True Positive Predictions Research Articles

A Digital Tool Supporting Pathology Practice and Identifying Leucocytes.

The aim of this work is to deliver an all-in-one package that contains both the part where the pathologist can manipulate the data as well as predefined models, altogether with the digital pathology interface with a comprehensive component that provides traceability between the identified leucocytes and the underlying possible outcomes of the potential disease. The aim is to directly provide the number of leucocytes and the mass of the cell, only from the image, with minimal intervention from the pathologist, necessary to have a PoC (Proof of Concept) or a prototype. The model was trained on a dataset of around 20,000 models, and the achieved accuracy was approximately 85%. Approximately 82% of the identified areas of interest, as determined by the models, were true positive predictions. The models correctly identified approximately 89% of the actual positive instances - areas of interest - identified by the pathologist. Approximately 6% of the total actual negative instances were incorrectly classified as positive by the models. The tool provides visual scripting, reducing the learning curve for pathology analysis techniques and offers an intuitive interface for healthcare professionals.

SimClone: Detecting Tabular Data Clones using Value Similarity

Data clones are defined as multiple copies of the same data among datasets. Presence of data clones between datasets can cause issues such as difficulties in managing data assets and data license violations when using datasets with clones to build AI software. However, detecting data clones is not trivial. Majority of the prior studies in this area rely on structural information to detect data clones (e.g., font size, column header). However, tabular datasets used to build AI software are typically stored without any structural information. In this paper, we propose a novel method called SimClone for data clone detection in tabular datasets without relying on structural information. SimClone method utilizes value similarities for data clone detection. We also propose a visualization approach as a part of our SimClone method to help locate the exact position of the cloned data between a dataset pair. Our results show that our SimClone outperforms the current state-of-the-art method by at least 20% in terms of both F1-score and AUC. In addition, SimClone’s visualization component helps identify the exact location of the data clone in a dataset with a Precision@10 value of 0.80 in the top 20 true positive predictions.

Strengthening Digital Security: Dynamic Attack Detection with LSTM, KNN, and Random Forest

Digital security is an ever-escalating concern in today's interconnected world, necessitating advanced intrusion detection systems. This research focuses on fortifying digital security through the integration of Long Short-Term Memory (LSTM), K-Nearest Neighbors (KNN), and Random Forest for dynamic attack detection. Leveraging a robust dataset, the models were subjected to rigorous evaluation, considering metrics such as accuracy, precision, recall, F1-score, and AUC-ROC. The LSTM model exhibited exceptional proficiency in capturing intricate sequential dependencies within network traffic, attaining a commendable accuracy of 99.11%. KNN, with its non-parametric adaptability, demonstrated resilience with a high accuracy of 99.23%. However, the Random Forest model emerged as the standout performer, boasting an accuracy of 99.63% and showcasing exceptional precision, recall, and F1-score metrics. Comparative analyses unveiled nuanced differences, guiding the selection of models based on specific security requirements. The AUC-ROC comparison reinforced the discriminative power of the models, with Random Forest consistently excelling. While all models excelled in true positive predictions, detailed scrutiny of confusion matrices offered insights into areas for refinement. In conclusion, the integration of LSTM, KNN, and Random Forest presents a robust and adaptive approach to dynamic attack detection. This research contributes valuable insights to the evolving landscape of digital security, emphasizing the significance of leveraging advanced machine learning techniques in constructing resilient defenses against cyber adversaries. The findings underscore the need for adaptive security solutions as the cyber threat landscape continues to evolve, with implications for practitioners, researchers, and policymakers in the field of cybersecurity.

