A simple plug-in bagging ensemble based on threshold-moving for classifying binary and multiclass imbalanced data

A reinforcement learning model for AI-based decision support in skin cancer

We investigated whether human preferences hold the potential to improve diagnostic artificial intelligence (AI)-based decision support using skin cancer diagnosis as a use case. We utilized nonuniform rewards and penalties based on expert-generated tables, balancing the benefits and harms of various diagnostic errors, which were applied using reinforcement learning. Compared with supervised learning, the reinforcement learning model improved the sensitivity for melanoma from 61.4% to 79.5% (95% confidence interval (CI): 73.5-85.6%) and for basal cell carcinoma from 79.4% to 87.1% (95% CI: 80.3-93.9%). AI overconfidence was also reduced while simultaneously maintaining accuracy. Reinforcement learning increased the rate of correct diagnoses made by dermatologists by 12.0% (95% CI: 8.8-15.1%) and improved the rate of optimal management decisions from 57.4% to 65.3% (95% CI: 61.7-68.9%). We further demonstrated that the reward-adjusted reinforcement learning model and a threshold-based model outperformed naïve supervised learning in various clinical scenarios. Our findings suggest the potential for incorporating human preferences into image-based diagnostic algorithms.

Cost-Sensitive YOLOv5 for Detecting Surface Defects of Industrial Products

Owing to the remarkable development of deep learning algorithms, defect detection techniques based on deep neural networks have been extensively applied in industrial production. Most existing surface defect detection models assign equal costs to the classification errors among different defect categories but do not strictly distinguish them. However, various errors can generate a great discrepancy in decision risk or classification costs and then produce a cost-sensitive issue that is crucial to the manufacturing process. To address this engineering challenge, we propose a novel supervised classification cost-sensitive learning method (SCCS) and apply it to improve YOLOv5 as CS-YOLOv5, where the classification loss function of object detection was reconstructed according to a new cost-sensitive learning criterion explained by a label-cost vector selection method. In this way, the classification risk information from a cost matrix is directly introduced into the detection model and fully exploited in training. As a result, the developed approach can make low-risk classification decisions for defect detection. It is applicable for direct cost-sensitive learning based on a cost matrix to implement detection tasks. Using two datasets of a painting surface and a hot-rolled steel strip surface, our CS-YOLOv5 model outperforms the original version with respect to cost under different positive classes, coefficients, and weight ratios, but also maintains effective detection performance measured by mAP and F1 scores.

Multiclass classification of environmental chemical stimuli from unbalanced plant electrophysiological data

Plant electrophysiological response contains useful signature of its environment and health which can be utilized using suitable statistical analysis for developing an inverse model to classify the stimulus applied to the plant. In this paper, we have presented a statistical analysis pipeline to tackle a multiclass environmental stimuli classification problem with unbalanced plant electrophysiological data. The objective here is to classify three different environmental chemical stimuli, using fifteen statistical features, extracted from the plant electrical signals and compare the performance of eight different classification algorithms. A comparison using reduced dimensional projection of the high dimensional features via principal component analysis (PCA) has also been presented. Since the experimental data is highly unbalanced due to varying length of the experiments, we employ a random under-sampling approach for the two majority classes to create an ensemble of confusion matrices to compare the classification performances. Along with this, three other multi-classification performance metrics commonly used for unbalanced data viz. balanced accuracy, F1-score and Matthews correlation coefficient have also been analyzed. From the stacked confusion matrices and the derived performance metrics, we choose the best feature-classifier setting in terms of the classification performances carried out in the original high dimensional vs. the reduced feature space, for this highly unbalanced multiclass problem of plant signal classification due to different chemical stress. Difference in the classification performances in the high vs. reduced dimensions are also quantified using the multivariate analysis of variance (MANOVA) hypothesis testing. Our findings have potential real-world applications in precision agriculture for exploring multiclass classification problems with highly unbalanced datasets, employing a combination of existing machine learning algorithms. This work also advances existing studies on environmental pollution level monitoring using plant electrophysiological data.

Confound-leakage: confound removal in machine learning leads to leakage

BACKGROUND

Machine learning (ML) approaches are a crucial component of modern data analysis in many fields, including epidemiology and medicine. Nonlinear ML methods often achieve accurate predictions, for instance, in personalized medicine, as they are capable of modeling complex relationships between features and the target. Problematically, ML models and their predictions can be biased by confounding information present in the features. To remove this spurious signal, researchers often employ featurewise linear confound regression (CR). While this is considered a standard approach for dealing with confounding, possible pitfalls of using CR in ML pipelines are not fully understood.

