Pretreatment Contrast-Enhanced Computed Tomography Radiomics for Prediction of Pathological Regression Following Neoadjuvant Chemotherapy in Locally Advanced Gastric Cancer: A Preliminary Multicenter Study

BackgroundSensitivity to neoadjuvant chemotherapy in locally advanced gastric cancer patients varies; however, an effective predictive marker is currently lacking. We aimed to propose and validate a practical treatment efficacy prediction method based on contrast-enhanced computed tomography (CECT) radiomics.MethodData of l24 locally advanced gastric carcinoma patients who underwent neoadjuvant chemotherapy were acquired retrospectively between December 2012 and August 2020 from three different cancer centers. In total, 1216 radiomics features were initially extracted from each lesion’s pretreatment portal venous phase computed tomography image. Subsequently, a radiomics predictive model was constructed using machine learning software. Clinicopathological data and radiological parameters of the enrolled patients were collected and analyzed retrospectively. Univariate and multivariate logistic regression analyses were performed to screen for independent predictive indices. Finally, we developed an integrated model combining clinicopathological predictive parameters and radiomics features.ResultIn the training set, 10 (14.9%) patients achieved a good response (GR) after preoperative neoadjuvant chemotherapy (n = 77), whereas in the testing set, seven (17.5%) patients achieved a GR (n = 47). The radiomics predictive model showed competitive prediction efficacy in both the training and independent external validation sets. The areas under the curve (AUC) values were 0.827 (95% confidence interval [CI]: 0.609–1.000) and 0.854 (95% CI: 0.610–1.000), respectively. Similarly, when only the single hospital data were included as an independent external validation set (testing set 2), AUC values of the models were 0.827 (95% CI: 0.650–0.952) and 0.889 (95% CI: 0.663–1.000) in the training set and testing set 2, respectively.ConclusionOur study is the first to discover that CECT radiomics could provide powerful and consistent predictions of therapeutic sensitivity to neoadjuvant chemotherapy among gastric cancer patients across different hospitals.

A pilot study of machine-learning based automated planning for primary brain tumours

Purpose High-quality radiotherapy (RT) planning for children and young adults with primary brain tumours is essential to minimize the risk of late treatment effects. The feasibility of using automated machine-learning (ML) to aid RT planning in this population has not previously been studied. Methods and materials We developed a ML model that identifies learned relationships between image features and expected dose in a training set of 95 patients with a primary brain tumour treated with focal radiotherapy to a dose of 54 Gy in 30 fractions. This ML method was then used to create predicted dose distributions for 15 previously-treated brain tumour patients across two institutions, as a testing set. Dosimetry to target volumes and organs-at-risk (OARs) were compared between the clinically-delivered (human-generated) plans versus the ML plans. Results The ML method was able to create deliverable plans in all 15 patients in the testing set. All ML plans were generated within 30 min of initiating planning. Planning target volume coverage with 95% of the prescription dose was attained in all plans. OAR doses were similar across most structures evaluated; mean doses to brain and left temporal lobe were lower in ML plans than manual plans (mean difference to left temporal, – 2.3 Gy, p = 0.006; mean differences to brain, – 1.3 Gy, p = 0.017), whereas mean doses to right cochlea and lenses were higher in ML plans (+ 1.6–2.2 Gy, p < 0.05 for each). Conclusions Use of an automated ML method to aid RT planning for children and young adults with primary brain tumours is dosimetrically feasible and can be successfully used to create high-quality 54 Gy RT plans. Further evaluation after clinical implementation is planned.

Activity Cliffs As Protein-Related Phenomenon: Investigation Using Machine Learning Against Numerous Protein Kinases

Activity cliffs (ACs) are analogous compounds of significant affinity discrepancies against certain biotarget. We propose that the ACs phenomenon is protein-related and that the propensity of certain target to have ACs can be predicted by some intrinsic protein properties. We pursued this assumption by collecting the crystallographic structures of 84 protein kinases, each of which has numerous reported inhibitors (hundreds). Following data augmentation using synthetic minority oversampling technique (SMOTE), we attempted to correlate the presence/absence of ACs within the ligand pools of collected protein kinases with their corresponding protein properties using genetic algorithm (GA) coupled with variety of machine learners (MLs). Very good GA-ML models were achieved with accuracies of around 75% against external testing set. The models were further validated by Y-scrambling. Shapely additive explanations highlighted the significance of protein rotatable bonds, hydrophobic and acidic residues in relation to the presence of ACs. These results support the hypothesis that ACs are protein-related.

