Improving Known-Unknown Cattle's Face Recognition for Smart Livestock Farm Management

Accurate identification of individual cattle is of paramount importance in precision livestock farming, enabling the monitoring of cattle behavior, disease prevention, and enhanced animal welfare. Unlike human faces, the faces of most Hanwoo cattle, a native breed of Korea, exhibit significant similarities and have the same body color, posing a substantial challenge in accurately distinguishing between individual cattle. In this study, we sought to extend the closed-set scope (only including identifying known individuals) to a more-adaptable open-set recognition scenario (identifying both known and unknown individuals) termed Cattle's Face Open-Set Recognition (CFOSR). By integrating open-set techniques to enhance the closed-set accuracy, the proposed method simultaneously addresses the open-set scenario. In CFOSR, the objective is to develop a trained model capable of accurately identifying known individuals, while effectively handling unknown or novel individuals, even in cases where the model has been trained solely on known individuals. To address this challenge, we propose a novel approach that integrates Adversarial Reciprocal Points Learning (ARPL), a state-of-the-art open-set recognition method, with the effectiveness of Additive Margin Softmax loss (AM-Softmax). ARPL was leveraged to mitigate the overlap between spaces of known and unknown or unregistered cattle. At the same time, AM-Softmax was chosen over the conventional Cross-Entropy loss (CE) to classify known individuals. The empirical results obtained from a real-world dataset demonstrated the effectiveness of the ARPL and AM-Softmax techniques in achieving both intra-class compactness and inter-class separability. Notably, the results of the open-set recognition and closed-set recognition validated the superior performance of our proposed method compared to existing algorithms. To be more precise, our method achieved an AUROC of 91.84 and an OSCR of 87.85 in the context of open-set recognition on a complex dataset. Simultaneously, it demonstrated an accuracy of 94.46 for closed-set recognition. We believe that our study provides a novel vision to improve the classification accuracy of the closed set. Simultaneously, it holds the potential to significantly contribute to herd monitoring and inventory management, especially in scenarios involving the presence of unknown or novel cattle.

An iterative noisy annotation correction model for robust plant disease detection

Previous work on plant disease detection demonstrated that object detectors generally suffer from degraded training data, and annotations with noise may cause the training task to fail. Well-annotated datasets are therefore crucial to build a robust detector. However, a good label set generally requires much expert knowledge and meticulous work, which is expensive and time-consuming. This paper aims to learn robust feature representations with inaccurate bounding boxes, thereby reducing the model requirements for annotation quality. Specifically, we analyze the distribution of noisy annotations in the real world. A teacher-student learning paradigm is proposed to correct inaccurate bounding boxes. The teacher model is used to rectify the degraded bounding boxes, and the student model extracts more robust feature representations from the corrected bounding boxes. Furthermore, the method can be easily generalized to semi-supervised learning paradigms and auto-labeling techniques. Experimental results show that applying our method to the Faster-RCNN detector achieves a 26% performance improvement on the noisy dataset. Besides, our method achieves approximately 75% of the performance of a fully supervised object detector when 1% of the labels are available. Overall, this work provides a robust solution to real-world location noise. It alleviates the challenges posed by noisy data to precision agriculture, optimizes data labeling technology, and encourages practitioners to further investigate plant disease detection and intelligent agriculture at a lower cost. The code will be released at https://github.com/JiuqingDong/TS_OAMIL-for-Plant-disease-detection.

A New Deep Learning-based Dynamic Paradigm Towards Open-World Plant Disease Detection

Plant disease detection has made significant strides thanks to the emergence of deep learning. However, existing methods have been limited to closed-set and static learning settings, where models are trained using a specific dataset. This confinement restricts the model's adaptability when encountering samples from unseen disease categories. Additionally, there is a challenge of knowledge degradation for these static learning settings, as the acquisition of new knowledge tends to overwrite the old when learning new categories. To overcome these limitations, this study introduces a novel paradigm for plant disease detection called open-world setting. Our approach can infer disease categories that have never been seen during the model training phase and gradually learn these unseen diseases through dynamic knowledge updates in the next training phase. Specifically, we utilize a well-trained unknown-aware region proposal network to generate pseudo-labels for unknown diseases during training and employ a class-agnostic classifier to enhance the recall rate for unknown diseases. Besides, we employ a sample replay strategy to maintain recognition ability for previously learned classes. Extensive experimental evaluation and ablation studies investigate the efficacy of our method in detecting old and unknown classes. Remarkably, our method demonstrates robust generalization ability even in cross-species disease detection experiments. Overall, this open-world and dynamically updated detection method shows promising potential to become the future paradigm for plant disease detection. We discuss open issues including classification and localization, and propose promising approaches to address them. We encourage further research in the community to tackle the crucial challenges in open-world plant disease detection. The code will be released at https://github.com/JiuqingDong/OWPDD.

Embracing limited and imperfect training datasets: opportunities and challenges in plant disease recognition using deep learning

Recent advancements in deep learning have brought significant improvements to plant disease recognition. However, achieving satisfactory performance often requires high-quality training datasets, which are challenging and expensive to collect. Consequently, the practical application of current deep learning-based methods in real-world scenarios is hindered by the scarcity of high-quality datasets. In this paper, we argue that embracing poor datasets is viable and aims to explicitly define the challenges associated with using these datasets. To delve into this topic, we analyze the characteristics of high-quality datasets, namely, large-scale images and desired annotation, and contrast them with the limited and imperfect nature of poor datasets. Challenges arise when the training datasets deviate from these characteristics. To provide a comprehensive understanding, we propose a novel and informative taxonomy that categorizes these challenges. Furthermore, we offer a brief overview of existing studies and approaches that address these challenges. We point out that our paper sheds light on the importance of embracing poor datasets, enhances the understanding of the associated challenges, and contributes to the ambitious objective of deploying deep learning in real-world applications. To facilitate the progress, we finally describe several outstanding questions and point out potential future directions. Although our primary focus is on plant disease recognition, we emphasize that the principles of embracing and analyzing poor datasets are applicable to a wider range of domains, including agriculture. Our project is public available at https://github.com/xml94/EmbracingLimitedImperfectTrainingDatasets.

