Convolutional Neural Network Predictions Research Articles

Abstract PO5-07-03: Integrating mechanism-based and data-driven modeling to predict the response of triple negative breast cancer to therapy

Abstract Introduction Neoadjuvant chemotherapy (NAC) is the standard-of-care (SOC) for patients with locally advanced triple-negative breast cancer (TNBC) [1]. Unfortunately, only about half of TNBC patients achieve a pathological complete response (pCR) at the completion of NAC [2]. With the recent FDA approval of immunotherapy, the response rate improved by about 8% [3]. To try to further improve NAC response rates, we build upon methods that predict treatment response to optimize interventions and outcomes. Our previous studies have demonstrated that excellent predictive accuracy of TNBC response to NAC was achieved through calibrating mechanism-based models to both pre- and on- treatment imaging data [4]. However, by requiring on-treatment imaging data, therapy optimization can only occur after a patient begins NAC. Here, our goal is to relax the requirement of on-treatment imaging data such that we can predict, and potentially optimize, a patient’s response to NAC before initiating it. We will accomplish this through combining data-driven approaches with mechanism-based modeling to obtain spatiotemporal predictions of tumor response prior to NAC. More specifically, we integrate a reaction-diffusion equation based mathematical model with a U-Net based convolutional neural network (CNN). Methods The reaction-diffusion equation models the development of tumor cellularity over time as the sum of tumor cell diffusion, controlled by a fixed diffusivity, and tumor cell proliferation, controlled by a calibrated net proliferation rate. The net proliferation rate implicitly accounts for both proliferative and drug induced death effects. We calibrated the net proliferation rates in this model on a patient-specific basis for 128 patients from the ARTEMIS trial (NCT02276443) using three imaging timepoints – pre- (V1), post two cycles (V2), and post four cycles (V3) of A/C. We then employed a CNN to characterize the relationship between the calibrated net proliferation rates and the pre-treatment imaging data. The CNN takes pretreatment imaging inputs of the apparent diffusion coefficient map, percent enhancement maps from the dynamic contrast enhanced MRI timecourse, and precontrast T1 map. The CNN outputs an estimate of the calibrated net proliferation rate which can be used to run the mathematical model forward in time to predict cellularity maps at V2/3. In training the CNN, our loss considers both the concordance correlation coefficient (CCC) and the normalized root mean square error (NRMSE). We calculated loss on the net proliferation rate predictions as NRMSE with a CCC penalty to aid in characterizing the non-Gaussian distribution of the net proliferation rate. We calculated loss on the cellularity predictions as the CCC between predicted and measured voxel-wise cellularity. The CNN was trained using the Adam optimizer and a fivefold cross validation with a withheld test set of 26 patients. Results Using the CNN predictions to forward run the mechanism-based model in time to V3 we calculated change in total tumor cellularity and volume from V1 to V3. In the test cohort, we obtained CCC values between the predicted and measured changes of 0.95 and 0.91 for cellularity and volume, respectively. Discussion and summary Through integrating mechanistic modeling and deep learning, we can accurately predict the spatiotemporal development of TNBC response to NAC using only pretreatment imaging data. We will build upon these results by expanding our model to include separate proliferation and drug induced death terms to allow for therapy intervention and optimization. Additionally, we will incorporate other pretreatment data types including genetic data. [1] Liedtke et al., J of Clin Oncol, 2008. [2] Spring et al., Clin Cancer Res, 2020. [3] Schmid et al., NEJM, 2020. [4] Wu et al., Cancer Res, 2022. Citation Format: Casey Stowers, Chengyue Wu, Sidharth Kumar, Zhan Xu, Clinton Yam, Jong Bum Son, Jingfei Ma, Jonathan Tamir, Gaiane Rauch, Thomas Yankeelov. Integrating mechanism-based and data-driven modeling to predict the response of triple negative breast cancer to therapy [abstract]. In: Proceedings of the 2023 San Antonio Breast Cancer Symposium; 2023 Dec 5-9; San Antonio, TX. Philadelphia (PA): AACR; Cancer Res 2024;84(9 Suppl):Abstract nr PO5-07-03.