Group Penalized Logistic Regressions Predict Ovarian Cancer

Objectives: Ovarian cancer ranks first among gynecological cancers in terms of the mortality rate. Accurately diagnosing ovarian benign tumors and malignant tumors is of immense important. The goal of this paper is to combine group LASSO/SCAD/MCP penalized logistic regression with machine learning procedure to further improve the prediction accuracy to ovarian benign tumors and malignant tumors prediction problem. Methods: We combine group LASSO/SCAD/MCP penalty with logistic regression, and propose group LASSO/SCAD/MCP penalized logistic regression to predict the benign and malignant ovarian cancer. Firstly, we select 349 ovarian cancer patients data and divide them into two sets: one is the training set for learning, and the other is the testing set for checking, and then choose 46 explanatory variables and divide them into 11 different groups. Secondly, we apply the training set and group coordinate descent algorithm to obtain group LASSO/SCAD/MCP estimator, and apply the testing set to compute confusion matrix, accuracy, sensitivity and specificity. Finally, we compare the prediction performance for group LASSO/SCAD/MCP penalized logistic regression with that for artificial neural network (ANN) and support vector machine (SVM).Results: Group LASSO/SCAD/MCP/ penalized logistic regression selects 6/4/1 groups. The prediction accuracy and AUC for group MCP/SCAD/LASSO penalized logistic regression/SVM/ANN is 93.33%/85.71%/82.26%/74.29%/72.38% and 0.892/0.852/0.823/0.639/0.789, respectively.Conclusions: Group MCP/SCAD/LASSO penalized logistic regression performs than SVM and ANN in terms of prediction accuracy and AUC. In particular, group MCP penalized logistic regression predicts the best. Therefore, we suggest group MCP penalized logistic regression to predict ovarian tumors.

Score for the Risk and Overall Survival of Lung Metastasis in Patients First Diagnosed With Soft Tissue Sarcoma: A Novel Nomogram-Based Risk Assessment System

Background: Metastatic soft tissue sarcoma (STS) patients have a poor prognosis with a 3-year survival rate of 25%. About 30% of them present lung metastases (LM). This study aimed to construct 2 nomograms to predict the risk of LM and overall survival of STS patients with LM. Materials and Methods: The data of patients were derived from the Surveillance, Epidemiology, and End Results database during the period of 2010 to 2015. Logistic and Cox analysis was performed to determine the independent risk factors and prognostic factors of STS patients with LM, respectively. Afterward, 2 nomograms were, respectively, established based on these factors. The performance of the developed nomogram was evaluated with receiver operating characteristic curves, area under the curve (AUC) calibration curves, and decision curve analysis (DCA). Results: A total of 7643 patients with STS were included in this study. The independent predictors of LM in first-diagnosed STS patients were N stage, grade, histologic type, and tumor size. The independent prognostic factors for STS patients with LM were age, N stage, surgery, and chemotherapy. The AUCs of the diagnostic nomogram were 0.806 in the training set and 0.799 in the testing set. For the prognostic nomogram, the time-dependent AUC values of the training and testing set suggested a favorable performance and discrimination of the nomogram. The 1-, 2-, and 3-year AUC values were 0.698, 0.718, and 0.715 in the training set, and 0.669, 0.612, and 0717 in the testing set, respectively. Furthermore, for the 2 nomograms, calibration curves indicated satisfactory agreement between prediction and actual survival, and DCA indicated its clinical usefulness. Conclusion: In this study, grade, histology, N stage, and tumor size were identified as independent risk factors of LM in STS patients, age, chemotherapy surgery, and N stage were identified as independent prognostic factors of STS patients with LM, these developed nomograms may be an effective tool for accurately predicting the risk and prognosis of newly diagnosed patients with LM.