Local refinement mechanism for improved plant leaf segmentation in cluttered backgrounds

Plant phenotyping is a critical field in agriculture, aiming to understand crop growth under specific conditions. Recent research uses images to describe plant characteristics by detecting visual information within organs such as leaves, flowers, stems, and fruits. However, processing data in real field conditions, with challenges such as image blurring and occlusion, requires improvement. This paper proposes a deep learning-based approach for leaf instance segmentation with a local refinement mechanism to enhance performance in cluttered backgrounds. The refinement mechanism employs Gaussian low-pass and High-boost filters to enhance target instances and can be applied to the training or testing dataset. An instance segmentation architecture generates segmented masks and detected areas, facilitating the derivation of phenotypic information, such as leaf count and size. Experimental results on a tomato leaf dataset demonstrate the system's accuracy in segmenting target leaves despite complex backgrounds. The investigation of the refinement mechanism with different kernel sizes reveals that larger kernel sizes benefit the system's ability to generate more leaf instances when using a High-boost filter, while prediction performance decays with larger Gaussian low-pass filter kernel sizes. This research addresses challenges in real greenhouse scenarios and enables automatic recognition of phenotypic data for smart agriculture. The proposed approach has the potential to enhance agricultural practices, ultimately leading to improved crop yields and productivity.

Super-resolution and apodization with discrete adaptive optics

High-resolution imaging is of great importance in various fields. The use of pupil phase-only filters (PPF) exceeds the diffraction limit of the imaging system in a simple way. When dealing with distorted wavefronts, however, PPF require that aberrations be compensated for. In this paper, we introduce a novel technique consisting of the use of discrete adaptive optics with PPFs so that the compensating device implements the PPF at the same time. Analysis of the theory for point spread function reshaping using PPFs has enabled us to develop a new approach to characterizing apodizing filters. A validation experiment has been carried out, the first of its kind to our knowledge, in which a number of PPFs were combined with two levels of compensation. Our experimental results are discussed.

Multiview Monitoring of Individual Cattle Behavior Based on Action Recognition in Closed Barns Using Deep Learning

Cattle behavior recognition is essential for monitoring their health and welfare. Existing techniques for behavior recognition in closed barns typically rely on direct observation to detect changes using wearable devices or surveillance cameras. While promising progress has been made in this field, monitoring individual cattle, especially those with similar visual characteristics, remains challenging due to numerous factors such as occlusion, scale variations, and pose changes. Accurate and consistent individual identification over time is therefore essential to overcome these challenges. To address this issue, this paper introduces an approach for multiview monitoring of individual cattle behavior based on action recognition using video data. The proposed system takes an image sequence as input and utilizes a detector to identify hierarchical actions categorized as part and individual actions. These regions of interest are then inputted into a tracking and identification mechanism, enabling the system to continuously track each individual in the scene and assign them a unique identification number. By implementing this approach, cattle behavior is continuously monitored, and statistical analysis is conducted to assess changes in behavior in the time domain. The effectiveness of the proposed framework is demonstrated through quantitative and qualitative experimental results obtained from our Hanwoo cattle video database. Overall, this study tackles the challenges encountered in real farm indoor scenarios, capturing spatiotemporal information and enabling automatic recognition of cattle behavior for precision livestock farming.

Causal Inference with Multilevel Data: A Comparison of Different Propensity Score Weighting Approaches

Propensity score methods are a widely recommended approach to adjust for confounding and to recover treatment effects with non-experimental, single-level data. This article reviews propensity score weighting estimators for multilevel data in which individuals (level 1) are nested in clusters (level 2) and nonrandomly assigned to either a treatment or control condition at level 1. We address the choice of a weighting strategy (inverse probability weights, trimming, overlap weights, calibration weights) and discuss key issues related to the specification of the propensity score model (fixed-effects model, multilevel random-effects model) in the context of multilevel data. In three simulation studies, we show that estimates based on calibration weights, which prioritize balancing the sample distribution of level-1 and (unmeasured) level-2 covariates, should be preferred under many scenarios (i.e., treatment effect heterogeneity, presence of strong level-2 confounding) and can accommodate covariate-by-cluster interactions. However, when level-1 covariate effects vary strongly across clusters (i.e., under random slopes), and this variation is present in both the treatment and outcome data-generating mechanisms, large cluster sizes are needed to obtain accurate estimates of the treatment effect. We also discuss the implementation of survey weights and present a real-data example that illustrates the different methods.

Identification of new variants and candidate genes in women with familial premature ovarian insufficiency using whole-exome sequencing

PURPOSE

To identify candidate variants in genes possibly associated with premature ovarian insufficiency (POI).

METHODS

Fourteen women, from 7 families, affected by idiopathic POI were included. Additionally, 98 oocyte donors of the same ethnicity were enrolled as a control group. Whole-exome sequencing (WES) was performed in 14 women with POI to identify possibly pathogenic variants in genes potentially associated with the ovarian function. The candidate genes selected in POI patients were analysed within the exome results of oocyte donors.

RESULTS

After the variant filtering in the WES analysis of 7 POI families, 23 possibly damaging genetic variants were identified in 22 genes related to POI or linked to ovarian physiology. All variants were heterozygous and five of the seven families carried two or more variants in different genes. We have described genes that have never been associated to POI pathology; however, they are involved in important biological processes for ovarian function. In the 98 oocyte donors of the control group, we found no potentially pathogenic variants among the 22 candidate genes.

CONCLUSION

WES has previously shown as an efficient tool to identify causative genes for ovarian failure. Although some studies have focused on it, and many genes are identified, this study proposes new candidate genes and variants, having potentially moderate/strong functional effects, associated with POI, and argues for a polygenic etiology of POI in some cases.

P-340 The endometrial switch following progesterone exposure correlates with uterine peristalsis

Study question

What is the effect of the changes of the endometrium following progesterone exposure on uterine peristalsis?

Summary answer

Uteri with a greater increase endometrial thickness after progesterone exposure have significantly greater uterine peristalsis (UP). Therefore, endometrial compaction is associated with quieter uteri.

What is known already

Endometrial compaction is considered a sign of adequate response to progesterone. However, it is not always possible to visualize it. Little is known about endometrial switch following progesterone and its consequences after frozen embryo transfer, although it is hypothesized that increased endometrial growth after the window of implantation may be related to an increase in endogenous or exogenous oestrogen causing impairment of progesterone function. On the other hand, the role of progesterone in inhibiting endometrial contractile function is well known. Consequently, both the change in endometrial thickness and uterine contractility are indicators of progesterone function.

Study design, size, duration

This retrospective observational was carried out in Instituto Bernabeu of Alicante. The study included 215 patients with at least three previous implantation failures after egg donation treatment, which underwent uterine peristalsis assessment the day of embryo transfer from June 2017 to December 2021.

Participants/materials, setting, methods

UP assessment was performed using 4D ultrasound, recording a video for 6 minutes. UP variable was split in quartiles, and then the last quartile (UP ≥ 1,5 contractions per minute) was considered the hypercontractility group. All patients had performed an ultrasound to assess endometrial thickness between 7 to 10 days before embryo transfer. Endometrial compaction has been considered when it has decreased. Endometrial switch (ES) is considered the percentage between both endometrial measures.