Research of data mining methods for classification of imbalanced data sets

With the rapid development of information technology, which is widely used in all spheres of human life and activity, extremely large amounts of data have been accumulated today. By applying machine learning methods to this data, new practically useful knowledge can be obtained. The main goal of this paper is to study different machine learning methods for solving the classification problem and compare their efficiency and accuracy. A separate task is data pre-processing aimed at solving the problem of sample imbalance, as well as identifying the principal components that will be used to solve the classification problem. For this purpose, an information system for classifying the bankruptcy of a company with specified economic and financial characteristics was researched and developed. The study uses a dataset on the basis of which the efficiency and quality of application of several existing classification algorithms are evaluated. These classifiers are: conventional and linear Support Vector Machine, Extra Trees, Random Forest, Decision Tree, Logistic Regression, Multilayer perceptron Classifier, Gradient Boosting, Naive Bayes Classifier. For data pre-processing, we scaled the data, used the SMOTE method to get rid of the imbalance of the training sample, and performed principal component analysis and L1 regularisation. Principal component analysis allowed us to identify 15 principal components that have the greatest impact on classification accuracy and, accordingly, use them in the classification process. Analysing the results, we found that the best classifier was Random Forest with 95.9 % accuracy, and the worst was Naive Bayes with 85.1 %. To evaluate the quality of classification and select the best classifier, the Confusion matrix is used, which takes into account the number of true positive (TP) and true negative (TN) values, as well as the number of false negative (FN) and false positive (FP) classification results, and the values of such metrics as accuracy, precision, sensitivity, F1, and ROC. Accuracy is the percentage of correct answers given by the algorithm, while Recall is the number of TPs divided by the number of TPs plus the number of FNs. F1 indicates the balance between accuracy and sensitivity. Precision is the number of true positive predictions divided by the number of false positive and true negative predictions. ROC AUC is a tool for measuring performance for classification tasks at different thresholds. It shows how well a model can distinguish between classes. The conclusions present the main results of the study and indicate the main future direction of the work, namely, the study of classification results for other datasets and more efficient processing and analysis.

Towards adequate policy enhancement: An AI-driven decision tree model for efficient recognition and classification of EPA status via multi-emission parameters

Accurate and timely evaluation and assessment of emission data and its impact on environmental status has been a key challenge due to the conventional manual approach utilized for independently computing most emission parameters. To resolve this long-standing issue, we proposed an Artificial Intelligence (AI)-driven Decision Tree model to adequately classify Environmental Protection Agency (EPA) status based on multiple Emission Parameters. The model's performance was systematically evaluated using multiple emission parameters obtained from a two-stroke motorcycle dataset collected in Nigeria across various metrics such as K-S Statistics, Confusion Matrix, Correlation Heat Map, Decision Tree, Validation Curve, and Threshold Plot. The K-S Statistics plot's experimental results showed a considerable correlation between HC, CO, and the target variable, with values ranging from 0.75 to 0.80. At the same time, CO2 and O2 do not correlate with the target variable with values between 0.00 and 0.09. The Confusion Matrix revealed that the proposed model has an overall accuracy of 99.9% with 481 true positive predictions and 75 true negative predictions, indicating the effectiveness of the proposed AI-driven model. In conclusion, our proposed AI-driven model can effectively classify EPA status based on multiple emission parameters with high accuracy, which may spur positive advancement in policy enhancement for proper environmental management.

Robust Detection of Chronic Lymphocytic Leukemia with Support Vector Machines and Flow Cytometry

Abstract
 Aim: Our aim is to build a precise automatic tool for the diagnosis of CLL with the help of machine learning algorithms and flow cytometry immunophenotypic data.
 Methods: We run experiments with two machine learning methods. First one is decision tree which was previously used in other similar works and second one is support vector machines which is considered to be a more robust classification method.
 Results : Among the 40 CLL patients from the test set, the model correctly predicts 38 of them and among the 20 other B-CLPD patients, the model predicts 18 of them correctly. Its sensitivity, which is the fraction of true positive predictions among all positive samples, is 95% (38/40).
 Conclusion : The model achieves very high accuracies on our leave out test set. This model can be a useful tool for automatic CLL diagnosis.

Predicting the recurrence risk of renal cell carcinoma after nephrectomy: potential role of CT-radiomics for adjuvant treatment decisions.