Transformer Encoder Enhanced by an Adaptive Graph Convolutional Neural Network for Prediction of Aero-Engines’ Remaining Useful Life

Accurate prediction of remaining useful life (RUL) plays a significant role in ensuring the safe flight of aircraft. With the recent rapid development of deep learning, there has been a growing trend towards more precise RUL prediction. However, while many current deep learning methods are capable of extracting spatial features—those along the sensor dimension—through convolutional kernels or fully connected layers, their extraction capacity is often limited due to the small scale of kernels and the high uncertainty associated with linear weights. Graph neural networks (GNNs), emerging as effective approaches for processing graph-structured data, explicitly consider the relationships between sensors. This is akin to imposing a constraint on the training process, thereby allowing the learned results to better approximate real-world situations. In order to address the challenge of GNNs in extracting temporal features, we augment our proposed framework for RUL prediction with a Transformer encoder, resulting in the adaptive graph convolutional transformer encoder (AGCTE). A case study using the C-MAPSS dataset is conducted to validate the effectiveness of our proposed model.

Enhancing Urban Traffic Management Through Hybrid Convolutional and Graph Neural Network Integration

Traffic congestion has made city planning and citizen well-being difficult due to fast city growth and the increasing number of vehicles. Traditional traffic management fails to solve urban transportation's ever-changing issues. Traffic prediction and control systems are vital for enhancing Traffic Flow (TF) and minimizing congestion. Smart cities need advanced prediction models to regulate urban TF as traffic management becomes more complex. This paper introduces a hybrid Convolutional Neural Networks (CNN) and Graph Neural Networks (GNN) model for better real-time traffic management. The hybrid model combines CNNs' spatial feature extraction with GNNs' structural and relational data processing to analyze and predict traffic conditions. Traffic camera images are pre-processed to extract spatial characteristics. Traffic network graph construction is used for structural research. The model accurately captures traffic topology and space. The proposed method sequentially processes spatial data with CNNs and integrates them with GNNs. The final hybrid model is trained on one year of traffic data from diverse circumstances and events. The hybrid model is compared to CNN, GNN, and traditional Traffic Prediction Models (TPM) like ARIMA and SVM utilizing MAE, RMSE, and MAPE. The hybrid GNN+CNN model outperforms benchmark models with lower MAE, RMSE, and MAPE across several prediction intervals.

The Prediction of Pervious Concrete Compressive Strength Based on a Convolutional Neural Network

To overcome limitations inherent in existing mechanical performance prediction models for pervious concrete, including material constraints, limited applicability, and inadequate accuracy, this study employs a deep learning approach to construct a Convolutional Neural Network (CNN) model with three convolutional modules. The primary objective of the model is to precisely predict the 28-day compressive strength of pervious concrete. Eight input variables, encompassing coarse and fine aggregate content, water content, admixture content, cement content, fly ash content, and silica fume content, were selected for the model. The dataset utilized for both model training and testing consists of 111 sample sets. To ensure the model’s coverage within the practical range of pervious concrete strength and to enhance its robustness in real-world applications, an additional 12 sets of experimental data were incorporated for training and testing. The research findings indicate that, in comparison to the conventional machine learning method of Backpropagation (BP) neural networks, the developed CNN prediction model in this paper demonstrates a higher coefficient of determination, reaching 0.938, on the test dataset. The mean absolute percentage error is 9.13%, signifying that the proposed prediction model exhibits notable accuracy and universality in predicting the 28-day compressive strength of pervious concrete, regardless of the materials used in its preparation.