Main results and the role of chance

The mean age of patients was 40,11 years who underwent an average of 3,75 embryo transfers. The average UP was 1,09 contractions per minute. The average ES was 9% of increased endometrium. Only 83 (38,6%) patients had endometrial compaction. To assess the association between UP and endometrial switch after progesterone exposure a univariate assessment was performed using Pearson’s correlation resulting in a negative correlation (r=-0,16; p = 0,019). It was used the hypercontractility group for performing bivariate logistic regression was performed to examine the effect of independent variables (previous miscarriages, previous pregnancies, C-section, endometriosis, adenomyosis, myomatosis and endometrial preparation) on ES. Greater ES is statistically associated to greater UP with OR 1,013 (95%IC: 1,002 to 1,025; p = 0,041).

Limitations, reasons for caution

It is a retrospective study based in patients with multiple implantation failure. It is possible that our conclusions couldn't be the same in patients with good prognosis.

Wider implications of the findings

Assessment of endometrial changes and uterine peristalsis provide information about the response of the uterus to progesterone exposure. However, more studies assessing it prospectively both would be interesting to define which population has high risk to develop situations with inadequate progesterone response.

Trial registration number

Not applicable

P-370 The vaginal microbiome in the first trimester of pregnancy is different in spontaneous versus IVF gestation

Study question

Are there differences in the vaginal microbiome of pregnant women who had a spontaneous pregnancy compared to those who required IVF?

Summary answer

The composition of the vaginal microbiome at 12 week's gestation is different in women who achieve the pregnancy spontaneously or by IVF.

What is known already

The vaginal microbiome plays an important role in women's reproductive health, finding associations between different microbiome patterns and the presence of infertility and embryo implantation failure in IVF. Additionally, recent studies show a correlation between obstetrics and perinatal outcomes and the composition of vaginal microbiota in pregnant women, as well as an increased risk of obstetrics complications in pregnant women after IVF.

Study design, size, duration

Observational, prospective and multicentre study. A total of 64 women were enrolled between January 2020 and June 2021. Spontaneous pregnancies n = 30; and IVF pregnancies n = 34.

Participants/materials, setting, methods

Vaginal swabs were obtained by speculum exam at 12 weeks of gestation in two public hospitals and a fertility private clinic in Spain, to evaluate the differences in vaginal microbiome between both cohorts. The microbiome composition was analyzed by sequencing the V3-V4 region of the 16S rRNA on the Illumina MiSeq platform.

Main results and the role of chance

There were no significant differences in socio-demographic characteristics between groups, except for an expected higher maternal age in the IVF cohort.

Lactobacillus was the most prevalent genus in both groups. When we compared the beta diversity of vaginal microbial by cohort a significant difference was obtained (p = 0.001).

Gardenella, Neisseria, Prevotella and Staphyloccocus were significantly enriched in the IVF group (p = 0.01).

A further evaluation of the four most abundant Lactobacillus species showed that Lactobacillus iners was dominant in IVF pregnancies (15.2%) compared to spontaneous (9.8%) (p = 0.002). On the other hand, Lactobacillus  gasseri showed a lower abundance in vaginal microbiome from women belonged to IVF (9.2%) vs spontaneous pregnant group (13.8%) (p = 0.005).

These findings allowed us to create a model to identify a microbial signature. This model is able to discriminate between IVF and spontaneous pregnancies.

Limitations, reasons for caution

The main limitation of our study is the small sample size. Larger studies are needed to corroborate our findings and their relationship with important aspects such as obstetric and perinatal complications.

Wider implications of the findings

The microbiome composition is different between both cohorts. The microbiome found in our IVF cohort has been also associated with obstetric complications as preterm delivery in previous studies. This suggest that the microbiome composition could be a plausible etiology for a higher risk of adverse pregnancy outcomes in IVF patients.

Trial registration number

Not applicable

P-024 Identification of spermatogenic infertility phenotypes using next generation sequencing

Study question

Can next generation sequencing (NGS) contribute to diagnoses male idiopathic infertility?

Summary answer

A male factor gene panel identifies pathogenic variants associated to spermatogenic failure in oligozoospermia and cryptozoospermia patients.

What is known already

In 50% of cases, infertility is due to a male factor problem. Although the causes of male infertility are heterogeneous, genetic causes account for approximately 30% of cases. Some phenotypes have been associated with specific genetic disorders such as chromosomal abnormalities and chromosome Y microdeletions. However, current genetic studies explain only 4% of cases, whilst most cases of male factor infertility remain without a clear diagnosis. Therefore, new techniques that explain the cause of male infertility are needed. Advances in NGS allowed us to study a large number of genes involved in spermatogenesis process in patients with idiopathic infertility.

Study design, size, duration

A retrospective study was performed from July 2020 until May 2021. A total of 30 patients with abnormal seminal count parameters (oligozoospermic and cryptozoospermic) were included in the male factor gene panel study. Patients carrying Y-chromosome microdeletions or abnormal karyotype were excluded. The control group included 20 normozoospermic healthy donors selected on the basis of normal semen parameters according to the WHO criteria (2010).

Participants/materials, setting, methods

Genomic DNA extraction from blood-EDTA of the patients was performed using the commercial MagMax DNA MultiSample Ultra kit and the King-Fisher automated extractor (ThermoFisher®). Next Generation Sequencing (NGS) was done using a panel with 426 genes involved in the spermatogenesis process. Panel sequencing for identification of genetic variants was performed using Nextera Enrichment technology (Illumina®). FASTAQ data were processed using BWA and GATK algorithms. VCF files were analyzed using Variant Interpreter software.

Main results and the role of chance

After data analysis, we observed that eight of the thirty patients studied were carriers of mutations in least one of the genes included in the panel (8/30, 26.7%). We identified the following pathogenic variants: a missense mutation (Phe1052Val) and a deletion (Phe508del) of CFTR gene (2/30, 6.6%), two frameshifts (Asp128GlufsTer34 and Lys1299Ter) of CEP290 (2/30, 6.6%), a missense mutation (Tyr284Cys) of GNRHR gene (1/30, 3.3%), a missense mutation (Tyr416Cys) of SCN5A gene (1/30, 3.3%), a deletion (Ser83del) of NANOS1 gene (1/30, 3.3%), a stop gained in splice region Arg341Ter of TEX14 gene (1/30, 3.3%), a splicing donor c.362 + 2T>C of ESR2 gene (1/30, 3.3%) and a missense mutation (Ser321Leu) of DNAH5 gene (1/30, 3.3%), which are related to spermatogenesis failure. Additionally, some variants classified as benign have been identified, which are not associated with pathogenicity. All the variants identified are related with male infertility, affecting spermatogenesis process, such as congenital bilateral absence of the vas deferens (CFTR), reproductive system syndrome (CEP190), endocrine disorder (GNRHR, hypogonadotropic hypogonadism), testis expressed (SCN5A), spermatogenic failure (NANOS1, TEX14 and ESR2) and syndromic infertility (DNAH5). Nevertheless, no pathogenic mutations associated to spermatogenic failure were observed in the control group.

