Conventional Features Research Articles

An integrated nomogram combining clinical and radiomic features of hyperattenuated imaging markers to predict malignant cerebral edema following endovascular thrombectomy.

Malignant cerebral edema (MCE), a potential complication following endovascular thrombectomy (EVT) in the treatment of acute ischemic stroke (AIS), can result in significant disability and mortality. This study aimed to develop a nomogram model based on the hyperattenuated imaging marker (HIM), characterized by hyperattenuation on head noncontrast computed tomography (CT) immediately after thrombectomy, to predict MCE in patients receiving EVT. In this retrospective cohort study, we selected 151 patients with anterior circulation large-vessel occlusion who received endovascular treatment. The patients were randomly allocated into training (n=121) and test (n=30) cohorts. HIM was used to extract radiomics characteristics. Conventional clinical and radiological features associated with MCE were also extracted. A model based on extreme gradient boosting (XGBoost) machine learning using fivefold cross-validation was employed to acquire radiomics and clinical features. Based on HIM, clinical and radiological signatures were used to construct a prediction nomogram for MCE. Subsequently, the signatures were merged through logistic regression (LR) analysis in order to create a comprehensive clinical radiomics nomogram. A total of 28 patients out of 151 (18.54%) developed MCE. The analysis of the receiver operating characteristic curve indicated an area under the curve (AUC) of 0.999 for the prediction of MCE in the training group and an AUC of 0.938 in the test group. The clinical and radiomics nomogram together showed the highest accuracy in predicting outcomes in both the training and test groups. The novel nomogram, which combines clinical manifestations and imaging findings based on postinterventional HIM, may serve as a predictor for MCE in patients experiencing AIS after EVT.

Intelligent Bladder Volume Monitoring for Wearable Ultrasound Devices: Enhancing Accuracy Through Deep Learning-Based Coarse-to-Fine Shape Estimation.

Accurate and continuous bladder volume monitoring is crucial for managing urinary dysfunctions. Wearable ultrasound (US) devices offer a solution by enabling noninvasive and real-time monitoring. Previous studies have limitations in power consumption and computation cost or quantitative volume estimation capability. To alleviate this, we present a novel pipeline that effectively integrates conventional feature extraction and deep learning (DL) to achieve continuous quantitative bladder volume monitoring efficiently. Particularly, in the proposed pipeline, bladder shape is coarsely estimated by a simple bladder wall detection algorithm in wearable devices, and the bladder wall coordinates are wirelessly transferred to an external server. Subsequently, a roughly estimated bladder shape from the wall coordinates is refined in an external server with a diffusion-based model. With this approach, power consumption and computation costs on wearable devices remained low, while fully harnessing the potential of DL for accurate shape estimation. To evaluate the proposed pipeline, we collected a dataset of bladder US images and RF signals from 250 patients. By simulating data acquisition from wearable devices using the dataset, we replicated real-world scenarios and validated the proposed method within these scenarios. Experimental results exhibit superior improvements, including +9.32% of IoU value in 2-D segmentation and -22.06 of RMSE in bladder volume regression compared to state-of-the-art (SOTA) performance from alternative methods, emphasizing the potential of this approach in continuous bladder volume monitoring in clinical settings. Therefore, this study effectively bridges the gap between accurate bladder volume estimation and the practical deployment of wearable US devices, promising improved patient care and quality of life.

Predicting MET exon 14 skipping mutation in pulmonary sarcomatoid carcinoma by whole-tumour texture analysis combined with clinical and conventional contrast-enhanced computed tomography features.

Pulmonary sarcomatoid carcinoma (PSC) is a rare, highly malignant type of non-small cell lung cancer (NSCLC) with a poor prognosis. Targeted drugs for MET exon 14 (METex14) skipping mutation can have considerable clinical benefits. This study aimed to predict METex14 skipping mutation in PSC patients by whole-tumour texture analysis combined with clinical and conventional contrast-enhanced computed tomography (CECT) features. This retrospective study included 56 patients with PSC diagnosed by pathology. All patients underwent CECT before surgery or other treatment, and both targeted DNA- and RNA-based next-generation sequencing (NGS) were used to detect METex14 skipping mutation status. The patients were divided into two groups: METex14 skipping mutation and nonmutation groups. Overall, 1,316 texture features of the whole tumour were extracted. We also collected 12 clinical and 20 conventional CECT features. After dimensionality reduction and selection, predictive models were established by multivariate logistic regression analysis. Models were evaluated using the area under the curve (AUC), and the clinical utility of the model was assessed by decision curve analysis. METex14 skipping mutation was detected in 17.9% of PSCs. Mutations were found more frequently in those (I) who had smaller long- or short-axis diameters (P=0.02, P=0.01); (II) who had lower T stages (I, II) (P=0.02); and (III) with pseudocapsular or annular enhancement (P=0.03). The combined model based on the conventional and texture models yielded the best performance in predicting METex14 skipping mutation with the highest AUC (0.89). The conventional and texture models also had good performance (AUC =0.83 conventional; =0.88 texture). Whole-tumour texture analysis combined with clinical and conventional CECT features may serve as a noninvasive tool to predict the METex14 skipping mutation status in PSC.

