lipid fractions and adipose tissue.     

A note on a motion control problem for a placement machine

Assembling printed circuit boards efficiently using automated placement machines is a challenging task. Here, we focus on a motion control problem for a specific type of placement machines. More specifically, the problem is to establish movement patterns for the robot arm, the feeder rack, and—when appropriate—the worktable, of a sequential pick-and-place machine. In this note we show that a (popular) greedy strategy may not always yield an optimum solution. However, under the relevant Tchebychev metric, we can model the problem as a linear program, thereby establishing the existence of a polynomial time algorithm for this motion control problem. Finally, we give experimental evidence that computing optimal solutions to this motion control problem can yield significantly better solutions than those found by a greedy method.     

Surface scanning analysis of planar arrays of analytes with desorption electrospray ionization-mass spectrometry

The analysis of analytes deposited on, separated on, or otherwise distributed about a planar surface using desorption electrospray ionization mass spectrometry in a surface scanning sampling mode was investigated. The physical regions of the surface-impacting solvent/gas jet desorption/ionization plume were described. Under the conditions typical for desorption electrospray ionization used here, the impact plume formed an elliptical desorption/ionization region on the surface. Most effective desorption/ionization was obtained from a smaller elliptical area within the larger impact region that was centered on a point on-axis from the sprayer tip to the surface. Maximum signal from a given amount of material on a surface was observed with proper plume and sample alignment when the diameter of the sample spot was less than the diameter of the central high-efficiency desorption/ionization region of the impact plume. Solvent and gas flow out of this high-efficiency desorption/ionization region was found to limit analyte accessibility to this region via a "washing effect" when analytes were on smooth surfaces or on surfaces for which the analyte had little affinity. As such, the direction of surface scanning and scan speed during an analysis was found to be important for maximizing signal levels and signal reproducibility on particular surface types. Overall, the results presented illustrate means to improve analysis of sample spots on various types of surfaces using desorption electrospray ionization mass spectrometry in a surface scanning mode.     

Comparison of interleukin-1 beta expression by in situ hybridization in monoclonal gammopathy of undetermined significance and multiple myeloma

We investigated whether interleukin-1beta (IL-1beta) is differentially expressed in plasma cells from monoclonal gammopathy of undetermined significance (MGUS) and multiple myeloma (MM) patients because IL-1beta appears to play a major role in the development of lytic bone lesions, the major clinical feature distinguishing MGUS from myeloma. In situ hybridization (ISH) for IL-1beta was performed using bone marrow aspirates from 51 MM, 7 smoldering MM, 21 MGUS, and 5 normal control samples. Using the ISH technique IL-1beta mRNA was detectable in the plasma cells from 49 of 51 patients with active myeloma and 7 of 7 patients with smoldering myeloma. In contrast, 5 of 21 patients with MGUS and 0 of 5 normal controls had detectable IL-1beta message. Bone lesions were present in 40 of the 51 MM patients analyzed, and all 40 patients had IL-1beta mRNA by ISH. These results show that greater than 95% of MM patients but less than 25% of MGUS patients are positive for IL-1beta production. In the future, continued follow-up of IL-1beta positive and negative MGUS patients should determine whether aberrant expression of plasma cell IL-1beta is predictive of those MGUS patients that will eventually progress to active myeloma.     

Long-term survival (10 years or more) in 30 patients with primary amyloidosis

The median survival in primary systemic (AL) amyloidosis is less than 18 months. No published series of patients with AL amyloidosis have reported survival of more than 10 years. The records of all Mayo Clinic patients with a diagnosis of AL amyloidosis between January 1, 1966 and March 1, 1987 were reviewed. Patients with secondary amyloidosis, familial amyloidosis, senile systemic amyloidosis, and localized amyloidosis were excluded. During the 21 years of the study, 841 patients with AL amyloidosis were seen. Of these, 29 were excluded because the diagnosis was made at autopsy, and 2 others were excluded because no follow-up data were available. Actuarial survival for the 810 patients was 51% at 1 year, 16% at 5 years, and 4.7% at 10 years. Thirty patients survived for 10 years or more after the histologic diagnosis of AL amyloidosis; all received alkylating-agent therapy. In 14 patients, the monoclonal protein disappeared from the serum or urine. Of 10 patients with nephrotic syndrome, 4 had an objective response. Congestive heart failure, older age, creatinine value of 2 mg/dL or more, bone marrow plasma cell value of 20% or more, platelet count of 500 x 10(9)/L or less, and the presence of peripheral neuropathy were underrepresented in the 10-year survivors and are unfavorable prognostic features. Five percent of patients with AL amyloidosis survived for 10 years or more.     