Limitations, reasons for caution

The main limitation of this study is the small number of patients included. Further studies including a higher number of males with idiopathic infertility are warranted to confidently link the genetic variants included in our gene panel to spermatogenic failure.

Wider implications of the findings

The gene list included in our panel represents a step-forward in the diagnosis screening of males with altered sperm parameters. Our results may add in the knowledge of male factor infertility in order to provide etiologic factors towards a personalized treatment and adequate genetic counselling.

Trial registration number

Not applicable

Identification of vaginal microbiome associated with IVF pregnancy

The factors that cause a preterm birth (PTB) are not completely understood up to date. Moreover, PTB is more common in pregnancies achieved by in-vitro fertilization (IVF) than in spontaneous pregnancies. Our aim was to compare the composition of vaginal microbiome at 12 weeks of gestation between women who conceived naturally or through IVF in order to study whether IVF PTB-risk could be related to vaginal microbiome composition. We performed an observational, prospective and multicentre study among two public hospitals and a fertility private clinic in Spain. Vaginal swabs from 64 pregnant women at 12 weeks of gestation were collected to analyse the microbiome composition by sequencing the V3-V4 region of the 16S rRNA. Our results showed that the vaginal microbiome signature at 12 weeks of pregnancy was different from women who conceived naturally or through IVF. The beta diversity and the genus composition were different between both cohorts. Gardnerella, Neisseria, Prevotella, and Staphylococcus genus were enriched genus in the vaginal microbiome from the IVF group, allowing us to create a balance model to predict both cohorts. Moreover, at species level the L. iners abundance was higher and L. gasseri was lower in the IVF group. As a conclusion, our findings were consistent with a proposed framework in which IVF pregnancy are related to risk for preterm birth (PTB) suggesting vaginal microbiome could be the reason to the relation between IVF pregnancy and risk for PTB.

Data-centric annotation analysis for plant disease detection: Strategy, consistency, and performance

Object detection models have become the current tool of choice for plant disease detection in precision agriculture. Most existing research improved the performance by ameliorating networks and optimizing the loss function. However, because of the vast influence of data annotation quality and the cost of annotation, the data-centric part of a project also needs more investigation. We should further consider the relationship between data annotation strategies, annotation quality, and the model's performance. In this paper, a systematic strategy with four annotation strategies for plant disease detection is proposed: local, semi-global, global, and symptom-adaptive annotation. Labels with different annotation strategies will result in distinct models' performance, and their contrasts are remarkable. An interpretability study of the annotation strategy is conducted by using class activation maps. In addition, we define five types of inconsistencies in the annotation process and investigate the severity of the impact of inconsistent labels on model's performance. Finally, we discuss the problem of label inconsistency during data augmentation. Overall, this data-centric quantitative analysis helps us to understand the significance of annotation strategies, which provides practitioners a way to obtain higher performance and reduce annotation costs on plant disease detection. Our work encourages researchers to pay more attention to annotation consistency and the essential issues of annotation strategy. The code will be released at: https://github.com/JiuqingDong/PlantDiseaseDetection_Yolov5 .

Improving Accuracy of Tomato Plant Disease Diagnosis Based on Deep Learning With Explicit Control of Hidden Classes

Recognizing plant diseases is a major challenge in agriculture, and recent works based on deep learning have shown high efficiency in addressing problems directly related to this area. Nonetheless, weak performance has been observed when a model trained on a particular dataset is evaluated in new greenhouse environments. Therefore, in this work, we take a step towards these issues and present a strategy to improve model accuracy by applying techniques that can help refine the model's generalization capability to deal with complex changes in new greenhouse environments. We propose a paradigm called "control to target classes." The core of our approach is to train and validate a deep learning-based detector using target and control classes on images collected in various greenhouses. Then, we apply the generated features for testing the inference of the system on data from new greenhouse conditions where the goal is to detect target classes exclusively. Therefore, by having explicit control over inter- and intra-class variations, our model can distinguish data variations that make the system more robust when applied to new scenarios. Experiments demonstrate the effectiveness and efficiency of the proposed approach on our extended tomato plant diseases dataset with 14 classes, from which 5 are target classes and the rest are control classes. Our detector achieves a recognition rate of target classes of 93.37% mean average precision on the inference dataset. Finally, we believe that our study offers valuable guidelines for researchers working in plant disease recognition with complex input data.

P–546 Exome sequencing and preimplantation genetic testing for unexplained recurrent fetal malformations

Study question

Could patient suffering unexplained recurrent fetal malformations be benefit of PGT-M by exome sequencing mutations identification?

Summary answer

Patients suffering unexplained recurrent fetal malformations could be benefit of the use of exome sequencing in combination to PGT-M to have a healthy live birth.

What is known already

Fetal malformations account for approximately 3% of live births and causes include: chromosomal abnormalities, exposure to toxic substances or teratogens and infections. Recently, studies have shown that several monogenic diseases are linked to fetal abnormalities. However, because of the large number of potential genes, genetic testing is challenging. Exome sequencing is widely used to detect genetic mutations and has emerged as a useful tool for finding the genetic cause of fetal abnormalities. The aim of this study was to show how exome sequencing in patients suffering unexplained recurrent fetal malformations in combination to PGT-M could lead to successful healthy newborn.

Study design, size, duration

Case report of a non-consanguineous couple with unexplained, recurrent fetal malformations. Couple were recruited during clinical consultation for unexplained recurrent fetal malformations at a private reproductive medicine clinic. The couple had two malformed fetus with the same congenital abnormalities: hydrocephalus, cerebellar vermis agenesis, cerebellar hypoplasia and enlarged cisterna magna. Patients signed written informed consent regarding to exome testing. For fetal sample, informed consent was obtained from parents.

Participants/materials, setting, methods

Sample of the affected fetus were provided. Parental genomic DNA was extracted from peripheral blood. Exome sequencing was performed using TrusightOne (Illumina®). FASTAQ data were processed through BWA and GATK algorithm. VCF files were analysed using Variant Interpreter software. After genetic counselling, PGT-M was performed using linkage polymorphic markers analysis and mutation sequencing. Embryo biopsy was carried at blastocyst stage. Embryos were vitrified and one healthy embryo was thaw and transfer in a subsequent cycle.