Group Penalized Logistic Regressions Predict Ovarian Cancer ∗

Objectives: Ovarian cancer ranks fifirst among gynecological cancers in terms of the mortality rate. Accurately diagnosing ovarian benign tumors and malignant tumors is of immense important. The goal of this paper is to combine group LASSO/SCAD/MCP penalized logistic regression with machine learning procedure to further improve the prediction accuracy to ovarian benign tumors and malignant tumors prediction problem. Methods: We combine group LASSO/SCAD/MCP penalty with logistic regression, and propose group LASSO/SCAD/MCP penalized logistic regression to predict the benign and malignant ovarian cancer. Firstly, we select 349 ovarian cancer patients data and divide them into two sets: one is the training set for learning, and the other is the testing set for checking, and then choose 46 explanatory variables and divide them into 11 difffferent groups. Secondly, we apply the training set and group coordinate descent algorithm to obtain group LASSO/SCAD/MCP estimator, and apply the testing set to compute confusion matrix, accuracy, sensitivity and specifificity. Finally, we compare the prediction performance for group LASSO/SCAD/MCP penalized logistic regression with that for artifificial neural network (ANN) and support vector machine (SVM). Results: Group LASSO/SCAD/MCP/ penalized logistic regression selects 6/4/1 groups. The prediction accuracy and AUC for group MCP/SCAD/LASSO penalized logistic regression/SVM/ANN is 93.33%/85.71%/82.26%/74.29%/72.38% and 0.892/0.852/0.823/0.639/0.789, respectively. Conclusions: Group MCP/SCAD/LASSO penalized logistic regression performs than SVM and ANN in terms of prediction accuracy and AUC. In particular, group MCP penalized logistic regression predicts the best. Therefore, we suggest group MCP penalized logistic regression to predict ovarian tumors.

Hypoxic Characteristic Genes Predict Response to Immunotherapy for Urothelial Carcinoma

Objective: Resistance to immune checkpoint inhibitors (ICIs) has been a massive obstacle to ICI treatment in metastatic urothelial carcinoma (MUC). Recently, increasing evidence indicates the clinical importance of the association between hypoxia and immune status in tumor patients. Therefore, it is necessary to investigate the relationship between hypoxia and prognosis in metastatic urothelial carcinoma.Methods: Transcriptomic and clinical data from 348 MUC patients who underwent ICI treatment from a large phase 2 trial (IMvigor210) were investigated in this study. The cohort was randomly divided into two datasets, a training set (n = 213) and a testing set (n = 135). Data of hypoxia-related genes were downloaded from the molecular signatures database (MSigDB), and screened by univariate and multivariate Cox regression analysis to construct a prognosis-predictive model. The robustness of the model was evaluated in two melanoma cohorts. Furthermore, an external validation cohort, the bladder cancer cohort, from the Cancer Genome Atlas (TCGA) database, was t used to explore the mechanism of gene mutation, immune cell infiltration, signaling pathway enrichment, and drug sensitivity.Results: We categorized patients as the high- or low- risk group using a four-gene hypoxia risk model which we constructed. It was found that patients with high-risk scores had significantly worse overall survival (OS) compared with those with low-risk scores. The prognostic model covers 0.71 of the area under the ROC curve in the training set and 0.59 in the testing set, which is better than the survival prediction of MUC patients using the clinical characteristics. Mutation analysis results showed that deletion mutations in RB1, TP53, TSC1 and KDM6A were correlated with hypoxic status. Immune cell infiltration analysis illustrated that the infiltration T cells, B cells, Treg cells, and macrophages was correlated with hypoxia. Functional enrichment analysis revealed that a hypoxic microenvironment activated inflammatory pathways, glucose metabolism pathways, and immune-related pathways.Conclusion: In this investigation, a four-gene hypoxia risk model was developed to evaluate the degree of hypoxia and prognosis of ICI treatment, which showed a promising clinical prediction value in MUC. Furthermore, the hypoxia risk model revealed a close relationship between hypoxia and the tumor immune microenvironment.