RESULTS

We provide new evidence that, contrary to general expectations, linear confound regression can increase the risk of confounding when combined with nonlinear ML approaches. Using a simple framework that uses the target as a confound, we show that information leaked via CR can increase null or moderate effects to near-perfect prediction. By shuffling the features, we provide evidence that this increase is indeed due to confound-leakage and not due to revealing of information. We then demonstrate the danger of confound-leakage in a real-world clinical application where the accuracy of predicting attention-deficit/hyperactivity disorder is overestimated using speech-derived features when using depression as a confound.

CONCLUSIONS

Mishandling or even amplifying confounding effects when building ML models due to confound-leakage, as shown, can lead to untrustworthy, biased, and unfair predictions. Our expose of the confound-leakage pitfall and provided guidelines for dealing with it can help create more robust and trustworthy ML models.

Toxic mechanism of the Mongolian medicine "Hunqile-7" based on metabonomics and the metabolism of intestinal flora

The traditional Mongolian medicine Hunqile-7 (HQL-7), which is mainly used to relieve pain in clinic, has certain toxicity. Therefore, toxicological investigation of HQL-7 is of great significance to its safety assessment. In this study, the toxic mechanism of HQL-7 was explored based on a combination of metabolomics and intestinal flora metabolism. UHPLC-MS was used to analyze the serum, liver and kidney samples of rats after intragastric administration of HQL-7. The decision tree and K Nearest Neighbor (KNN) model were established based on the bootstrap aggregation (bagging) algorithm to classify the omics data. After samples were extracted from rat feces, the high-throughput sequencing platform was used to analyze the 16s rRNA V3-V4 region of bacteria. The experimental results confirm that the bagging algorithm improved the classification accuracy. The toxic dose, toxic intensity, and toxic target organ of HQL-7 were determined in toxicity tests. Seventeen biomarkers were identified and the metabolism dysregulation of these biomarkers may be responsible for the toxicity of HQL-7 in vivo. Several kinds of bacteria was demonstrated to be closely related to the physiological indices of renal and liver function, indicating liver and kidney damage induced by HQL-7 may be related to the disturbance of these intestinal bacteria. Overall, the toxic mechanism of HQL-7 was revealed in vivo, which not only provides a scientific basis for the safe and rational clinical use of HQL-7, but also opens up a new field of research on big data for Mongolian medicine.

Analyzing the Therapeutic Mechanism of Mongolian Medicine Zhonglun-5 in Rheumatoid Arthritis Using a Bagging Algorithm with Serum Metabonomics

Rheumatoid arthritis (RA) is a complex autoimmune disorder. Zhonglun-5 (ZL), a traditional Mongolian medicine, exhibits an excellent clinical effect on RA; however, its molecular mechanism remains unclear. In this study, rat serum metabolomic analysis was performed to identify potential biomarkers for RA and investigate its treatment mechanism. A Dionex Ultimate 3000 ultrahigh-performance liquid chromatography system coupled with a Q-Exactive Focus Orbitrap mass spectrometer was used for metabonomics analysis. Bootstrap aggregation (bagging) classification algorithm was applied to process data from control (CG), model (MG), and treatment administration groups. The classification accuracy was 100.00% (6/6) in the decision tree model and 83.33% (5/6) in the K-nearest neighbor (KNN) model, accompanied by 18 training samples and 6 testing samples. Using volcanic map analysis, 24 biomarkers were identified between CG and MG, including those related to glycosphingolipid biosynthesis, arachidonic acid, fatty acids, amino acids, bile acids, vitamins, and sphingolipids. A set diagram of the heatmap and drug-biomarker network of potential biomarkers was constructed. After ZL administration, the levels of these biomarkers returned to normal, indicating that ZL had a therapeutic effect in rats with RA. This study established a solid theoretical foundation to promote further research on the clinical applicability of ZL.