Previous trial results suggest that only a small number of patients with non-metastatic renal cell carcinoma (RCC) benefit from adjuvant therapy. We assessed whether the addition of CT-based radiomics to established clinico-pathological biomarkers improves recurrence risk prediction for adjuvant treatment decisions. This retrospective study included 453 patients with non-metastatic RCC undergoing nephrectomy. Cox models were trained to predict disease-free survival (DFS) using post-operative biomarkers (age, stage, tumor size and grade) with and without radiomics selected on pre-operative CT. Models were assessed using C-statistic, calibration, and decision curve analyses (repeated tenfold cross-validation). At multivariable analysis, one of four selected radiomic features (wavelet-HHL_glcm_ClusterShade) was prognostic for DFS with an adjusted hazard ratio (HR) of 0.44 (p = 0.02), along with American Joint Committee on Cancer (AJCC) stage group (III versus I, HR 2.90; p = 0.002), grade 4 (versus grade 1, HR 8.90; p = 0.001), age (per 10 years HR 1.29; p = 0.03), and tumor size (per cm HR 1.13; p = 0.003). The discriminatory ability of the combined clinical-radiomic model (C = 0.80) was superior to that of the clinical model (C = 0.78; p < 0.001). Decision curve analysis revealed a net benefit of the combined model when used for adjuvant treatment decisions. At an exemplary threshold probability of ≥25% for disease recurrence within 5 years, using the combined versus the clinical model was equivalent to treating 9 additional patients (per 1000 assessed) who would recur without treatment (i.e., true-positive predictions) with no increase in false-positive predictions. Adding CT-based radiomic features to established prognostic biomarkers improved post-operative recurrence risk assessment in our internal validation study and may help guide decisions regarding adjuvant therapy. In patients with non-metastatic renal cell carcinoma undergoing nephrectomy, CT-based radiomics combined with established clinical and pathological biomarkers improved recurrence risk assessment. Compared to a clinical base model, the combined risk model enabled superior clinical utility if used to guide decisions on adjuvant treatment.

Influence of uncertainty estimation techniques on false-positive reduction in liver lesion detection

Deep learning techniques show success in detecting objects in medical images, but still suffer from false-positive predictions that may hinder accurate diagnosis. The estimated uncertainty of the neural network output has been used to flag incorrect predictions. We study the role played by features computed from neural network uncertainty estimates and shape-based features computed from binary predictions in reducing false positives in liver lesion detection by developing a classification-based post-processing step for different uncertainty estimation methods. We demonstrate an improvement in the lesion detection performance of the neural network (with respect to F1-score) for all uncertainty estimation methods on two datasets, comprising abdominal MR and CT images respectively. We show that features computed from neural network uncertainty estimates tend not to contribute much toward reducing false positives. Our results show that factors like class imbalance (true over false positive ratio) and shape-based features extracted from uncertainty maps play an important role in distinguishing false positive from true positive predictions.

Novel Techniques to Assess Predictive Systems and Reduce Their Alarm Burden.

Machine prediction algorithms (e.g., binary classifiers) often are adopted on the basis of claimed performance using classic metrics such as precision and recall. However, classifier performance depends heavily upon the context (workflow) in which the classifier operates. Classic metrics do not reflect the realized performance of a predictor unless certain implicit assumptions are met, and these assumptions cannot be met in many common clinical scenarios. This often results in suboptimal implementations and in disappointment when expected outcomes are not achieved. One common failure mode for classic metrics arises when multiple predictions can be made for the same event, particularly when redundant true positive predictions produce little additional value. This describes many clinical alerting systems. We explain why classic metrics cannot correctly represent predictor performance in such contexts, and introduce an improved performance assessment technique using utility functions to score predictions based on their utility in a specific workflow context. The resulting utility metrics (u-metrics) explicitly account for the effects of temporal relationships and other sources of variability in prediction utility. Compared to traditional measures, u-metrics more accurately reflect the real-world costs and benefits of a predictor operating in a realized context. The improvement can be significant. We also describe a formal approach to snoozing, a mitigation strategy in which some predictions are suppressed to improve predictor performance by reducing false positives while retaining event capture. Snoozing is especially useful for predictors that generate interruptive alarms. U-metrics correctly measure and predict the performance benefits of snoozing, whereas traditional metrics do not.