The new SUMPOT to predict postoperative complications using an Artificial Neural Network

An accurate assessment of preoperative risk may improve use of hospital resources and reduce morbidity and mortality in high-risk surgical patients. This study aims at implementing an automated surgical risk calculator based on Artificial Neural Network technology to identify patients at risk for postoperative complications. We developed the new SUMPOT based on risk factors previously used in other scoring systems and tested it in a cohort of 560 surgical patients undergoing elective or emergency procedures and subsequently admitted to intensive care units, high-dependency units or standard wards. The whole dataset was divided into a training set, to train the predictive model, and a testing set, to assess generalization performance. The effectiveness of the Artificial Neural Network is a measure of the accuracy in detecting those patients who will develop postoperative complications. A total of 560 surgical patients entered the analysis. Among them, 77 patients (13.7%) suffered from one or more postoperative complications (PoCs), while 483 patients (86.3%) did not. The trained Artificial Neural Network returned an average classification accuracy of 90% in the testing set. Specifically, classification accuracy was 90.2% in the control group (46 patients out of 51 were correctly classified) and 88.9% in the PoC group (8 patients out of 9 were correctly classified). The Artificial Neural Network showed good performance in predicting presence/absence of postoperative complications, suggesting its potential value for perioperative management of surgical patients. Further clinical studies are required to confirm its applicability in routine clinical practice.

A supervised machine learning semantic segmentation approach for detecting artifacts in plethysmography signals from wearables

Objective. Wearable devices equipped with plethysmography (PPG) sensors provided a low-cost, long-term solution to early diagnosis and continuous screening of heart conditions. However PPG signals collected from such devices often suffer from corruption caused by artifacts. The objective of this study is to develop an effective supervised algorithm to locate the regions of artifacts within PPG signals. Approach. We treat artifact detection as a 1D segmentation problem. We solve it via a novel combination of an active-contour-based loss and an adapted U-Net architecture. The proposed algorithm was trained on the PPG DaLiA training set, and further evaluated on the PPG DaLiA testing set, WESAD dataset and TROIKA dataset. Main results. We evaluated with the DICE score, a well-established metric for segmentation accuracy evaluation in the field of computer vision. The proposed method outperforms baseline methods on all three datasets by a large margin (≈ 7 percentage points above the next best method). On the PPG DaLiA testing set, WESAD dataset and TROIKA dataset, the proposed method achieved 0.8734±0.0018, 0.9114±0.0033 and 0.8050±0.0116 respectively. The next best method only achieved 0.8068±0.0014, 0.8446±0.0013 and 0.7247±0.0050. Significance. The proposed method is able to pinpoint exact locations of artifacts with high precision; in the past, we had only a binary classification of whether a PPG signal has good or poor quality. This more nuanced information will be critical to further inform the design of algorithms to detect cardiac arrhythmia.

Development of a Radiomic-Based Model Predicting Lymph Node Involvement in Prostate Cancer Patients

Significant advances in lymph node involvement (LNI) risk modeling in prostate cancer (PCa) have been achieved with the addition of visual interpretation of magnetic resonance imaging (MRI) data, but it is likely that quantitative analysis could further improve prediction models. In this study, we aimed to develop and internally validate a novel LNI risk prediction model based on radiomic features extracted from preoperative multimodal MRI. All patients who underwent a preoperative MRI and radical prostatectomy with extensive lymph node dissection were retrospectively included in a single institution. Patients were randomly divided into the training (60%) and testing (40%) sets. Radiomic features were extracted from the index tumor volumes, delineated on the apparent diffusion coefficient corrected map and the T2 sequences. A ComBat harmonization method was applied to account for inter-site heterogeneity. A prediction model was trained using a neural network approach (Multilayer Perceptron Network, SPSS v24.0©) combining clinical, radiomic and all features. It was then evaluated on the testing set and compared to the current available models using the Receiver Operative Characteristics and the C-Index. Two hundred and eighty patients were included, with a median age of 65.2 y (45.3–79.6), a mean PSA level of 9.5 ng/mL (1.04–63.0) and 79.6% of ISUP ≥ 2 tumors. LNI occurred in 51 patients (18.2%), with a median number of extracted nodes of 15 (10–19). In the testing set, with their respective cutoffs applied, the Partin, Roach, Yale, MSKCC, Briganti 2012 and 2017 models resulted in a C-Index of 0.71, 0.66, 0.55, 0.67, 0.65 and 0.73, respectively, while our proposed combined model resulted in a C-Index of 0.89 in the testing set. Radiomic features extracted from the preoperative MRI scans and combined with clinical features through a neural network seem to provide added predictive performance compared to state of the art models regarding LNI risk prediction in PCa.