A survey of multi-class imbalanced data classification methods

In reality, the data generated in many fields are often imbalanced, such as fraud detection, network intrusion detection and disease diagnosis. The class with fewer instances in the data is called the minority class, and the minority class in some applications contains the significant information. So far, many classification methods and strategies for binary imbalanced data have been proposed, but there are still many problems and challenges in multi-class imbalanced data that need to be solved urgently. The classification methods for multi-class imbalanced data are analyzed and summarized in terms of data preprocessing methods and algorithm-level classification methods, and the performance of the algorithms using the same dataset is compared separately. In the data preprocessing methods, the methods of oversampling, under-sampling, hybrid sampling and feature selection are mainly introduced. Algorithm-level classification methods are comprehensively introduced in four aspects: ensemble learning, neural network, support vector machine and multi-class decomposition technique. At the same time, all data preprocessing methods and algorithm-level classification methods are analyzed in detail in terms of the techniques used, comparison algorithms, pros and cons, respectively. Moreover, the evaluation metrics commonly used for multi-class imbalanced data classification methods are described comprehensively. Finally, the future directions of multi-class imbalanced data classification are given.

Detection of Common Cold from Speech Signals using Deep Neural Network

This paper presents a deep learning-based analysis and classification of cold speech observed when a person is diagnosed with the common cold. The common cold is a viral infectious disease that affects the throat and the nose. Since speech is produced by the vocal tract after linear filtering of excitation source information, during a common cold, its attributes are impacted by the throat and the nose. The proposed study attempts to develop a deep learning-based classification model that can accurately predict whether a person has a cold or not based on their speech. The common cold-related information is captured using Mel-frequency cepstral coefficients (MFCC) and linear predictive coding (LPC) from the speech signal. The data imbalance is handled using the sampling strategy, SMOTE-Tomek links. Then, utilizing MFCC and LPC features, a deep learning-based model is trained and then used to categorize cold speech. The performance of a deep learning-based method is compared to logistic regression, random forest, and gradient boosted tree classifiers. The proposed model is less complex and uses a smaller feature set while giving comparable results to other state-of-the-art methods. The proposed method gives an UAR of 67.71 % , higher than the benchmark OpenSMILE SVM result of 64 % . The study's success will yield a noninvasive method for cold detection, which can further be extended to detect other speech-affecting pathologies.

Subwavelength ultrasonic imaging using a deep convolutional neural network trained on structural noise

Subwavelength ultrasonic imaging (SUI) can detect subwavelength flaws beyond the diffraction limit, however, SUI sometimes fails to clearly reveal flaws in C-scans when the signal-to-noise ratio (SNR) is low. In this work, a convolutional neural network (CNN) that takes structural noise into account is developed for SUI to distinguish flaw echoes from structural noise. The network contains a regression CNN for learning features from the structural noise and a learnable soft thresholding layer for classification. Experiments show that the proposed method performs well for imaging subwavelength flaws at different depths and of different sizes. It achieved an F₁ score of 97.69 ± 1.56% in detecting flaws as compared to the enhanced ultrasonic flaw detection method with time-dependent threshold. As an example of general application of the method, we also performed SUI on natural flaws in a spheroidal graphite cast iron specimen. The results show that the method can achieve SUI without a theoretical backscattering model and is not limited by noise distribution, multiple scattering, or complex microstructures. Furthermore, the network does not need to prepare flaw echoes for training.

Crash Injury Severity Prediction Using an Ordinal Classification Machine Learning Approach

In many related works, nominal classification algorithms ignore the order between injury severity levels and make sub-optimal predictions. Existing ordinal classification methods suffer rank inconsistency and rank non-monotonicity. The aim of this paper is to propose an ordinal classification approach to predict traffic crash injury severity and to test its performance over existing machine learning classification methods. First, we compare the performance of the neural network, XGBoost, and SVM classifiers in injury severity prediction. Second, we utilize a severity category-combination method with oversampling to relieve the class-imbalance problem prevalent in crash data. Third, we take advantage of probability calibration and the optimal probability threshold moving to improve the prediction ability of ordinal classification. The proposed approach can satisfy the rank consistency and rank monotonicity requirement and is proved to be superior to other ordinal classification methods and nominal classification machine learning by statistical significance test. Important factors relating to injury severity are selected based on their permutation feature importance scores. We find that converting severity levels into three classes, minor injury, moderate injury, and serious injury, can substantially improve the prediction precision.

Cost-Sensitive Siamese Network for PCB Defect Classification

After the production of printed circuit boards (PCB), PCB manufacturers need to remove defected boards by conducting rigorous testing, while manual inspection is time-consuming and laborious. Many PCB factories employ automatic optical inspection (AOI), but this pixel-based comparison method has a high false alarm rate, thus requiring intensive human inspection to determine whether alarms raised from it resemble true or pseudo defects. In this paper, we propose a new cost-sensitive deep learning model: cost-sensitive siamese network (CSS-Net) based on siamese network, transfer learning and threshold moving methods to distinguish between true and pseudo PCB defects as a cost-sensitive classification problem. We use optimization algorithms such as NSGA-II to determine the optimal cost-sensitive threshold. Results show that our model improves true defects prediction accuracy to 97.60%, and it maintains relatively high pseudo defect prediction accuracy, 61.24% in real-production scenario. Furthermore, our model also outperforms its state-of-the-art competitor models in other comprehensive cost-sensitive metrics, with an average of 33.32% shorter training time.

