Immune-related lncRNA classification of head and neck squamous cell carcinoma

Background Long noncoding RNAs (lncRNAs) play a critical role in innate and adaptive immune responses. Thus, we aimed to identify ideal subtypes for head and neck squamous cell carcinoma (HNSCC) based on immune-related lncRNAs. Methods TCGA HNSCC cohort was divided into two datasets (training and validation dataset), and 960 previously characterized immune-related lncRNAs were extracted for non-negative matrix factorization analysis. We characterized our HNSCC subtypes based on biological behaviors, immune landscape and response to immunotherapy in both training and validation cohort. A lncRNA-signature was generated to predict our HNSCC subtypes, and essential lncRNAs involved in tumor microenvironment (TME) were identified. Results We developed and validated two HNSCC subtypes (C1 and C2) based on the 70 lncRNAs in the training and validation cohort. C2 subtype displayed good prognosis, high immune cell infiltration, immune-related genes expression and sensitivity to PD-1 blockade. C1 subtype was associated with high activity of mTORC1 signaling and glycolysis as well as high fraction of inactive immune cells. Finally, we generated a 31-lncRNA signature that could predict our above subtypes with high accurate. Additionally, TRG-AS1 was identified as the essential lncRNA involving TME formation. Knockdown of TRG-AS1 inhibited the expression of HLA-A, HLA-B, HLA-C, CXCL9, CXCL10 and CXCL11. High expression of TRG-AS1 indicated a favorable prognosis in HNSCC and anti-PD-L1 cohort (IMvigor210). Conclusions Our study establishes a novel HNSCC classification on the basis of 31-lncRNA, helping to identify beneficiaries for anti-PD-1 treatment. In addition, a critical lncRNA TRG-AS1 is identified as a new potential prognosis biomarker as well as therapeutic target.

Generalization enhancement of artificial neural network for turbulence closure by feature selection

Feature selection targets for selecting relevant and useful features, and is a vital challenge in turbulence modeling by machine learning methods. In this paper, a new posterior feature selection method based on validation dataset is proposed, which is an efficient and universal method for complex systems including turbulence. Different from the priori feature importance ranking of the filter method and the exhaustive search for feature subset of the wrapper method, the proposed method ranks the features according to the model performance on the validation dataset, and generates the feature subsets in the order of feature importance. Using the features from the proposed method, a black-box model is built by artificial neural network (ANN) to reproduce the behavior of Spalart-Allmaras (S-A) turbulence model for high Reynolds number (Re) airfoil flows in aeronautical engineering. The results show that compared with the model without feature selection, the generalization ability of the model after feature selection is significantly improved. To some extent, it is also demonstrated that although the feature importance can be reflected by the model parameters during the training process, artificial feature selection is still very necessary.

AFD-Net: Apple Foliar Disease Multi Classification using Deep Learning on Plant Pathology Dataset

Plant diseases significantly affect the crop, so their identification is very important. Correct identification of these diseases is crucial for establishing a good disease control strategy to avoid time and financial losses. In general, machines can greatly reduce the possibility of human error. In particular, computer vision techniques developed through deep learning have paved a way to detect and diagnose these plant diseases on the leaf. In this work, the model AFD-Net was developed to detect and identify various leaf diseases in apple trees. The dataset is from Kaggle 2020 and 2021 and was financially supported by the Cornell Initiative for Digital Agriculture. A AFD-Net was proposed for leaf disease classification in apple trees and the results of the efficiency of the model are compared with other state-of-the-art deep learning approaches. The results of the experiments in the validation dataset show that the proposed AFD-Net model achieves the highest values compared to other deep learning models in the original and extended datasets with 98.7% accuracy for Plant Pathology 2020 and 92.6% for Plant Pathology 2021.

CLASSIFICATION OF TRANSLATIONAL LANDSLIDE ACTIVITY USING VEGETATION ANOMALIES INDICATOR (VAI) IN KUNDASANG, SABAH

This paper introduced a novel method of landslide activity mapping using vegetation anomalies indicators (VAIs) obtained from high resolution remotely sensed data. The study area was located in a tectonically active area of Kundasang, Sabah, Malaysia. High resolution remotely sensed data were used to assist manual landslide inventory process and production on VAIs. The inventory process identified 33, 139, and 31 of active, dormant, and relict landslides, respectively. Landslide inventory map were randomly divided into two groups for training (70%) and validation (30%) datasets. Overall, 7 group of VAIs were derived including (i) tree height irregularities; (ii) tree canopy gap; (iii) density of different layer of vegetation; (iv) vegetation type distribution; (v) vegetation indices (VIs); (vi) root strength index (RSI); and (vii) distribution of water-loving trees. The VAIs were used as the feature layer input of the classification process with landslide activity as the target results. The landslide activity of the study area was classified using support vector machine (SVM) approach. SVM parameter optimization was applied by using Grid Search (GS) and Genetic Algorithm (GA) techniques. The results showed that the overall accuracy of the validation dataset is between 61.4–86%, and kappa is between 0.335–0.769 for deep-seated translational landslide. SVM RBF-GS with 0.5m spatial resolution produced highest overall accuracy and kappa values. Also, the overall accuracy of the validation dataset for shallow translational is between 49.8–71.3%, and kappa is between 0.243–0.563 where SVM RBF-GS with 0.5m resolution recorded the best result. In conclusion, this study provides a novel framework in utilizing high resolution remote sensing to support labour intensive process of landslide inventory. The nature-based vegetation anomalies indicators have been proved to be reliable for landslide activity identification in Malaysia.