A deep learning approach to median nerve evaluation in ultrasound images of carpal tunnel inlet

Ultrasound (US) imaging is recognized as a useful support for Carpal Tunnel Syndrome (CTS) assessment through the evaluation of median nerve morphology. However, US is still far to be systematically adopted to evaluate this common entrapment neuropathy, due to US intrinsic challenges, such as its operator dependency and the lack of standard protocols. To support sonographers, the present study proposes a fully-automatic deep learning approach to median nerve segmentation from US images. We collected and annotated a dataset of 246 images acquired in clinical practice involving 103 rheumatic patients, regardless of anatomical variants (bifid nerve, closed vessels). We developed a Mask R-CNN with two additional transposed layers at segmentation head to accurately segment the median nerve directly on transverse US images. We calculated the cross-sectional area (CSA) of the predicted median nerve. Proposed model achieved good performances both in median nerve detection and segmentation: Precision (Prec), Recall (Rec), Mean Average Precision (mAP) and Dice Similarity Coefficient (DSC) values are 0.916 ± 0.245, 0.938 ± 0.233, 0.936 ± 0.235 and 0.868 ± 0.201, respectively. The CSA values measured on true positive predictions were comparable with the sonographer manual measurements with a mean absolute error (MAE) of 0.918 mm2. Experimental results showed the potential of proposed model, which identified and segmented the median nerve section in normal anatomy images, while still struggling when dealing with infrequent anatomical variants. Future research will expand the dataset including a wider spectrum of normal anatomy and pathology to support sonographers in daily practice.Graphical abstract

Network Architecture Influence on Facial Emotion Recognition

Abstract Artificial Intelligence has been blending into daily life by means of many useful applications from voice command to facial recognition. One therapeutic application to be supported by AI solutions is treatment of people with Autism Spectrum Disorder. A closed loop feedback system is planned in conjunction with a novel reward system that will encourage the user to express emotions and be rewarded for it in a virtual environment. In this work five popular neural network architectures of VGG16, ResNet50, GoogleNet, ShuffleNet and EfficientNetb0 are studied and compared, with the aim of finding a relation between accuracy and developed features based on the architecture, for the application in Facial Emotion Recognition (FER). Three datasets were used, the OULU-CASIA for training and validation, alongside FACES and JAFFE for robustness analysis. The images were first preprocessed to eliminate background noise. The performance of the model was based on the true positive predictions with Grad-CAM prediction visualizations to visualize the focus of the networks in making decisions for classification. Results showed that deep network architectures with high parameter space performed best, with architecture design showing more influence on the region of focus than on classification results. This is attributed to the different layer combinations as well as parameters used for feature extraction. Shallow depth networks with high parameter space performed better than deep networks with low parameter space for FER application.

Verification of subclinical carotid atherosclerosis as part of risk stratification in overweight and obesity: the role of machine learning in the development of a diagnostic algorithm

Aim. Comparative analysis of mathematical models obtained using multivariate logistic regression (MLR) with stepwise inclusion of predictors and machine learning (ML) for assessing the probability of subclinical carotid atherosclerosis in normotensive overweight and obese patients without cardiovascular diseases and/or diabetes.Material and methods. We received data on patients from the Webiomed platform database. The inclusion criteria were age ≥18 years, body mass index ≥25 kg/m2, extracranial artery ultrasound results, while the exclusion criteria included diabetes and/or cardiovascular disease. MLR analysis was carried out with stepwise inclusion of predictors. ML algorithms were used to create an alternative model.Results. The overall percentage of true results for MLR model was 73,2%, while the proportion of true negative and positive predictions was 80,1% and 63,4%, respectively. Mathematical models created using ML methods are characterized by a predictive value from 75 to 97% with a sensitivity of 77 to 92% and a specificity of 80 to 98%.Conclusion. A significant superiority of ML models was revealed in the study of available clinical and paraclinical parameters. Integration of ML mathematical models into a diagnostic algorithm for making a decision to refer a low-risk patient for extracranial artery ultrasound will significantly improve its accuracy and cost efficiency.