Method for short-term prediction of milk yield at the quarter level to improve udder health monitoring

Udder health problems are often associated with milk losses. These losses are different between quarters, as infected quarters are affected both by systemic and pathogen-specific local effects, whereas noninfected quarters are only subject to systemic effects. To gain insight in these losses and the milk yield dynamics during disease, it is essential to have a reliable reference for quarter-level milk yield in an unperturbed state, mimicking its potential yield. We developed a novel methodology to predict this quarter milk yield per milking session, using an historical data set of 504 lactations collected on a test farm by an automated milking system from DeLaval (Tumba, Sweden). Using a linear mixed model framework in which covariates associated with the linearized Wood model and the milking interval are included, we were able to describe quarter-level yield per milking session with a proportional error below 10%. Applying this model enables us to predict the milk yield of individual quarters 1 to 50 d ahead with a mean prediction error ranging between 8 and 20%, depending on the amount of historical data available to estimate the random effect covariates for the predicted lactation. The developed methodology was illustrated using 2 examples for which quarter-level milk losses are calculated during clinical mastitis. These showed that the quarter-level mixed model allows us to gain insight in quarter lactation dynamics and enables to calculate milk losses in different situations.     

Genetic diversity and mycotoxin production of Fusarium lactis species complex isolates from sweet pepper

An internal fruit rot disease of sweet peppers was first detected in Belgium in 2003. Research conducted mostly in Canada indicates that this disease is primarily caused by Fusarium lactis Pirotta. Ninety-eight Fusarium isolates obtained from diseased sweet peppers from Belgium, as well as from other countries (Canada, the Netherlands and the United Kingdom) were identified by sequencing the translation elongation factor 1α (EF). Of these 98 isolates, 13 were identified as F. oxysporum Schltdl., nine as F. proliferatum (Matsush.) Nirenberg and two belonged to clade 3 of the F. solani species complex. Of the 74 remaining isolates, the EF sequence showed 97% to 98% similarity to F. lactis. Of these isolates, the β-tubulin (TUB), calmodulin (CAM) and the second largest subunit of RNA polymerase II (RPB2) genes were also sequenced. Analysis of the combined sequences revealed that the 74 isolates share nine combined sequences that correspond to nine multilocus sequence types (STs), while the F. lactis neotype strain and one other strain, both isolated from figs, form a separate ST. Together, these 10 STs represent a monophyletic F. lactis species complex (FLASC). An unusually high level of genetic diversity was observed between (groups of) these STs. Two of them (ST5 and ST6) fulfilled the criteria for species recognition based on genealogical exclusivity and together represent a new monophyletic species lineage (FLASC-1). The seven other STs, together with the F. lactis neotype ST, form a paraphyletic species lineage in the African clade of the Gibberella fujikuroi species complex (GFSC). From each of the 10 STs, the mycotoxin production was assessed using a multi-mycotoxin liquid chromatography mass spectrometry method. Out of the 27 analyzed mycotoxins, beauvericin and fumonisins were detected in sweet pepper tissue and in maize kernels. The 10 STs clearly differed in the amount of mycotoxin produced, but there was only limited congruence between the production profile and the phylogenetic analysis. Furthermore, the morphological characterization (based on mycelial growth rate and the length of macroconidia) showed distinct differences between the 10 STs, but again there was limited congruence with the phylogenetic results. In conclusion, the data presented in this study demonstrate that 75% of the isolates obtained from sweet pepper with internal fruit rot belong to a F. lactis species complex (FLASC), including a new FLASC-1 monophyletic species, and that the members of this complex display great genetic and phenotypic diversity.     

Pathogen-group specific association between CXCR1 polymorphisms and subclinical mastitis in dairy heifers

The chemokine (C-X-C motif) receptor 1 (CXCR1) gene encodes the homonymous receptor for interleukin 8 (IL8) on polymorphonuclear neutrophilic leucocytes (PMNL). Binding causes migration from blood to milk, activation and prolonged survival of PMNL, a crucial process in the innate immune defence of the bovine mammary gland against invading mastitis-causing pathogens. The main objective of this study was to screen the entire coding region of the CXCR1 gene for polymorphisms and to analyse their association with udder health of dairy heifers. One-hundred-and-forty Belgian Holstein heifers originating from 20 commercial dairy farms were genotyped by DNA sequencing. Detailed phenotypic data on udder health was available including quarter bacteriological culture results and somatic cell count (SCC) in early lactation and composite milk SCC during first lactation. In total, 16 polymorphisms (including 8 missense mutations) were detected. Polymorphism c.980A>G was associated with pathogen-group specific IMI: heifers with genotype AG were less likely to have an IMI due to major mastitis pathogens compared with heifers with genotype GG but did not have less IMI by coagulase-negative staphylococci, so-called minor pathogens. CXCR1 genotype was neither associated with quarter SCC in early lactation nor with composite SCC during lactation. Although mastitis susceptibility is influenced by many factors, some genetic polymorphisms potentially have major effects on udder health of heifers, as was shown here. These results trigger us to further study the relationship between CXCR1 polymorphisms and mastitis susceptibility in both observational and experimental trials.