A macro-microscopic traffic flow data-driven optimal control strategy for freeway

To estimate the amount of emissions, most state-of-the-art microscopic emission models, such as VT-micro, takes the individual vehicle speed and acceleration as the model input, which can be collected efficiently with V2I technology. However, there is a gap in freeway traffic control since most of them rely on the macroscopic traffic model and omit the individual vehicle status. To fill this gap, this study proposed an individual vehicle status prediction method that utilized the convolutional neural network (CNN) for freeway proactive controls. Then the overall performance of the road network in multi-objective, namely mobility, safety, and emissions, will be evaluated to determine the optimal control signal. The proposed CNN enabled individual vehicle status prediction method reported a good match to the ground truth data compared with the support vector machine and artificial neural network. Furthermore, a field data-based simulation platform was established to implement the proposed control algorithm with the CNN prediction network. The result showed that the multi-objective performance was significantly improved compared with the uncontrolled case and achieved further optimization of multi-objective compared with the original model.

A Combined Method for Short-Term Load Forecasting Considering the Characteristics of Components of Seasonal and Trend Decomposition Using Local Regression

In response to the complexity and high volatility of original load data affecting the accuracy of load forecasting, a combined method for short-term load forecasting considering the characteristics of components of seasonal and trend decomposition using local regression (STL) is proposed. The original load data are decomposed into a trend component, seasonal component, and residual component using STL. Then, considering the characteristics of each component, a long short-term memory (LSTM) neural network, a convolutional neural network (CNN), and Gaussian process regression (GPR) are used to predict the trend component, seasonal component, and residual component, respectively. The final outcome of the load forecasting is obtained by summing the forecasted results of each individual component. A specific case study is conducted to compare the proposed combined method with LSTM, CNN, GPR, STL-LSTM, STL-CNN, and STL-GPR prediction methods. Through comparison, the proposed combined method exhibits lower errors and higher accuracy, demonstrating the effectiveness of this method.

Dislocation descriptors of low and high angle grain boundaries with convolutional neural networks

Multiscale modeling of plasticity in polycrystalline metals is a long-standing challenge in part because of the lack of an accepted grain boundary descriptor, which has hindered the bridging of scales between atomistic simulations and meso-scale discrete dislocation dynamics (DDD) models. While grain boundary dislocations (GBDs) for low angle grain boundaries can be ascertained by Burgers circuit analyses, the dislocation structures of high angle grain boundaries have remained elusive because of overlapping dislocation core fields. Here, we use convolutional neural networks (CNNs) to establish the locations of GBDs responsible for the misorientations of <001> symmetrical-tilt Cu grain boundaries, from the local atomistic stress fields modeled with molecular dynamics (MD) simulations. We achieve accurate CNN predictions of GBDs with sub-angstrom resolution across an extensive set of low- to high-tilt-angle grain boundaries through a training dataset, generated from superposing continuum-representations of the MD stress fields of single dislocations that consider the contributions of both the Volterra and dislocation core fields. The approach paves the way for dislocation representations of the atomistic grain boundary structures modeled by MD simulations or density functional theory (DFT) calculations in mesoscale DDD models to elucidate multiscale plasticity effects.