Denoising and Dehazing an Image in a Cascaded Pattern for Continuous Casting

Automatic vision systems have been widely used in the continuous casting of the steel industry, which improve efficiency and reduce labor. At present, high temperatures with evaporating fog cause images to be noisy and hazy, impeding the usage of advanced machine learning algorithms in this task. Instead of considering denoising and dehazing separately like previous papers, we established that by taking advantage of deep learning in a modeling complex formulation, our proposed algorithm, called Cascaded Denoising and Dehazing Net (CDDNet) reduces noise and hazy in a cascading pattern. Experimental results on both synthesized images and a pragmatic video from a continuous casting factory demonstrate our method’s superior performance in various metrics. Compared with existing methods, CDDNet achieved a 50% improvement in terms of peak signal-to-noise ratio on the validation dataset, and a nearly 5% improvement on a dataset that has never seen before. Besides, our model generalizes so well that processing a video from an operating continuous casting factory with CDDNet resulted in high visual quality.

Increases in Picocyanobacteria Abundance in Agriculturally Eutrophic Pampean Lakes Inferred from Historical Records of Secchi Depth and Chlorophyll-a

Phytoplankton size structure has profound consequences on food-web organization and energy transfer. Presently, picocyanobacteria (size < 2 µm) represent a major fraction of the autotrophic plankton of Pampean lakes. Glyphosate is known to stimulate the development of picocyanobacteria capable of degrading the herbicide. Due to the worldwide adoption of glyphosate-resistant crops, herbicide usage has increased sharply since the mid-1990s. Unfortunately, there are very few studies (none for the Pampa region) reporting picocyanobacteria abundance before 2000. The proliferation of µm sized particles should decrease Secchi disc depth (ZSD). Therefore ZSD, conditional to chlorophyll-a, may serve as an indicator of picocyanobacteria abundance. We use generalized additive models (GAMs) to analyze a “validation” dataset consisting of 82 records of ZSD, chlorophyll-a, and picocyanobacteria abundance from two Pampean lakes surveys (2009 and 2015). In support of the hypothesis, ZSD was negatively related to picocyanobacteria after accounting for the effect of chlorophyll-a. We then fitted a “historical” dataset using hierarchical GAMs to compare ZSD conditional to chlorophyll-a, before and after 2000. We estimated that ZSD levels during 2000–2021 were, on average, only about half as deep as those during 1980–1999. We conclude that the adoption of glyphosate-resistant crops has stimulated outbreaks of picocyanobacteria populations, resulting in lower water transparency.

Assimilation of sea ice thickness derived from CryoSat-2 along-track freeboard measurements into the Met Office's Forecast Ocean Assimilation Model (FOAM)

The feasibility of assimilating sea ice thickness (SIT) observations derived from CryoSat-2 along-track measurements of sea ice freeboard is successfully demonstrated using a 3D-Var assimilation scheme, NEMOVAR, within the Met Office's global, coupled ocean–sea-ice model, Forecast Ocean Assimilation Model (FOAM). The CryoSat-2 Arctic freeboard measurements are produced by the Centre for Polar Observation and Modelling (CPOM) and are converted to SIT within FOAM using modelled snow depth. This is the first time along-track observations of SIT have been used in this way, with other centres assimilating gridded and temporally averaged observations. The assimilation leads to improvements in the SIT analysis and forecast fields generated by FOAM, particularly in the Canadian Arctic. Arctic-wide observation-minus-background assimilation statistics for 2015–2017 show improvements of 0.75 m mean difference and 0.41 m root-mean-square difference (RMSD) in the freeze-up period and 0.46 m mean difference and 0.33 m RMSD in the ice break-up period. Validation of the SIT analysis against independent springtime in situ SIT observations from NASA Operation IceBridge (OIB) shows improvement in the SIT analysis of 0.61 m mean difference (0.42 m RMSD) compared to a control without SIT assimilation. Similar improvements are seen in the FOAM 5 d SIT forecast. Validation of the SIT assimilation with independent Beaufort Gyre Exploration Project (BGEP) sea ice draft observations does not show an improvement, since the assimilated CryoSat-2 observations compare similarly to the model without assimilation in this region. Comparison with airborne electromagnetic induction (Air-EM) combined measurements of SIT and snow depth shows poorer results for the assimilation compared to the control, despite covering similar locations to the OIB and BGEP datasets. This may be evidence of sampling uncertainty in the matchups with the Air-EM validation dataset, owing to the limited number of observations available over the time period of interest. This may also be evidence of noise in the SIT analysis or uncertainties in the modelled snow depth, in the assimilated SIT observations, or in the data used for validation. The SIT analysis could be improved by upgrading the observation uncertainties used in the assimilation. Despite the lack of CryoSat-2 SIT observations available for assimilation over the summer due to the detrimental effect of melt ponds on retrievals, it is shown that the model is able to retain improvements to the SIT field throughout the summer months due to prior, wintertime SIT assimilation. This also results in regional improvements to the July modelled sea ice concentration (SIC) of 5 % RMSD in the European sector, due to slower melt of the thicker sea ice.