Using Support Vector Machine (SVM) with GPS Ionospheric TEC Estimations to Potentially Predict Earthquake Events

There are significant controversies surrounding the detection of precursors that may precede earthquakes. Natural hazard signatures associated with strong earthquakes can appear in the lithosphere, troposphere, and ionosphere, where current remote sensing technologies have become valuable tools for detecting and measuring early warning signals of stress build-up deep in the Earth’s crust (presumably associated with earthquake events). Here, we propose implementing a machine learning support vector machine (SVM) technique, applied with GPS ionospheric total electron content (TEC) pre-processed time series estimations, to evaluate potential precursors caused by earthquakes and manifested as disturbances in the TEC data. After filtering and screening our data for solar or geomagnetic influences at different time scales, our results indicate that for large earthquakes (&gt;Mw 6), true negative predictions can be achieved with 85.7% accuracy, and true positive predictions with an accuracy of 80%. We tested our method with different skill scores, such as accuracy (0.83), precision (0.85), recall (0.8), the Heidke skill score (0.66), and true skill statistics (0.66).

Precursors-driven machine learning prediction of chaotic extreme pulses in Kerr resonators

Machine learning algorithms have opened a breach in the prediction's fortress of high-dimensional chaotic systems. Their ability to find hidden correlations in data can be exploited to perform model-free forecasting of spatiotemporal chaos and extreme events. However, the extensive feature of these evolutions makes up a critical limitation for full-size forecasting processes. Hence, the main challenge for forecasting relevant events is to establish the set of pertinent information. Here, we identify precursors from the transfer entropy of the system and a deep Long Short-Term Memory network to forecast the complex dynamics of a system evolving in a high-dimensional spatiotemporal chaotic regime. Performances of this triggerable model-free prediction protocol based on the information flowing map are tested from experimental data originating from a passive resonator operating in such a complex nonlinear regime. We have been able to predict the occurrence of extreme events up to 9 round trips after the detection of precursor, i.e., 3 times the horizon provided by Lyapunov exponents, with 92% of true positive predictions leading to 60% of accuracy.We have implemented a process to forecast extreme events in a fully developed turbulent flow. The novelty of our strategy lies in the information's use theory method to detect precursors and use them as the input of a neural network to infer the incoming extreme events. Our process is suitable for all the extended dissipative systems that can only be partially observed or real-world data.

Seizure Prediction in Genetic Rat Models of Absence Epilepsy: Improved Performance through Multiple-Site Cortico-Thalamic Recordings Combined with Machine Learning.

Seizure prediction is the grand challenge of epileptology. However, effort was devoted to prediction of focal seizures, while generalized seizures were regarded as stochastic events. Long-lasting local field potential (LFP) recordings containing several hundred generalized spike and wave discharges (SWDs), acquired at eight locations in the cortico-thalamic system of absence epileptic rats, were iteratively analyzed in all possible combinations of either two or three recording sites, by a wavelet-based algorithm, calculating the product of the wavelet-energy signaling increases in synchronicity. Sensitivity and false alarm rate of prediction were compared between various combinations, and wavelet spectra of true and false positive predictions were fed to a random forest machine learning algorithm to further differentiate between them. Wavelet analysis of intracortical and cortico-thalamic LFP traces showed a significantly smaller number of false alarms compared with intrathalamic combinations, while predictions based on recordings in Layers IV, V, and VI of the somatosensory-cortex significantly outreached all other combinations in terms of prediction sensitivity. In 24-h out-of-sample recordings of nine Genetic Absence Epilepsy Rats from Strasbourg (GAERS), containing diurnal fluctuations of SWD occurrence, classification of true and false positives by the trained random forest further reduced the false alarm rate by 71%, although at some trade-off between false alarms and sensitivity of prediction, as reflected in relatively low F1 score values. Results provide support for the cortical-focus theory of absence epilepsy and allow the conclusion that SWDs are predictable to some degree. The latter paves the way for the development of closed-loop SWD prediction-prevention systems. Suggestions for a possible translation to human data are outlined.