Main results and the role of chance

An homozygous novel pathogenic mutation c.641 C>T (p.Ala214Val) in FVLCR2 gene was found. The parents were heterozygous carriers revealing that the detected variant follow an autosomal recessive pattern. The FLVCR2 (14q24.3) gene encodes a transmembrane protein that belongs to the major facilitator superfamily of secondary carriers that transport small solutes in response to chemiosmosis ion gradients, such as calcium. Mutations in this gene are related to fetal central nervous system defects. This disorder is diagnosed prenatally and is lethal. PGT-M was recommended during genetic counselling. After control ovarian stimulation 14 oocytes were retrieved and finally 4 embryos were suitable for embryo biopsy at blastocyst stage. One embryo was diagnosed as healthy, two affected and one heterozygous carrier. The healthy embryo was thaw and transferred and a healthy male baby was born.

Limitations, reasons for caution

Exome sequencing has technical limitations: only covers mutations in coding regions and does not cover noncoding regions of the genome. It also cannot reliably detect copy-number variants at single gene level.

Wider implications of the findings: This study offers strong evidence of exome-sequencing as a new diagnostic strategy and powerful tool discovering the underlying etiology of recurrent fetal malformations and identifying new genes important for human development. Using this strategy in combination with PGT-M, clinicians can help couples with recurrent fetal malformations to have healthy newborns.

Trial registration number

Not applicable

P–541 Identification of novel variants and candidate genes in women with familial idiopathic premature ovarian failure using whole-exome sequencing

Study question

Is it possible to identify a genetic cause of familial premature ovarian failure (POF) with whole-exome sequencing (WES)?

Summary answer

Whole-exome sequencing is the most efficient strategy to identify probably pathogenic mutations in different genes in pathologies of polygenic etiology such as premature ovarian failure.

What is known already

Premature ovarian failure is the loss of ovarian function before the age of 40, and it is a common cause of infertility in women. This pathology has a heterogeneous etiology. Some chromosomal and genetic alterations have been described, and could explain approximately 20% of cases. However, in most patients the origin remains unknown. Recent studies with next-generation sequencing (NGS) have identified new variants in candidate genes related with premature ovarian insufficiency (POI) or premature ovarian failure (POF). These genes are not only involved in processes such as folliculogenesis, but also with DNA damage repair, homologous recombination, and meiosis.

Study design, size, duration

Fourteen women, from 7 families, affected by idiopathic POF were included in the study from October 2019 to September 2020. Seven POF patients were recruited when they came to our clinic to undergo assisted reproductive treatment. In the anamnesis, it was found that they had relatives with a diagnosis of POF, who were also recruited for the study. The inclusion criteria were amenorrhea before 38 years old and analytical and ultrasound signs of ovarian failure.

Participants/materials, setting, methods

WES was performed using TrusightOne (Illumina®). Sequenced data were aligned through BWA tool and GATK algorithm was used for SNVs/InDel identification. VCF files were annotated using Variant Interpreter software. Only the variants shared by each family were extracted for analysis and these criteria were followed: (1) Exonic/splicing variants in genes related with POF or involved in biological ovarian functions (2) Variants with minor allele frequency (MAF) ≤0.05 and (3) having potentially moderate/strong functional effects.

Main results and the role of chance

Seventy-nine variants possibly related with the POF phenotype were identified in the seven families. All these variants had a minor allele frequency (MAF) ≤0.05 in the gnomAD database and 1000 genomes project. Among these candidate variants, two were nonsense, six splice region, one frameshift, two inframe deletion and 68 missense. Thirty-two of the missense variants were predicted to have deleterious effects by minimum two of the four in silico algorithms used (SIFT, PolyPhen–2, MutationTaster and PROVEAN). All variants were heterozygous, and all the families carried three or more candidate variants. Altogether, 43 probably damaging genetic variants were identified in 39 genes expressed in the ovary and related with POF/POI or linked to ovarian physiology. We have described genes that have never been associated to POF pathology, however they may be involved in key biological processes for ovarian function. Moreover, some of these genes were found in two families, for example DDX11, VWF, PIWIL3 and HSD3B1. DDX11 may function at the interface of replication-coupled DNA repair and sister chromatid cohesion. VWF gene is suggested to be associated with follicular atresia in previous studies. PIWIL3 functions in development and maintenance of germline stem cells, and HSD3B1 is implicated in ovarian steroidogenesis.

Limitations, reasons for caution

Whole-exome sequencing has some limitations: does not cover noncoding regions of the genome, it also cannot detect large rearrangements, copy-number variants (large deletions/duplications), mosaic mutations, mutations in repetitive or high GC rich regions and mutations in genes with corresponding pseudogenes or other highly homologous sequences.

Wider implications of the findings: WES has previously shown to be an efficient tool to identify genes as cause of POF, and has demonstrated the polygenic etiology. Although some studies have focused on it, and many genes are identified, this study proposes new candidate genes and variants, having potentially moderate/strong functional effects, associated with POF.

Trial registration number

Not applicable

P–286 Uterine vascularity in women with previous caesarean section and its potential role in implantation failure: a retrospective cohort study

Study question

Does a previous Caesarean section affect uterine vascularisation the day of embryo transfer?

Summary answer

3D vascularisation parameters show less uterine irrigation in patients with previous Caesarean section

What is known already

A recent retrospective cohort study demonstrates that previous Caesarean section impairs live birth rates after assisted reproductive treatment (ART) compared to a previous vaginal delivery. Furthermore, it has been hypothesized about the mechanisms by which post-cesarean section niche may diminish clinical pregnancy rates. One of the hypothetical process mentioned has been a distorted contractility of the uterus caused by fibrosis, which can influence in the vascularisation of the endometrium.

Study design, size, duration

We retrospectively studied the uterine contractility and 3D vascularisation parameters in women who had an embryo transfer at the Instituto Bernabeu of Alicante, between 2018 and 2020 with one recurrent implantation failure (at least two good quality blastocysts transferred from egg donation treatment).

Participants/materials, setting, methods

Patients with large myomas (more than 4 cm), adenomyosis or polyp were excluded. In total, 196 patients were assessed on the day of embryo transfer which 12 patients had a previous caesarean section. Uterine contractility was analyzed using 4D ultrasound after 6 minutes of video recording. Vascularisation index and vascularisation flow index were assessed after the endometrial volume definition.