Prediction of Diabetic Retinopathy using Deep Learning with Preprocessing

INTRODUCTION: When Diabetic Retinopathy (DR) is not identified promptly; it frequently results in sight impairment. To properly diagnose and treat DR, preprocessing of picture methods and precise prediction models are essential. With the help of numerous well-liked filters and a Deep CNN (Convolutional Neural Network) model, the comprehensive method for DR image preparation and prognosis presented in this research is described. Using the filters that focus boundaries and contours in the ocular pictures is the first step in the initial processing stage. This procedure tries to find anomalies linked to DR. By the usage of filters, the excellence of pictures can be developed and minimize disturbances, preserving critical information. The Deep CNN algorithm has been trained to generate forecasts on the cleaned retinal pictures following the phase of preprocessing. The filters efficiently eliminate interference without sacrificing vital data. Convolutional type layers, pooling type layers, and fully associated layers are used in the CNN framework, which was created especially for image categorization tasks, to acquire data and understand the relationships associated with DR.&#x0D; OBJECTIVES: Using image preprocessing techniques such as the Sobel, Wiener, Gaussian, and non-local mean filters is a promising approach for DR analysis. Then, predicting using a CNN completes the approach. These preprocessing filters enhance the images and prepare them for further examination. The pre-processed images are fed into a CNN model. The model extracts significant information from the images by identifying complex patterns. DR or classification may be predicted by the CNN model through training on a labeled dataset.&#x0D; METHODS: The Method Preprocessing is employed for enhancing the clarity and difference of retina fundus picture by removing noise and fluctuation. The preprocessing stage is utilized for the normalization of the pictures and non-uniform brightness adjustment in addition to contrast augmentation and noise mitigation to remove noises and improve the rate of precision of the subsequent processing stages.&#x0D; RESULTS: To improve image quality and reduce noise, preprocessing techniques including Sobel, Wiener, Gaussian, and non-local mean filters are frequently employed in image processing jobs. For a particular task, the non-local mean filter produces superior results; for enhanced performance, it may be advantageous to combine it with a CNN. Before supplying the processed images to the CNN for prediction, the non-local mean filter can assist reduce noise and improve image details.&#x0D; CONCLUSION: A promising method for DR analysis entails the use of image preprocessing methods such as the Sobel, Wiener, Gaussian, and non-local mean filters, followed by prediction using a CNN. These preprocessing filters improve the photos and get them ready for analysis. After being pre-processed, the photos are sent into a CNN model, which uses its capacity to discover intricate patterns to draw out important elements from the images. The CNN model may predict DR or classification by training it on a labeled dataset. The development of computer-aided diagnosis systems for DR is facilitated by the integration of CNN prediction with image preprocessing filters. This strategy may increase the effectiveness of healthcare workers, boost patient outcomes, and lessen the burden of DR.

Artificial neural network and convolutional neural network for prediction of dental caries

Dental caries has high prevalence among kids and adults thus it has become one of the global health concerns. The current modern dentistry focused on the preventives measures to reduce the number of dental caries cases. The employment of machine learning coupled with UV spectroscopy plays a crucial role to detect the early stage of caries. Artificial neural network with hyperparameter tuning was employed to train spectral data for the classification based on the International Caries Detection and Assesment System (ICDAS). Spectra preprocessing namely mean center (MC), autoscale (AS) and Savitzky Golay smoothing (SG) were applied on the data for spectra correction. The best performance of ANN model obtained has accuracy of 0.85 with precision of 1.00. Convolutional neural network (CNN) combined with Savitzky Golay smoothing performed on the spectral data has accuracy, precision, sensitivity and specificity for validation data of 1.00 respectively. The result obtained shows that the application of ANN and CNN capable to produce robust model to be used as an early screening of dental caries.

The model transferability of AI in digital pathology : Potential and reality

Artificial intelligence (AI) holds the potential to make significant advancements in pathology. However, its actual implementation and certification for practical use are currently limited, often due to challenges related to model transferability. In this context, we investigate the factors influencing transferability and present methods aimed at enhancing the utilization of AI algorithms in pathology. Various convolutional neural networks (CNNs) and vision transformers (ViTs) were trained using datasets from two institutions, along with the publicly available TCGA-MIBC dataset. These networks conducted predictions in urothelial tissue and intrahepatic cholangiocarcinoma (iCCA). The objective was to illustrate the impact of stain normalization, the influence of various artifacts during both training and testing, as well as the effects of the NoisyEnsemble method. We were able to demonstrate that stain normalization of slides from different institutions has asignificant positive effect on the inter-institutional transferability of CNNs and ViTs (respectively +13% and +10%). In addition, ViTs usually achieve ahigher accuracy in the external test (here +1.5%). Similarly, we showcased how artifacts in test data can negatively affect CNN predictions and how incorporating these artifacts during training leads to improvements. Lastly, NoisyEnsembles of CNNs (better than ViTs) were shown to enhance transferability across different tissues and research questions (+7% Bladder, +15% iCCA). It is crucial to be aware of the transferability challenge: achieving good performance during development does not necessarily translate to good performance in real-world applications. The inclusion of existing methods to enhance transferability, such as stain normalization and NoisyEnsemble, and their ongoing refinement, is of importance.