DNCON2_Inter: predicting interchain contacts for homodimeric and homomultimeric protein complexes using multiple sequence alignments of monomers and deep learning

Deep learning methods that achieved great success in predicting intrachain residue-residue contacts have been applied to predict interchain contacts between proteins. However, these methods require multiple sequence alignments (MSAs) of a pair of interacting proteins (dimers) as input, which are often difficult to obtain because there are not many known protein complexes available to generate MSAs of sufficient depth for a pair of proteins. In recognizing that multiple sequence alignments of a monomer that forms homomultimers contain the co-evolutionary signals of both intrachain and interchain residue pairs in contact, we applied DNCON2 (a deep learning-based protein intrachain residue-residue contact predictor) to predict both intrachain and interchain contacts for homomultimers using multiple sequence alignment (MSA) and other co-evolutionary features of a single monomer followed by discrimination of interchain and intrachain contacts according to the tertiary structure of the monomer. We name this tool DNCON2_Inter. Allowing true-positive predictions within two residue shifts, the best average precision was obtained for the Top-L/10 predictions of 22.9% for homodimers and 17.0% for higher-order homomultimers. In some instances, especially where interchain contact densities are high, DNCON2_Inter predicted interchain contacts with 100% precision. We also developed Con_Complex, a complex structure reconstruction tool that uses predicted contacts to produce the structure of the complex. Using Con_Complex, we show that the predicted contacts can be used to accurately construct the structure of some complexes. Our experiment demonstrates that monomeric multiple sequence alignments can be used with deep learning to predict interchain contacts of homomeric proteins.

Sequence Neighborhoods Enable Reliable Prediction of Pathogenic Mutations in Cancer Genomes.

Simple SummaryCancer is caused by the accumulation of somatic mutations, some of which are responsible for the disease’s progression (drivers) while others are functionally neutral (passengers). Although several methods have been developed to distinguish between the two classes of mutations, very few have concentrated on using the neighborhood nucleotide sequences as potential discrimination features. In this study, we show that driver mutations’ neighborhood is significantly different from that of passengers. We further develop a novel machine learning tool, NBDriver, which is highly efficient at identifying pathogenic variants from multiple independent test datasets. Efficient and accurate identification of novel pathogenic variants from sequenced cancer genomes would help facilitate more effective therapies tailored to patients’ mutational profiles.Identifying cancer-causing mutations from sequenced cancer genomes hold much promise for targeted therapy and precision medicine. “Driver” mutations are primarily responsible for cancer progression, while “passengers” are functionally neutral. Although several computational approaches have been developed for distinguishing between driver and passenger mutations, very few have concentrated on using the raw nucleotide sequences surrounding a particular mutation as potential features for building predictive models. Using experimentally validated cancer mutation data in this study, we explored various string-based feature representation techniques to incorporate information on the neighborhood bases immediately 5′ and 3′ from each mutated position. Density estimation methods showed significant distributional differences between the neighborhood bases surrounding driver and passenger mutations. Binary classification models derived using repeated cross-validation experiments provided comparable performances across all window sizes. Integrating sequence features derived from raw nucleotide sequences with other genomic, structural, and evolutionary features resulted in the development of a pan-cancer mutation effect prediction tool, NBDriver, which was highly efficient in identifying pathogenic variants from five independent validation datasets. An ensemble predictor obtained by combining the predictions from NBDriver with three other commonly used driver prediction tools (FATHMM (cancer), CONDEL, and MutationTaster) significantly outperformed existing pan-cancer models in prioritizing a literature-curated list of driver and passenger mutations. Using the list of true positive mutation predictions derived from NBDriver, we identified a list of 138 known driver genes with functional evidence from various sources. Overall, our study underscores the efficacy of using raw nucleotide sequences as features to distinguish between driver and passenger mutations from sequenced cancer genomes.