Improving the phishing website detection using empirical analysis of Function Tree and its variants

The phishing attack is one of the most complex threats that have put internet users and legitimate web resource owners at risk. The recent rise in the number of phishing attacks has instilled distrust in legitimate internet users, making them feel less safe even in the presence of powerful antivirus apps. Reports of a rise in financial damages as a result of phishing website attacks have caused grave concern. Several methods, including blacklists and machine learning-based models, have been proposed to combat phishing website attacks. The blacklist anti-phishing method has been faulted for failure to detect new phishing URLs due to its reliance on compiled blacklisted phishing URLs. Many ML methods for detecting phishing websites have been reported with relatively low detection accuracy and high false alarm. Hence, this research proposed a Functional Tree (FT) based meta-learning models for detecting phishing websites. That is, this study investigated improving the phishing website detection using empirical analysis of FT and its variants. The proposed models outperformed baseline classifiers, meta-learners and hybrid models that are used for phishing websites detection in existing studies. Besides, the proposed FT based meta-learners are effective for detecting legitimate and phishing websites with accuracy as high as 98.51% and a false positive rate as low as 0.015. Hence, the deployment and adoption of FT and its meta-learner variants for phishing website detection and applicable cybersecurity attacks are recommended.

Plankton classification with high-throughput submersible holographic microscopy and transfer learning

BACKGROUND

Plankton are foundational to marine food webs and an important feature for characterizing ocean health. Recent developments in quantitative imaging devices provide in-flow high-throughput sampling from bulk volumes-opening new ecological challenges exploring microbial eukaryotic variation and diversity, alongside technical hurdles to automate classification from large datasets. However, a limited number of deployable imaging instruments have been coupled with the most prominent classification algorithms-effectively limiting the extraction of curated observations from field deployments. Holography offers relatively simple coherent microscopy designs with non-intrusive 3-D image information, and rapid frame rates that support data-driven plankton imaging tasks. Classification benchmarks across different domains have been set with transfer learning approaches, focused on repurposing pre-trained, state-of-the-art deep learning models as classifiers to learn new image features without protracted model training times. Combining the data production of holography, digital image processing, and computer vision could improve in-situ monitoring of plankton communities and contribute to sampling the diversity of microbial eukaryotes.

RESULTS

Here we use a light and portable digital in-line holographic microscope (The HoloSea) with maximum optical resolution of 1.5 μm, intensity-based object detection through a volume, and four different pre-trained convolutional neural networks to classify > 3800 micro-mesoplankton (> 20 μm) images across 19 classes. The maximum classifier performance was quickly achieved for each convolutional neural network during training and reached F1-scores > 89%. Taking classification further, we show that off-the-shelf classifiers perform strongly across every decision threshold for ranking a majority of the plankton classes.

CONCLUSION

These results show compelling baselines for classifying holographic plankton images, both rare and plentiful, including several dinoflagellate and diatom groups. These results also support a broader potential for deployable holographic microscopes to sample diverse microbial eukaryotic communities, and its use for high-throughput plankton monitoring.

Assessing the predictive ability of the Suicide Crisis Inventory for near-term suicidal behavior using machine learning approaches

OBJECTIVE

This study explores the prediction of near-term suicidal behavior using machine learning (ML) analyses of the Suicide Crisis Inventory (SCI), which measures the Suicide Crisis Syndrome, a presuicidal mental state.

METHODS

SCI data were collected from high-risk psychiatric inpatients (N = 591) grouped based on their short-term suicidal behavior, that is, those who attempted suicide between intake and 1-month follow-up dates (N = 20) and those who did not (N = 571). Data were analyzed using three predictive algorithms (logistic regression, random forest, and gradient boosting) and three sampling approaches (split sample, Synthetic minority oversampling technique, and enhanced bootstrap).

