Application of Machine Learning to Predict Cervical Lymph Node Metastasis from Micropapillary Thyroid Carcinoma with Ultrasound

Background This study aims to determine the prediction performance of a machine learning-based clinical model for cervical lymph node metastasis (CLNM) in micropapillary thyroid carcinoma (MPTC) with ultrasound (US).Methods Patients with MPTC who underwent total or hemithyroidectomy with unilateral or bilateral prophylactic central neck dissection were included (n = 692). Nodal status was pathologically determined. Clinical and US features and thyroid function markers were extracted to build a random forest model. A nomogram with the significant predictive risk factors from multivariable logistic regression analysis was built to visualize hazard rates. Finally, the predictive performances of the models were compared.Results Overall, 332 patients (47.98%) had CLNM. In multiple logistic regression, the strong predictive risk factors for CLNM were younger age, larger anteroposterior diameter, lower anteroposterior/transverse diameter (A/T) ratio, and higher thyroglobulin (TG) concentration (P < 0.05). The random forest and nomogram models showed good predictive performance with the area under the curves (AUCs) of 0.836 and 0.780, respectively, which were significantly higher than those without A/T ratio in the models (AUCs: 0.807 vs. 0.722, all P < 0.05). The AUC of the A/T ratio as a single feature for predicting CLNM was 0.744, while A/T ratio (≤ 0.828) combined with anteroposterior diameter (≥ 10 mm) yielded a higher AUC of 0.754 for predicting CLNM.Conclusions The machine learning-based clinical model with US had a good predictive performance for CLNM in MPTC patients. This clinical model may facilitate surgical decision-making for MPTC, especially regarding whether cervical lymph node dissection is warranted.

Peeking inside the Black Box: Interpreting Deep Learning Models for Exoplanet Atmospheric Retrievals

&lt;p&gt;Deep learning algorithms are growing in popularity in the field of exoplanetary science due to their ability to model highly non-linear relations and solve interesting problems in a data-driven manner. Several works have attempted to perform fast retrievals of atmospheric parameters with the use of machine learning algorithms like deep neural networks (DNNs). Yet, despite their high predictive power, DNNs are also infamous for being 'black boxes'. It is their apparent lack of explainability that makes the astrophysics community reluctant to adopt them. What are their predictions based on? How confident should we be in them? When are they wrong and how wrong can they be? In this work, we present a number of general evaluation methodologies that can be applied to any trained model and answer questions like these. In particular, we train three different popular DNN architectures to retrieve atmospheric parameters from exoplanet spectra and show that all three achieve good predictive performance. We then present an extensive analysis of the predictions of DNNs, which can inform us - among other things - of the credibility limits for atmospheric parameters for a given instrument and model. Finally, we perform a perturbation-based sensitivity analysis to identify to which features of the spectrum the outcome of the retrieval is most sensitive. We conclude that for different molecules, the wavelength ranges to which the DNN's predictions are most sensitive, indeed coincide with their characteristic absorption regions. The methodologies presented in this work help to improve the evaluation of DNNs and to grant interpretability to their predictions.&lt;/p&gt;

A Combined RNA Signature Predicts Recurrence Risk of Stage I-IIIA Lung Squamous Cell Carcinoma