Convolutional neural network and transfer learning for dose volume histogram prediction for prostate cancer radiotherapy

To adopt a transfer learning approach and establish a convolutional neural network (CNN) model for the prediction of rectum and bladder dose-volume histograms (DVH) in prostate patients treated with a VMAT technique. One hundred forty-four VMAT patients with intermediate or high-risk prostate cancer were included in this study. Data were split into two sets: 120 and 24 patients, respectively. The second set was used for final validation. To ensure the accuracy of the training data, we developed a ground-truth analysis for detecting and correcting for all potential outliers. We used transfer learning in combination with a pre-trained VGG-16 network. We dropped the fully connected layers from the VGG-16 and added a new fully connected neural network. The inputs for the CNN were a 2D image of the volumes contoured in the CT, but we only retained the geometrical information of every CT-slice. The outputs were the corresponding rectum and bladder DVH for every slice. We used a confusion matrix to analyze the performance of our model. Our model achieved 100% and 81% of true positive and true negative predictions, respectively. We have an overall accuracy of 87.5%, a misclassification rate of 12.5%, and a precision of 100%. We have successfully developed a model for reliable prediction of rectum and bladder DVH in prostate patients by applying a previously pre-trained CNN. To our knowledge, this is the first attempt to apply transfer learning to the prediction of DVHs that accounts for the ground truth problem.

Mother’s Lifestyle Feature Relevance for NICU and Preterm Birth Prediction

Maternal health plays an important role in defining the health of mother, child and childbirth experience. With the change in lifestyle over the decades, there have been many health challenges faced by woman, which makes it important for women to understand the impact of their lifestyle and physical health features on their wellbeing. In this study, we have realised the importance of mother’s features with respect to preterm childbirth prediction and prediction for neonatal intensive care unit(NICU) facility requirement for newborn. Experiments are performed on MSF dataset which consists of records of 1000 women, 21 physical features and 78 lifestyle features are taken into consideration. Random forest based hybrid model using F-score and Mutual information is used to evaluate each features for their capability of True positive(TP) and False Negative(FN) predictions. For preterm birth prediction, out of all the features hypertension, diabetes, PCOS and consumption of outside food during teenage are found to be the most relevant features. While for NICU prediction diabetes, low amniotic fluid during pregnancy, exposure to air and noise pollution during teenage and consumption of alcohol after marriage are found to be relevant.

Predicting lymph node metastasis in patients with oropharyngeal cancer by using a convolutional neural network with associated epistemic and aleatoric uncertainty

There can be significant uncertainty when identifying cervical lymph node (LN) metastases in patients with oropharyngeal squamous cell carcinoma (OPSCC) despite the use of modern imaging modalities such as positron emission tomography (PET) and computed tomography (CT) scans. Grossly involved LNs are readily identifiable during routine imaging, but smaller and less PET-avid LNs are harder to classify. We trained a convolutional neural network (CNN) to detect malignant LNs in patients with OPSCC and used quantitative measures of uncertainty to identify the most reliable predictions. Our dataset consisted of images of 791 LNs from 129 patients with OPSCC who had preoperative PET/CT imaging and detailed pathological reports after neck dissections. These LNs were segmented on PET/CT imaging and then labeled according to the pathology reports. An AlexNet-like CNN was trained to classify LNs as malignant or benign. We estimated epistemic and aleatoric uncertainty by using dropout variational inference and test-time augmentation, respectively. CNN performance was stratified according to the median epistemic and aleatoric uncertainty values calculated using the validation cohort. Our model achieved an area under the receiver operating characteristic (ROC) curve (AUC) of 0.99 on the testing dataset. Sensitivity and specificity were 0.94 and 0.90, respectively. Epistemic and aleatoric uncertainty values were statistically larger for false negative and false positive predictions than for true negative and true positive predictions (p < 0.001). Model sensitivity and specificity were 1.0 and 0.98, respectively, for cases with epistemic uncertainty lower than the median value of the incorrect predictions in the validation dataset. For cases with higher epistemic uncertainty, sensitivity and specificity were 0.67 and 0.41, respectively. Model sensitivity and specificity were 1.0 and 0.98, respectively, for cases with aleatoric uncertainty lower than the median value of the incorrect predictions in the validation dataset. For cases with higher aleatoric uncertainty, sensitivity and specificity were 0.67 and 0.37, respectively. We used a CNN to predict the malignant status of LNs in patients with OPSCC with high accuracy, and we showed that uncertainty can be used to quantify a prediction’s reliability. Assigning measures of uncertainty to predictions could improve the accuracy of LN classification by efficiently identifying instances where expert evaluation is needed to corroborate a model’s prediction.