RESULTS

The enhanced bootstrap approach considerably outperformed the other sampling approaches, with random forest (98.0% precision; 33.9% recall; 71.0% Area under the precision-recall curve [AUPRC]; and 87.8% Area under the receiver operating characteristic [AUROC]) and gradient boosting (94.0% precision; 48.9% recall; 70.5% AUPRC; and 89.4% AUROC) algorithms performing best in predicting positive cases of near-term suicidal behavior using this dataset.

CONCLUSIONS

ML can be useful in analyzing data from psychometric scales, such as the SCI, and for predicting near-term suicidal behavior. However, in cases such as the current analysis where the data are highly imbalanced, the optimal method of measuring performance must be carefully considered and selected.

Estimating System State through Similarity Analysis of Signal Patterns

State prediction is not straightforward, particularly for complex systems that cannot provide sufficient amounts of training data. In particular, it is usually difficult to analyze some signal patterns for state prediction if they were observed in both normal and fault-states with a similar frequency or if they were rarely observed in any system state. In order to estimate the system status with imbalanced state data characterized insufficient fault occurrences, this paper proposes a state prediction method that employs discrete state vectors (DSVs) for pattern extraction and then applies a naïve Bayes classifier and Brier scores to interpolate untrained pattern information by using the trained ones probabilistically. Each Brier score is transformed into a more intuitive one, termed state prediction power (SPP). The SPP values represent the reliability of the system state prediction. A state prediction power map, which visualizes the DSVs and corresponding SPP values, is provided a more intuitive way of state prediction analysis. A case study using a car engine fault simulator was conducted to generate artificial engine knocking. The proposed method was evaluated using holdout cross-validation, defining specificity and sensitivity as indicators to represent state prediction success rates for no-fault and fault states, respectively. The results show that specificity and sensitivity are very high (equal to 1) for high limit values of SPP, but drop off dramatically for lower limit values.

Radial-Based Oversampling for Multiclass Imbalanced Data Classification

Learning from imbalanced data is among the most popular topics in the contemporary machine learning. However, the vast majority of attention in this field is given to binary problems, while their much more difficult multiclass counterparts are relatively unexplored. Handling data sets with multiple skewed classes poses various challenges and calls for a better understanding of the relationship among classes. In this paper, we propose multiclass radial-based oversampling (MC-RBO), a novel data-sampling algorithm dedicated to multiclass problems. The main novelty of our method lies in using potential functions for generating artificial instances. We take into account information coming from all of the classes, contrary to existing multiclass oversampling approaches that use only minority class characteristics. The process of artificial instance generation is guided by exploring areas where the value of the mutual class distribution is very small. This way, we ensure a smart oversampling procedure that can cope with difficult data distributions and alleviate the shortcomings of existing methods. The usefulness of the MC-RBO algorithm is evaluated on the basis of extensive experimental study and backed-up with a thorough statistical analysis. Obtained results show that by taking into account information coming from all of the classes and conducting a smart oversampling, we can significantly improve the process of learning from multiclass imbalanced data.

Business Analytics in Telemarketing: Cost-Sensitive Analysis of Bank Campaigns Using Artificial Neural Networks

The banking industry has been seeking novel ways to leverage database marketing efficiency. However, the nature of bank marketing data hindered the researchers in the process of finding a reliable analytical scheme. Various studies have attempted to improve the performance of Artificial Neural Networks in predicting clients’ intentions but did not resolve the issue of imbalanced data. This research aims at improving the performance of predicting the willingness of bank clients to apply for a term deposit in highly imbalanced datasets. It proposes enhanced Artificial Neural Network models (i.e., cost-sensitive) to mitigate the dramatic effects of highly imbalanced data, without distorting the original data samples. The generated models are evaluated, validated, and consequently compared to different machine-learning models. A real-world telemarketing dataset from a Portuguese bank is used in all the experiments. The best prediction model achieved 79% of geometric mean, and misclassification errors were minimized to 0.192, 0.229 of Type I & Type II Errors, respectively. In summary, an interesting Meta-Cost method improved the performance of the prediction model without imposing significant processing overhead or altering original data samples.

Computational Modeling and Prediction on Viscosity of Slags by Big Data Mining

The viscosity of slag is a key factor affecting metallurgical efficiency and recycling, such as metal-slag reaction and separation, as well as slag wool processing. In order to comprehensively clarify the variation of the slag viscosity, various data mining methods have been employed to predict the viscosity of the slag. In this study, a more advanced dual-stage predictive modeling approach is proposed in order to accurately analyze and predict the viscosity of slag. Compared with the traditional single data mining approach, the proposed method performs better with a higher recall rate and low misclassification rate. The simulation results show that temperature, SiO2, Al2O3, P2O5, and CaO have greater influences on the slag’s viscosity. The critical temperature for onset of the important influence of slag composition is 980 °C. Furthermore, it is found that SiO2 and P2O5 have positive correlations with slag’s viscosity, while temperature, Al2O3, and CaO have negative correlations. A two-equation model of six-degree polynomial combined with Arrhenius formula is also established for the purpose of providing theoretical guidance for industrial application and reutilization of slag.