ObjectiveRecurrence remains the main cause of the poor prognosis in stage I-IIIA lung squamous cell carcinoma (LUSC) after surgical resection. In the present study, we aimed to identify the long non-coding RNAs (lncRNAs), microRNAs (miRNAs), and messenger RNAs (mRNAs) related to the recurrence of stage I-IIIA LUSC. Moreover, we constructed a risk assessment model to predict the recurrence of LUSC patients.MethodsRNA sequencing data (including miRNAs, lncRNAs, and mRNAs) and relevant clinical information were obtained from The Cancer Genome Atlas (TCGA) database. The differentially expressed lncRNAs, miRNAs, and mRNAs were identified using the “DESeq2” package of the R language. Univariate Cox proportional hazards regression analysis and Kaplan-Meier curve were used to identify recurrence-related genes. Stepwise multivariate Cox regression analysis was carried out to establish a risk model for predicting recurrence in the training cohort. Moreover, Kaplan-Meier curves and receiver operating characteristic (ROC) curves were adopted to examine the predictive performance of the signature in the training cohort, validation cohort, and entire cohort.ResultsBased on the TCGA database, we analyzed the differentially expressed genes (DEGs) among 27 patients with recurrent stage I-IIIA LUSC and 134 patients with non-recurrent stage I-IIIA LUSC, and identified 431 lncRNAs, 36 miRNAs, and 746 mRNAs with different expression levels. Out of these DEGs, the optimal combination of DEGs was finally determined, and a nine-joint RNA molecular signature was constructed for clinical prediction of recurrence, including LINC02683, AC244517.5, LINC02418, LINC01322, AC011468.3, hsa-mir-6825, AC020637.1, AC027117.2, and SERPINB12. The ROC curve proved that the model had good predictive performance in predicting recurrence. The area under the curve (AUC) of the prognostic model for recurrence-free survival (RFS) was 0.989 at 3 years and 0.958 at 5 years (in the training set). The combined RNA signature also revealed good predictive performance in predicting the recurrence in the validation cohort and entire cohort.ConclusionsIn the present study, we constructed a nine-joint RNA molecular signature for recurrence prediction of stage I-IIIA LUSC. Collectively, our findings provided new and valuable clinical evidence for predicting the recurrence and targeted treatment of stage I-IIIA LUSC.

Study on the prognosis predictive model of COVID-19 patients based on CT radiomics

Making timely assessments of disease progression in patients with COVID-19 could help offer the best personalized treatment. The purpose of this study was to explore an effective model to predict the outcome of patients with COVID-19. We retrospectively included 188 patients (124 in the training set and 64 in the test set) diagnosed with COVID-19. Patients were divided into aggravation and improvement groups according to the disease progression. Three kinds of models were established, including the radiomics, clinical, and combined model. Receiver operating characteristic curves, decision curves, and Delong’s test were used to evaluate and compare the models. Our analysis showed that all the established prediction models had good predictive performance in predicting the progress and outcome of COVID-19.

Statistical Analysis of Nigeria’s Price Sector: An Econometric Approach

This study analyzed Nigeria’s price sector using a formulated model for the price sector of the Nigeria economy. A set of simultaneous equations were used to reflect the implicit gross domestic product deflators for each of the sectors of the Nigeria economy and was found to be over identified under the order condition for identification. The model was estimated by ordinary least square method and two stage least square methods. All the variables have expected signs and as indicated by the F –statistic, the overall performance of the entire regression is significant. The high measure of R2 and Ṝ2, in each case indicates that the explanatory variables included in the equation jointly account for the entire variation. The small RMSE also indicates that the equations have good fit. Durbin –Watson statistics shows that there is no positive first order autocorrelation. The small value of the Theil’s inequality indicates that the equation has good predictive performance. The researcher therefore recommends that government should employ the model so as to be able to monitor price of each of the sectors of the economy and put proper mechanism in place to control those sectors that affect the overall price sector of the economy.

Incremental predictive clustering trees for online semi-supervised multi-target regression

In many application settings, labeling data examples is a costly endeavor, while unlabeled examples are abundant and cheap to produce. Labeling examples can be particularly problematic in an online setting, where there can be arbitrarily many examples that arrive at high frequencies. It is also problematic when we need to predict complex values (e.g., multiple real values), a task that has started receiving considerable attention, but mostly in the batch setting. In this paper, we propose a method for online semi-supervised multi-target regression. It is based on incremental trees for multi-target regression and the predictive clustering framework. Furthermore, it utilizes unlabeled examples to improve its predictive performance as compared to using just the labeled examples. We compare the proposed iSOUP-PCT method with supervised tree methods, which do not use unlabeled examples, and to an oracle method, which uses unlabeled examples as though they were labeled. Additionally, we compare the proposed method to the available state-of-the-art methods. The method achieves good predictive performance on account of increased consumption of computational resources as compared to its supervised variant. The proposed method also beats the state-of-the-art in the case of very few labeled examples in terms of performance, while achieving comparable performance when the labeled examples are more common.