Prediction of seasonal sea surface temperature based on temperature and salinity of subsurface ocean using machine learning

AbstractThe sea surface temperature (SST) is not only a crucial external factor in the evolution of the atmosphere, but also a primary factor and premonition signal used in climate prediction. It is challenging to obtain a precise SST for generating accurate initial and boundary conditions in numerical models. This study employs a machine learning approach, that is, a convolutional neural network (CNN) algorithm, to predict SST on a seasonal scale. In particular, the subsurface ocean temperature (OT) and ocean salinity (OS) at depths of 5.02, 15.08, 25.16, 35.28, 45.45 and 76.55 m were used as training factors in developing a CNN prediction model. The results indicate that subsurface OT and OS can persist for 6 months or longer, with a maximum persistence of up to 12 months. Using the CNN prediction model, the SST can be reliably predicted 6 months in advance in most cases. The predicted SST has a mean bias of approximately 0–0.8 K on the globe. The bias is small (below 0.5 K) in the open ocean. The root mean square errors (RMSEs) of hindcasting for Interdecadal Pacific Oscillation, North Atlantic Oscillation (NAO) and Atlantic Multidecadal Oscillation indices are all less than 1.0 K. Specifically, the RMSE for El Niño prediction is less than 0.5 K. This study provides a viable method for establishing initial and boundary conditions for climate prediction.

Deep learning-based differentiation of peripheral high-flow and low-flow vascular malformations in T2-weighted short tau inversion recovery MRI.

Differentiation of high-flow from low-flow vascular malformations (VMs) is crucial for therapeutic management of this orphan disease. A convolutional neural network (CNN) was evaluated for differentiation of peripheral vascular malformations (VMs) on T2-weighted short tau inversion recovery (STIR) MRI. 527 MRIs (386 low-flow and 141 high-flow VMs) were randomly divided into training, validation and test set for this single-center study. 1) Results of the CNN's diagnostic performance were compared with that of two expert and four junior radiologists. 2) The influence of CNN's prediction on the radiologists' performance and diagnostic certainty was evaluated. 3) Junior radiologists' performance after self-training was compared with that of the CNN. Compared with the expert radiologists the CNN achieved similar accuracy (92% vs. 97%, p = 0.11), sensitivity (80% vs. 93%, p = 0.16) and specificity (97% vs. 100%, p = 0.50). In comparison to the junior radiologists, the CNN had a higher specificity and accuracy (97% vs. 80%, p < 0.001; 92% vs. 77%, p < 0.001). CNN assistance had no significant influence on their diagnostic performance and certainty. After self-training, the junior radiologists' specificity and accuracy improved and were comparable to that of the CNN. Diagnostic performance of the CNN for differentiating high-flow from low-flow VM was comparable to that of expert radiologists. CNN did not significantly improve the simulated daily practice of junior radiologists, self-training was more effective.

Stochastic properties of coastal flooding events – Part 1: convolutional-neural-network-based semantic segmentation for water detection

Abstract. The frequency and intensity of coastal flooding is expected to accelerate in low-elevation coastal areas due to sea level rise. Coastal flooding due to wave overtopping affects coastal communities and infrastructure; however, it can be difficult to monitor in remote and vulnerable areas. Here we use a camera-based system to measure beach and back-beach flooding as part of the after-storm recovery of an eroded beach on the Texas coast. We analyze high-temporal resolution images of the beach using convolutional neural network (CNN)-based semantic segmentation to study the stochastic properties of flooding events. In the first part of this work, we focus on the application of semantic segmentation to identify water and overtopping events. We train and validate a CNN with over 500 manually classified images and introduce a post-processing method to reduce false positives. We find that the accuracy of CNN predictions of water pixels is around 90 % and strongly depends on the number and diversity of images used for training.

Moisture content online detection system based on multi-sensor fusion and convolutional neural network.