Main results and the role of chance

Baseline characteristics of both groups were comparable. 3D vascularization parameters were significantly lower in women with a previous caesarean section. Vascularization Index (VI) reached 0,8% in caesarean section group (CS group) versus 2,3% (p = 0,038) and vascularization flow index (VFI) was 0,2 in CS group versus 0,8 (p = 0,038) Despite uterine peristalsis showed less contractility in those patients with previous caesarean section (0,8 contractions per minute versus 1,1 contractions per minute), non-statistical differences were demonstrated (p = 0,154)

Limitations, reasons for caution

This study is limited by its retrospective design and the low number of cases.

Wider implications of the findings: The lower 3D vascularisation indexes support a post-Caesarean section vascular-related impaired perfusion as a hypothetical mechanism. Its correlation with a possible impairment in the embryo implantation after fertility treatments warrants further studies.

Trial registration number

Not applicable

P–588 Follicle-stimulating hormone receptor genotype and its influence on the results of double ovarian stimulation in IVF cycles

Study question

Does the follicle-stimulating hormone receptor (FSHR) genotype influence the results of the ovarian stimulation treatment in the luteal phase?

Summary answer

All patients undergoing in-vitro fertilization benefit from luteal phase ovarian stimulation, regardless of their follicle-stimulating hormone receptor genotype.

What is known already

Previous studies suggest that FSH receptor polymorphism in position 680 influences the response to ovarian stimulation in the luteal phase. It was observed that patients with SS genotype seems to require a higher dose to obtain an optimal ovarian response. Later, it was reported that, in patients with SS genotype, a better performance seems to be obtained by administering highly purified urinary FSH while, in SN patients, better results were obtained with recombinant FSH. In patients with NN genotype, no differences were found. Our aim was to test whether this concept is applicable to ovarian stimulation in the luteal phase.

Study design, size, duration

One hundred and thirty-four patients were included in a retrospective study between July 2017 and September 2020. In these patients, a double stimulation protocol was carried out and the FSH receptor was genotyped either as part of the pre-treatment fertility tests or for the current study. Patients with a double stimulation treatment who could not be genotyped were excluded from the analysis.

Participants/materials, setting, methods

To genotype the 680 position of the FSH receptor, a real-time PCR for allelic discrimination was carried out using StepOnePlus™ Real-Time PCR System (Applied Biosystems™. Ref: 4376600). Non-parametic tests were used to study the differences between the groups. They were performed with the software R Statistical Software, version 4.0.3.

Main results and the role of chance

The results of ovarian stimulation in the luteal phase were better compared to the conventional follicular phase. Statistically significant differences (p < 0.001) were found in the number of retrieved oocytes (5.06 versus 3.51), retrieved MII (4.13 versus 2.91), fertilized oocytes (3.22 versus 1.81) and blastocysts formed (1.79 versus 0.62). Furthermore, these differences remained regardless of the genotype for the 680 position of the FSH receptor in all groups (p < 0.05).

In addition, better results were obtained in the luteal phase in patients who have been stimulated with the type of gonadotropin that already had better performance in the follicular phase for its genotype, that is, highly purified urinary FSH in SS patients and recombinant FSH in SN patients, compared to other types of gonadotropin (p < 0.05).

We also observed that stimulation in the luteal phase lasts longer and consume more gonadotropins than in the follicular phase. This is especially notable in the case of patients with SS genotype, who required slightly higher consumption of gonadotropins compared to the other genotypes in the luteal phase, as had previously been observed in the follicular phase for this genotype.

Limitations, reasons for caution

The retrospective study design and the sample size could be a limitation. Furthermore, we cannot determine whether the improvement in luteal phase performance is related to differences in the physiological environment between phases of the cycle or is caused by a possible activation of the ovary from the previous stimulation.

Wider implications of the findings: All patients undergoing in-vitro fertilization seems to benefit from luteal phase ovarian stimulation, regardless of their genotype for FSHR. In addition, the pharmacogenetic recommendation when choosing the type of FSH for ovarian stimulation should be the same both in the follicular phase and in the luteal phase.

Trial registration number

Not applicable

P–540 A feasible diagnostic approach for the cryptic subtelomeric traslocations in early recurrent miscarriage patients by preimplantation genetic testing (PGT)

Study question

Could cryptic subtelomeric traslocations in early recurrent miscarriage patients be diagnosed by preimplantation genetic testing?

Summary answer

PGT is a powerful tool to detect subtelomeric cryptic traslocations identifying the cause of early recurrent miscarriage and allowing subsequent genetic counselling. What is known already: Chromosome translocations are frequently associated with birth defects, spontaneous early pregnancy losses and infertility. However, submicroscopic traslocations (so-called cryptic traslocations) are too small to be detected by conventional karyotyping.. Due to balanced status, high resolution molecular techniques as arrayCGH are not able to detect it. Thus, cryptic traslocations detection is challenging. PGT is able to detect CNVs at higher resolution than routine karyotyping. Therefore, the recurrent diagnosis of CNV at embryo level could suggest a subchromosomal parental traslocation. The aim of this study is to investigate the feasibility of using PGT as an indicator of parental balanced cryptic traslocations.

Study design, size, duration

We included three couples who underwent PGT for unexplained repeated pregnancy loss (RPL) in our clinic from February 2020 to November 2020. Common established causes of RPL (uterine anomalies, antiphospholipid syndrome, immunological, hormonal and metabolic disorders) were previously rouled-out. Even couple karyotypes were normal. Twenty-three embryos from those couples were biopsied at blastocyst and analysed for CNVs detection using low coverage whole genome NGS.

Participants/materials, setting, methods

PGT by NGS was performed by Veriseq-NGS (Illumina), with previous whole genome amplification. Fluorescence in situ hybridization (FISH) using parental blood samples were performed to validate the origin of subchromosomal number variation. Commercially available subtelomeric specific probes were selected according to the CNV identified and the procedures were performed according to the manufacturer’s protocols.

Main results and the role of chance

Overall, CNVs of terminal duplication and deletion that imply unbalanced traslocation derivatives were detected in the 43.5% of biopsied embryos. For couple 1, 4 out of 5 embryos (80%) carried deletion of telomeric region on chromosomes 5 and 21. Three out of 6 biopsed embyos (50%) were diagnosed with subchromosomal copy variants at telomeric region on chromosomes 6 and 16 for couple 2. In the case of couple 3, three out of 12 embryos (25%) were carriers of CNV at subtelomeric region on chromosomes 2 and 6. The size of CNVs detected ranges from 8Mb to 20Mb. Accurate diagnosis with the parental study was made by FISH. The combination of probes to detect the structural chromosome alteration were: Tel5qter-LSI21q, Tel6pter-CEP16 and Tel6pter-CEP6 for each couple respectively. The FISH studies reveal that CNVs were inherited from one parent carrying the balanced cryptic traslocation. Ultimately, the abnormal karyotype from the carrier parent were 46,XY,t(5;21)(q33.2;q21.2) for couple 1, 46,XY,t(6;16)(p22.3;q22.1) for couple 2 and 46,XY,t(2;6)(p25.1;p24.2) for couple 3. Finally, each couple performed a cryotransfer of a single normal balanced embryo. Two pregnancies are ongoing.