Automated detection of diabetic retinopathy using custom convolutional neural network

Diabetic retinopathy is an eye deficiency that affects retina as a result of the patient having diabetes mellitus caused by high sugar levels, which may eventually lead to macular edema. The objective of this study is to design and compare several deep learning models that detect severity of diabetic retinopathy, determine risk of leading to macular edema, and segment different types of disease patterns using retina images. Indian Diabetic Retinopathy Image Dataset (IDRiD) dataset was used for disease grading and segmentation. Since images of the dataset have different brightness and contrast, we employed three techniques for generating processed images from the original images, which include brightness, color and, contrast (BCC) enhancing, color jitters (CJ), and contrast limited adaptive histogram equalization (CLAHE). After image preporcessing, we used pre-trained ResNet50, VGG16, and VGG19 models on these different preprocessed images both for determining the severity of the retinopathy and also the chances of macular edema. UNet was also applied to segment different types of diseases. To train and test these models, image dataset was divided into training, testing, and validation data at 70%, 20%, and 10% ratios, respectively. During model training, data augmentation method was also applied to increase the number of training images. Study results show that for detecting the severity of retinopathy and macular edema, ResNet50 showed the best accuracy using BCC and original images with an accuracy of 60.2% and 82.5%, respectively, on validation dataset. In segmenting different types of diseases, UNet yielded the highest testing accuracy of 65.22% and 91.09% for microaneurysms and hard exudates using BCC images, 84.83% for optic disc using CJ images, 59.35% and 89.69% for hemorrhages and soft exudates using CLAHE images, respectively. Thus, image preprocessing can play an important role to improve efficacy and performance of deep learning models.

A comparison of deep learning models for end-to-end face-based video retrieval in unconstrained videos

Face-based video retrieval (FBVR) is the task of retrieving videos that containing the same face shown in the query image. In this article, we present the first end-to-end FBVR pipeline that is able to operate on large datasets of unconstrained, multi-shot, multi-person videos. We adapt an existing audiovisual recognition dataset to the task of FBVR and use it to evaluate our proposed pipeline. We compare a number of deep learning models for shot detection, face detection, and face feature extraction as part of our pipeline on a validation dataset made of more than 4000 videos. We obtain 97.25% mean average precision on an independent test set, composed of more than 1000 videos. The pipeline is able to extract features from videos at $$\sim $$ ∼ 7 times the real-time speed, and it is able to perform a query on thousands of videos in less than 0.5 s.

Implementation and Comparison of Two Pharmacometric Tools for Model-Based Therapeutic Drug Monitoring and Precision Dosing of Daptomycin

Daptomycin is a candidate for therapeutic drug monitoring (TDM). The objectives of this work were to implement and compare two pharmacometric tools for daptomycin TDM and precision dosing. A nonparametric population PK model developed from patients with bone and joint infection was implemented into the BestDose software. A published parametric model was imported into Tucuxi. We compared the performance of the two models in a validation dataset based on mean error (ME) and mean absolute percent error (MAPE) of individual predictions, estimated exposure and predicted doses necessary to achieve daptomycin efficacy and safety PK/PD targets. The BestDose model described the data very well in the learning dataset. In the validation dataset (94 patients, 264 concentrations), 21.3% of patients were underexposed (AUC24h < 666 mg.h/L) and 31.9% of patients were overexposed (Cmin > 24.3 mg/L) on the first TDM occasion. The BestDose model performed slightly better than the model in Tucuxi (ME = −0.13 ± 5.16 vs. −1.90 ± 6.99 mg/L, p < 0.001), but overall results were in agreement between the two models. A significant proportion of patients exhibited underexposure or overexposure to daptomycin after the initial dosage, which supports TDM. The two models may be useful for model-informed precision dosing.