To monitor the moisture content of agricultural products in the drying process in real time, this study applied a model combining multi-sensor fusion and convolutional neural network (CNN) to moisture content online detection. This study built a multi-sensor data acquisition platform and established a CNN prediction model with the raw monitoring data of load sensor, air velocity sensor, temperature sensor, and the tray position as input and the weight of the material as output. The model's predictive performance was compared with that of the linear partial least squares regression (PLSR) and nonlinear support vector machine (SVM) models. A moisture content online detection system was established based on this model. Results of the model performance comparison showed that the CNN prediction model had the optimal prediction effect, with the determination coefficient (R2) and root mean square error (RMSE) of 0.9989 and 6.9, respectively, which were significantly better than those of the other two models. Results of validation experiments showed that the detection system met the requirements of moisture content online detection in the drying process of agricultural products. The R2 and RMSE were 0.9901 and 1.47, respectively, indicating the good performance of the model combining multi-sensor fusion and CNN in moisture content online detection for agricultural products in the drying process. The moisture content online detection system established in this study is of great significance for researching new drying processes and realizing the intelligent development of drying equipment. It also provides a reference for online detection of other indexes in the drying process of agricultural products.

Research on prediction model of coal spontaneous combustion temperature based on SSA-CNN

To predict the coal spontaneous combustion temperature accurately and efficiently, this study proposes a model based on the Sparrow Search Algorithm (SSA) and Convolutional Neural Network (CNN). Firstly, the study analyzes the main gas reactions during the coal oxidation to pyrolysis process. Six gas indicators, namely O2, CO, C2H4, CO/ΔO2, C2H4/C2H6, and C2H6, are closely related to coal temperature. Subsequently, a prediction indicator system is established. Then, the excellent data mining capabilities of CNN are leveraged through deep learning, along with their unique advantages in local perception and weight sharing, and a CNN prediction model framework is constructed. Moreover, the comparison between the algorithm performances is executed and SSA is selected for optimization. Utilizing its exceptional global search capability and adaptability, SSA optimizes the seven hyper-parameters of the model, significantly enhancing prediction accuracy. In the final step, SSA-CNN is compared with five reference models on test samples. The SSA-CNN model showcases a maximum relative error of 0.155, outperforming other models. Moreover, the RMSE of this model yields 8.4500, which is also lower than other models. The results suggest that the combination of the selected gas indicators with the SSA-CNN model can accurately predict the spontaneous combustion temperature of coal.

HACNet: End-to-end learning of interpretable table-to-image converter and convolutional neural network

Motivated by the high prediction performance of convolutional neural networks (CNNs), several works have applied them to tabular datasets. As CNNs are built to accept images, several transformations of tabular data have been proposed to obtain images. However, existing methods transform the tabular data into images prior to CNN training, which fails to take the prediction error into account. Additionally, they employ all features from the tables, including unimportant ones, to produce the images. Moreover, the created images might not become human-interpretable because they do not consider the interpretability of images as a metric. To overcome these problems, we propose a hard attention-based converter combined with a convolutional neural network (HACNet), consisting of an attention-based table-to-image converter and a CNN-based predictor. HACNet trains its components simultaneously by minimizing CNN prediction loss and mean squared error (MSE) between created and template images. Minimizing this MSE loss allows us to visually distinguish the created images with different labels. The attention-based converter selects exactly one feature for each pixel in the image via its hard attention mechanism with Gumbel-Softmax, enabling feature selection. We experimentally show that HACNet produces human-interpretable images, reduces used features, and achieves prediction performances comparative with existing methods on several benchmark datasets.

Deep learning prediction of galaxy stellar populations in the low-redshift Universe

ABSTRACT The increasing size and complexity of data provided by both ongoing and planned galaxy surveys greatly contribute to our understanding of galaxy evolution. Deep learning methods are particularly well suited for handling the complex and massive data. We train a convolutional neural network (CNN) to simultaneously predict the stellar populations in galaxies: age, metallicity, colour excess E(B − V), and central velocity dispersion (VD) using spectra with redshift ≤ 0.3 from the Sloan Digital Sky Survey. This is the first time to use spectra based on deep learning to derive the four galaxy properties. The testing results show that our CNN predictions of galaxy properties are in good consistent with values by the traditional stellar population synthesis method with little scatters (0.11 dex for age and metallicity, 0.018 mag for E(B − V), and 31 km s−1 for VD). In terms of the computational time, our method reduces by more than 10 times compared to traditional method. We further evaluate the performance of our CNN prediction model using spectra with different signal-to-noise ratios (S/Ns), redshifts, and spectral classes. We find that our model generally exhibits good performance, although the errors at different S/Ns, redshifts, and spectral classes vary slightly. Our well-trained CNN model and related codes are publicly available on https://github.com/sddzwll/CNNforStellarp.