Limitations, reasons for caution

The main limitation of this approach is the NGS- PGT resolution. CNVs smaller than 5Mb could not be detected.

Wider implications of the findings: This study shows the value of PGT for unexplained RPL, followed by parental FISH to better characterize CNVs and identify couples in whom one partner carries a cryptic translocation. Accurate diagnosis of parental chromosome translocation can achieve with FISH only, but FISH would not be performed unless PGT showed CNVs.

Trial registration number

Not applicable

Open Set Self and Across Domain Adaptation for Tomato Disease Recognition With Deep Learning Techniques

Recent advances in automatic recognition systems based on deep learning technology have shown the potential to provide environmental-friendly plant disease monitoring. These systems are able to reliably distinguish plant anomalies under varying environmental conditions as the basis for plant intervention using methods such as classification or detection. However, they often show a performance decay when applied under new field conditions and unseen data. Therefore, in this article, we propose an approach based on the concept of open-set domain adaptation to the task of plant disease recognition to allow existing systems to operate in new environments with unseen conditions and farms. Our system specifically copes diagnosis as an open set learning problem, and mainly operates in the target domain by exploiting a precise estimation of unknown data while maintaining the performance of the known classes. The main framework consists of two modules based on deep learning that perform bounding box detection and open set self and across domain adaptation. The detector is built based on our previous filter bank architecture for plant diseases recognition and enforces domain adaptation from the source to the target domain, by constraining data to be classified as one of the target classes or labeled as unknown otherwise. We perform an extensive evaluation on our tomato plant diseases dataset with three different domain farms, which indicates that our approach can efficiently cope with changes of new field environments during field-testing and observe consistent gains from explicit modeling of unseen data.

Style-Consistent Image Translation: A Novel Data Augmentation Paradigm to Improve Plant Disease Recognition

Deep learning shows its advantages and potentials in plant disease recognition and has witnessed a profound development in recent years. To obtain a competing performance with a deep learning algorithm, enough amount of annotated data is requested but in the natural world, scarce or imbalanced data are common, and annotated data is expensive or hard to collect. Data augmentation, aiming to create variations for training data, has shown its power for this issue. But there are still two challenges: creating more desirable variations for scarce and imbalanced data, and designing a data augmentation to ease object detection and instance segmentation. First, current algorithms made variations only inside one specific class, but more desirable variations can further promote performance. To address this issue, we propose a novel data augmentation paradigm that can adapt variations from one class to another. In the novel paradigm, an image in the source domain is translated into the target domain, while the variations unrelated to the domain are maintained. For example, an image with a healthy tomato leaf is translated into a powdery mildew image but the variations of the healthy leaf are maintained and transferred into the powdery mildew class, such as types of tomato leaf, sizes, and viewpoints. Second, current data augmentation is suitable to promote the image classification model but may not be appropriate to alleviate object detection and instance segmentation model, mainly because the necessary annotations can not be obtained. In this study, we leverage a prior mask as input to tell the area we are interested in and reuse the original annotations. In this way, our proposed algorithm can be utilized to do the three tasks simultaneously. Further, We collect 1,258 images of tomato leaves with 1,429 instance segmentation annotations as there is more than one instance in one single image, including five diseases and healthy leaves. Extensive experimental results on the collected images validate that our new data augmentation algorithm makes useful variations and contributes to improving performance for diverse deep learning-based methods.

Dry eye is matched by increased intrasubject variability in tear osmolarity as confirmed by machine learning approach

OBJECTIVE

Because of high variability, tear film osmolarity measures have been questioned in dry eye assessment. Understanding the origin of such variability would aid data interpretation. This study aims to evaluate osmolarity variability in a clinical setting.

MATERIAL AND METHODS

Twenty dry eyes and 20 control patients were evaluated. Three consecutive osmolarity measurements per eye at 5min intervals were obtained. Variability was represented by the difference between both extreme readings per eye. Machine learning techniques were used to quantify discrimination capacity of tear osmolarity for dry eye.

RESULTS

Mean osmolarities in the control and dry eye groups were 295.1±7.3mOsm/L and 300.6±11.2mOsm/L, respectively (P=.004). Osmolarity variabilities were 7.5±3.6mOsm/L and 16.7±11.9mOsm/L, for the control and dry eye groups, respectively (P<.001). Based on osmolarity, a logistic classifier showed an 85% classification accuracy.

CONCLUSIONS

In the clinical setting, both mean osmolarity and osmolarity variability in the dry eye group were significantly higher than in the control group. Machine learning techniques showed good classification accuracy. It is concluded that higher variability of tear osmolarity is a dry eye feature.

Deep Learning-Based Phenotyping System With Glocal Description of Plant Anomalies and Symptoms

Recent advances in Deep Neural Networks have allowed the development of efficient and automated diagnosis systems for plant anomalies recognition. Although existing methods have shown promising results, they present several limitations to provide an appropriate characterization of the problem, especially in real-field scenarios. To address this limitation, we propose an approach that besides being able to efficiently detect and localize plant anomalies, allows to generate more detailed information about their symptoms and interactions with the scene, by combining visual object recognition and language generation. It uses an image as input and generates a diagnosis result that shows the location of anomalies and sentences describing the symptoms as output. Our framework is divided into two main parts: First, a detector obtains a set of region features that contain the anomalies using a Region-based Deep Neural Network. Second, a language generator takes the features of the detector as input and generates descriptive sentences with details of the symptoms using Long-Short Term Memory (LSTM). Our loss metric allows the system to be trained end-to-end from the object detector to the language generator. Finally, the system outputs a set of bounding boxes along with the sentences that describe their symptoms using glocal criteria into two different ways: a set of specific descriptions of the anomalies detected in the plant and an abstract description that provides general information about the scene. We demonstrate the efficiency of our approach in the challenging tomato diseases and pests recognition task. We further show that our approach achieves a mean Average Precision (mAP) of 92.5% in our newly created Tomato Plant Anomalies Description Dataset. Our objective evaluation allows users to understand the relationships between pathologies and their evolution throughout their stage of infection, location in the plant, symptoms, etc. Our work introduces a cost-efficient tool that provides farmers with a technology that facilitates proper handling of crops.