Boosting Minority Class Prediction on Imbalanced Point Cloud Data

Data imbalance during the training of deep networks can cause the network to skip directly to learning minority classes. This paper presents a novel framework by which to train segmentation networks using imbalanced point cloud data. PointNet, an early deep network used for the segmentation of point cloud data, proved effective in the point-wise classification of balanced data; however, performance degraded when imbalanced data was used. The proposed approach involves removing between-class data point imbalances and guiding the network to pay more attention to majority classes. Data imbalance is alleviated using a hybrid-sampling method involving oversampling, as well as undersampling, respectively, to decrease the amount of data in majority classes and increase the amount of data in minority classes. A balanced focus loss function is also used to emphasize the minority classes through the automated assignment of costs to the various classes based on their density in the point cloud. Experiments demonstrate the effectiveness of the proposed training framework when provided a point cloud dataset pertaining to six objects. The mean intersection over union (mIoU) test accuracy results obtained using PointNet training were as follows: XYZRGB data (91%) and XYZ data (86%). The mIoU test accuracy results obtained using the proposed scheme were as follows: XYZRGB data (98%) and XYZ data (93%).

A descriptive study of variable discretization and cost-sensitive logistic regression on imbalanced credit data

Training classification models on imbalanced data tends to result in bias towards the majority class. In this paper, we demonstrate how variable discretization and cost-sensitive logistic regression help mitigate this bias on an imbalanced credit scoring dataset, and further show the application of the variable discretization technique on the data from other domains, demonstrating its potential as a generic technique for classifying imbalanced data beyond credit socring. The performance measurements include ROC curves, Area under ROC Curve (AUC), Type I Error, Type II Error, accuracy, and F1 score. The results show that proper variable discretization and cost-sensitive logistic regression with the best class weights can reduce the model bias and/or variance. From the perspective of the algorithm, cost-sensitive logistic regression is beneficial for increasing the value of predictors even if they are not in their optimized forms while maintaining monotonicity. From the perspective of predictors, the variable discretization performs better than cost-sensitive logistic regression, provides more reasonable coefficient estimates for predictors which have nonlinear relationships against their empirical logit, and is robust to penalty weights on misclassifications of events and non-events determined by their apriori proportions.

Intelligent Fault Diagnosis of HVCB with Feature Space Optimization-Based Random Forest

Mechanical faults of high-voltage circuit breakers (HVCBs) always happen over long-term operation, so extracting the fault features and identifying the fault type have become a key issue for ensuring the security and reliability of power supply. Based on wavelet packet decomposition technology and random forest algorithm, an effective identification system was developed in this paper. First, compared with the incomplete description of Shannon entropy, the wavelet packet time-frequency energy rate (WTFER) was adopted as the input vector for the classifier model in the feature selection procedure. Then, a random forest classifier was used to diagnose the HVCB fault, assess the importance of the feature variable and optimize the feature space. Finally, the approach was verified based on actual HVCB vibration signals by considering six typical fault classes. The comparative experiment results show that the classification accuracy of the proposed method with the origin feature space reached 93.33% and reached up to 95.56% with optimized input feature vector of classifier. This indicates that feature optimization procedure is successful, and the proposed diagnosis algorithm has higher efficiency and robustness than traditional methods.

Robust total retina thickness segmentation in optical coherence tomography images using convolutional neural networks

We developed a fully automated system using a convolutional neural network (CNN) for total retina segmentation in optical coherence tomography (OCT) that is robust to the presence of severe retinal pathology. A generalized U-net network architecture was introduced to include the large context needed to account for large retinal changes. The proposed algorithm outperformed qualitative and quantitatively two available algorithms. The algorithm accurately estimated macular thickness with an error of 14.0 ± 22.1 µm, substantially lower than the error obtained using the other algorithms (42.9 ± 116.0 µm and 27.1 ± 69.3 µm, respectively). These results highlighted the proposed algorithm's capability of modeling the wide variability in retinal appearance and obtained a robust and reliable retina segmentation even in severe pathological cases.