CNN-based flow field prediction for bus aerodynamics analysis

The aim of this article is to evaluate the ability of a convolutional neural network (CNN) to predict velocity and pressure aerodynamic fields in heavy vehicles. For training and testing the developed CNN, various CFD simulations of three different vehicle geometries have been conducted, considering the RANS-based k-ω SST turbulent model. Two geometries correspond to the SC7 and SC5 coach models of the bus manufacturer SUNSUNDEGUI and the third one corresponds to Ahmed body. By generating different variants of these three geometries, a large number of representations of the velocity and pressure fields are obtained that will be used to train, verify, and evaluate the convolutional neural network. To improve the accuracy of the CNN, the field representations obtained are discretized as a function of the expected velocity gradient, so that in the areas where there is a greater variation in velocity, the corresponding neuron is smaller. The results show good agreement between numerical results and CNN predictions, being the CNN able to accurately represent the velocity and pressure fields with very low errors. Additionally, a substantial improvement in the computational time needed for each simulation is appreciated, reducing it by four orders of magnitude.

Design and validation of an artificial intelligence system to detect the quality of colon cleansing before colonoscopy

Background and aimsPatients’ perception of their bowel cleansing quality may guide rescue cleansing strategies before colonoscopy. The main aim of this study was to train and validate a convolutional neural network (CNN) for classifying rectal effluent during bowel preparation intake as “adequate” or “inadequate” cleansing before colonoscopy. Patients and methodsPatients referred for outpatient colonoscopy were asked to provide images of their rectal effluent during the bowel preparation process. The images were categorized as adequate or inadequate cleansing based on a predefined 4-picture quality scale. A total of 1203 images were collected from 660 patients. The initial dataset (799 images), was split into a training set (80%) and a validation set (20%). The second dataset (404 images) was used to develop a second test of the CNN accuracy. Afterward, CNN prediction was prospectively compared with the Boston Bowel Preparation Scale (BBPS) in 200 additional patients who provided a picture of their last rectal effluent. ResultsOn the initial dataset, a global accuracy of 97.49%, a sensitivity of 98.17% and a specificity of 96.66% were obtained using the CNN model. On the second dataset, an accuracy of 95%, a sensitivity of 99.60% and a specificity of 87.41% were obtained. The results from the CNN model were significantly associated with those from the BBPS (P<0.001), and 77.78% of the patients with poor bowel preparation were correctly classified. ConclusionThe designed CNN is capable of classifying “adequate cleansing” and “inadequate cleansing” images with high accuracy.

Simultaneous Quantitative Determination of Low-Concentration Preservatives and Heavy Metals in Tricholoma Matsutakes Based on SERS and FLU Spectral Data Fusion.

As an ingredient of great economic value, Tricholoma matsutake has received widespread attention. However, heavy metal residues and preservatives in it will affect the quality of Tricholoma matsutake and endanger the health of consumers. Here, we present a method for the simultaneous detection of low concentrations of potassium sorbate and lead in Tricholoma matsutakes based on surface-enhanced Raman spectroscopy (SERS) and fluorescence (FLU) spectroscopy to test the safety of consumption. Data fusion strategies combined with multiple machine learning methods, including partial least-squares regression (PLSR), deep forest (DF) and convolutional neural networks (CNN) are used for model training. The results show that combined with reasonable band selection, the CNN prediction model based on decision-level fusion achieves the best performance, the correlation coefficients (R2) were increased to 0.9963 and 0.9934, and the root mean square errors (RMSE) were reduced to 0.0712 g·kg-1 and 0.0795 mg·kg-1, respectively. The method proposed in this paper accurately predicts preservatives and heavy metals remaining in Tricholoma matsutake and provides a reference for other food safety testing.