Genetic variation and population structure of Diaphorina citri using cytochrome oxidase I sequencing

Citrus greening disease, or huanglongbing (HLB), is currently one of the most devastating diseases of citrus. The bacteria thought to be responsible for the disease, Candidatus Liberibacter asiaticus impact the majority of commercial citrus species worldwide. These bacteria are transmitted by the Asian citrus psyllid (ACP), Diaphorina citri Kuwayama, which is now found in most citrus growing regions. With no known cure, ACP-vectored HLB is responsible for significant economic losses to the global citrus industry. A better understanding of the global genetic diversity of D. citri would improve current and future pest management and mitigation programs. To assess the genetic diversity of D. citri in worldwide collections, a total of 1,108 sequences belonging to ACP gathered from 27 countries in the Americas, the Caribbean, Southeast and Southwest Asia were examined for the study. 883 D. citri came from 98 locations in 18 different countries, and were sequenced using a 678bp fragment of the mitochondrial cytochrome oxidase I (COI) gene. Additionally, 225 previously-reported D. citri COI sequences, were also included in our analysis. Analyses revealed 28 haplotypes and a low genetic diversity. This is in accordance with previous reports on the little diversity of D. citri in worldwide populations. Our analyses reveal population structure with 21 haplotypes showing geographic association, increasing the resolution for the source estimation of ACP. This study reveals the distribution of haplotypes observed in different geographic regions and likely geographic sources for D. citri introductions.

Effect of thyme and oregano essential oils on the shelf life of salmon and seaweed burgers

The effect of different essential oils on the quality and shelf life of fish and seaweed burgers during storage was evaluated. For this purpose, thyme and oregano essential oils were selected at a concentration of 0.05% (v/w). Three types of salmon and seaweed burgers were prepared: without essential oil, burgers with red thyme essential oil (0.05% (v/w)) and burgers with oregano essential oil (0.05% (v/w)), which were vacuum packaged and stored at 4 ℃ for 17 days. Physicochemical and microbiological analyses were carried out periodically throughout storage. The addition of both essential oils did not have any effect on the evolution of the pH, the moisture content or texture parameters. Only the thyme essential oil managed to slightly slow down the increase of total volatile basic nitrogen and trimethylamine nitrogen. The samples with oregano essential oil and especially those with thyme essential oil showed minor oxidation. The salmon and seaweed burgers without essential oils and those which contained oregano essential oil showed a faster increase of mesophilic counts than those which had thyme essential oil, but no noticeable improvement was observed in the shelf life of the burgers with thyme essential oil. To improve the shelf life of the fish and seaweed burgers, it would be necessary to increase the concentration of both essential oils.

Blazar spectral variability as explained by a twisted inhomogeneous jet

Blazars are active galactic nuclei, which are powerful sources of radiation whose central engine is located in the core of the host galaxy. Blazar emission is dominated by non-thermal radiation from a jet that moves relativistically towards us, and therefore undergoes Doppler beaming. This beaming causes flux enhancement and contraction of the variability timescales, so that most blazars appear as luminous sources characterized by noticeable and fast changes in brightness at all frequencies. The mechanism that produces this unpredictable variability is under debate, but proposed mechanisms include injection, acceleration and cooling of particles, with possible intervention of shock waves or turbulence. Changes in the viewing angle of the observed emitting knots or jet regions have also been suggested as an explanation of flaring events and can also explain specific properties of blazar emission, such as intra-day variability, quasi-periodicity and the delay of radio flux variations relative to optical changes. Such a geometric interpretation, however, is not universally accepted because alternative explanations based on changes in physical conditions-such as the size and speed of the emitting zone, the magnetic field, the number of emitting particles and their energy distribution-can explain snapshots of the spectral behaviour of blazars in many cases. Here we report the results of optical-to-radio-wavelength monitoring of the blazar CTA 102 and show that the observed long-term trends of the flux and spectral variability are best explained by an inhomogeneous, curved jet that undergoes changes in orientation over time. We propose that magnetohydrodynamic instabilities or rotation of the twisted jet cause different jet regions to change their orientation and hence their relative Doppler factors. In particular, the extreme optical outburst of 2016-2017 (brightness increase of six magnitudes) occurred when the corresponding emitting region had a small viewing angle. The agreement between observations and theoretical predictions can be seen as further validation of the relativistic beaming theory.

A Robust Deep-Learning-Based Detector for Real-Time Tomato Plant Diseases and Pests Recognition

Plant Diseases and Pests are a major challenge in the agriculture sector. An accurate and a faster detection of diseases and pests in plants could help to develop an early treatment technique while substantially reducing economic losses. Recent developments in Deep Neural Networks have allowed researchers to drastically improve the accuracy of object detection and recognition systems. In this paper, we present a deep-learning-based approach to detect diseases and pests in tomato plants using images captured in-place by camera devices with various resolutions. Our goal is to find the more suitable deep-learning architecture for our task. Therefore, we consider three main families of detectors: Faster Region-based Convolutional Neural Network (Faster R-CNN), Region-based Fully Convolutional Network (R-FCN), and Single Shot Multibox Detector (SSD), which for the purpose of this work are called "deep learning meta-architectures". We combine each of these meta-architectures with "deep feature extractors" such as VGG net and Residual Network (ResNet). We demonstrate the performance of deep meta-architectures and feature extractors, and additionally propose a method for local and global class annotation and data augmentation to increase the accuracy and reduce the number of false positives during training. We train and test our systems end-to-end on our large Tomato Diseases and Pests Dataset, which contains challenging images with diseases and pests, including several inter- and extra-class variations, such as infection status and location in the plant. Experimental results show that our proposed system can effectively recognize nine different types of diseases and pests, with the ability to deal with complex scenarios from a plant's surrounding area.

Effect of partial sodium replacement on physicochemical parameters of smoked sea bass during storage

The objective of this work was to study the effect of partial sodium replacement by potassium and packaging conditions on the physicochemical properties of smoked sea bass during cold storage. Sea bass fillets were salted with 100% NaCl (Na samples) or with 50% NaCl-50% KCl (Na:K samples), smoked, packaged under three different conditions (air, vacuum and modified atmosphere) and stored at 4 °C for 42 days. Physicochemical parameters, color and texture were periodically determined in the raw material and in smoked samples during cold storage. The smoking process led to a reduction in moisture, pH and a(w) values, and an increase in water holding capacity, ash and mineral contents. Smoked fish exhibited significant differences in color and texture as compared to fresh fish. The type of packaging had an effect on the pH, water holding capacity and texture. Samples in air exhibited the highest pH values and water holding capacity in these samples gradually decreased during storage. Textural parameters decreased during storage in samples packaged in vacuum and modified atmosphere. The pH of Na samples was initially higher than in Na:K samples, and this difference remained over the rest of the study. The type of salt did not affect the texture or other physicochemical parameters.