Peeking inside the Black Box: Interpreting Deep Learning Models for Exoplanet Atmospheric Retrievals

&lt;p&gt;Deep learning algorithms are growing in popularity in the field of exoplanetary science due to their ability to model highly non-linear relations and solve interesting problems in a data-driven manner. Several works have attempted to perform fast retrieval of atmospheric parameters with the use of machine learning algorithms or deep neural networks (DNNs). &amp;#160;Yet, despite their high predictive power, &amp;#160;DNNs are also infamous for being `black boxes&amp;#8217;. It is their apparent lack of explainability that makes the astrophysics community reluctant to adopt them. What are their predictions based on? How confident should we be in them? When are they wrong and how wrong can they be? In this work, we present a number of general evaluation methodologies that can be applied to any trained model and answer questions like these. &amp;#160;In particular, we train 3 different popular DNN architectures to retrieve atmospheric parameters from exoplanet spectra and show that all 3 achieve good predictive performance. We then present an extensive analysis of the predictions of DNNs, which can inform us &amp;#8212;among other things &amp;#8212; of the credibility limit for atmospheric parameters for a given instrument and model. Finally, we perform a sensitivity analysis to identify to which features of the spectrum the outcome of the retrieval is most sensitive. We conclude that for different molecules, the wavelength ranges to which the DNN&amp;#8217;s predictions are most sensitive, indeed coincide with their characteristic absorption regions. The methodologies presented in this work help to improve the evaluation of DNNs and to grant interpretability to their predictions.&lt;/p&gt;

GenNet framework: interpretable neural networks for phenotype prediction

Neural networks have been seldomly leveraged in population genomics due to the computational burden and challenge of interpretability. Here, we propose GenNet, a novel open-source deep learning framework for predicting phenotype from genotype. In this framework, public prior biological knowledge is used to construct interpretable and memory-efficient neural network architectures. These architectures obtain good predictive performance for multiple traits and complex diseases, opening the door for neural networks in population genomics.

Bayesian arrival model for Atlantic salmon smolt counts powered by environmental covariates and expert knowledge

Annual run size and timing of Atlantic salmon (Salmo salar) smolt migration was estimated using Bayesian model framework and data from 6 years of a video monitoring survey. The model has a modular structure. It separates subprocesses of departing, traveling, and observing, of which the first two together define the arrival distribution. The subprocesses utilize biological background and expert knowledge about the migratory behavior of smolts and about the probability to observe them from the video footage under varying environmental conditions. Daily mean temperature and discharge were used as environmental covariates. The model framework does not require assuming a simple distributional shape for the arrival dynamics and thus also allows for multimodal arrival distributions. Results indicate that 20%–43% of smolts passed the Utsjoki monitoring site unobserved during the years of study. Predictive studies were made to estimate daily run size in cases with missing counts either at the beginning or in the middle of the run, indicating good predictive performance.

Individualized medication of digoxin based on the serum drug concentration and blood biochemical indexes

Aim: The dose of digoxin is often difficult to be determined precisely. The aim of this study was to retrospectively investigate the effect of blood biochemical indexes on the serum concentration of digoxin. Materials & methods: We collected the data of hospitalized patients treated orally with digoxin in Nanjing Drum Tower Hospital (Nanjing, China) from 2016 to 2018. Descriptive statistics was used to analyze the patients’ comprehensive condition. Results: A total of 425 patients were included in the study. Through analysis, nine factors were included in the regression model of the serum concentration of digoxin, and this regression model showed good predictive performance (r2 = 0.83138; p < 0.001). Conclusion: The regression model for the prediction of serum concentration of digoxin has clinical significance, and can provide research basis for individualized medication of digoxin.