Automated structural repair based on continuous carbon fiber reinforced plastic 3D printing and online model reconstruction

AbstractAdditive manufacturing of continuous carbon fiber reinforced plastic (CCFRP) has gained increasing prominence in structural repair through material accumulation. However, the repairing process entails laborious preliminary tasks, such as positioning, model reconstruction, and data input. In this article, computer vision is integrated into a CCFRP 3D printing device and a printing path‐selecting strategy is proposed to automate the repair process with accuracy and efficiency. An irregularly damaged structure is placed arbitrarily on the printing platform. Subsequently, the damaged area is extracted by semantic segmentation to construct a 3D model. The printing paths including zig‐zag and contour offset infilling for an irregular damage area are selected, according to the prediction of flexural modulus based on the finite element method. Finally, an in‐situ composite printing is conducted based on coaxial extrusion principle. The results show that the intersection over union (IOU) of regular damaged areas by the applied pyramid scene parsing network (PSPNet) segmentation model is 0.81. Semantic segmentation is more robust than conventional feature extraction based on filtering methods. The appearance of the repaired structure is consistent with that of the surrounding structure. The combination of CV and additive manufacturing technology presents a new way to simplify the repair process and improve repair effects.Highlights An automated repairing system for irregular damaged areas is established. Deep learning for semantic segmentation is integrated into 3D printing. Finite element submodule is introduced to select optimal printing paths. The CCFRP 3D printing is introduced to surface defect repairing.

Feature engineering and machine learning for electrochemical detection of rabies virus in graphene-based biosensors

Electrochemical biosensors are small analytical devices that convert a biological response into a processable signal with high sensitivity, ease of operation, cost-effectiveness, and miniaturization capability. The current study proposes enhancing the reliability of the electrochemical detection of enveloped viruses, such as the rabies virus using feature engineering and machine learning. Portable detection was achieved using a graphene-based microfluidic sensor and the current was recorded from electrochemical experiments using a portable potentiostat. After data mining and feature engineering, a dataset obtained from staircase cyclic voltammetry was used in different machine learning models. The features from the voltammogram were extracted following theoretic aspects to build a dataset with uncovering patterns and valuable information. Correlation-based and recursive feature elimination algorithms were used for feature engineering. Our analysis showed that the best F-measure score a model trained was obtained through a support vector machine with 0.9830 for the diagnostic tasks following the proposed feature engineering pipeline. On the other hand, the preprocessing pipeline which is commonly used in many studies obtained a score of 0.9394 using a decision tree model. In addition, the proposed method generates features with higher interpretability and lower model complexity than conventional feature reduction methods such as PCA. Our study shows the potential of machine learning models in the field of electrochemical biosensor development with a focus on feature engineering based on electroanalytical analysis, which can be extended to develop new tools for disease diagnosis in the future.

Non-intrusive fault detection in shipboard power systems using wavelet graph neural networks

Naval shipboard power systems (SPS) are rapidly embracing electrification, resulting in loads that generate pulsation currents and encounter substantial transients. However, conventional time-based features alone are inadequate for effectively monitoring and safeguarding these loads against faults. This highlights the critical requirement for advanced machine learning based methods to discern and differentiate between the various transient stages within the load profile. In this paper, we propose a Wavelet Graph Neural Network (WGNN) model for non-intrusive fault detection in SPS. The fault detection system leverages the dynamic model of the SPS to train and test performance with varying fault scenarios. The underlying structure and the interdependence among component states in the SPS network are effectively captured using the WGNN model, resulting in accuracies over 99% for intrusive fault detection and 97% for non-intrusive fault detection. The developed WGNN model has also shown to be robust in the presence of pulse loads and noise, achieving an accuracy of over 95%. At the end, a real-time simulation of the proposed method is validated on a hardware-in-the-loop system, guaranteeing the high fidelity and low latency of the proposed approach. These findings validate the effectiveness of the proposed WGNN model for fault detection and real-world applications in SPS.

Impact of quantitative CT texture analysis on the outcome of CT-guided bone biopsy

Texture analysis can provide new imaging-based biomarkers. Texture analysis derived from computed tomography (CT) might be able to better characterize patients undergoing CT-guided percutaneous bone biopsy. The present study evaluated this and correlated texture features with bioptic outcome in patients undergoing CT-guided bone biopsy. Overall, 123 patients (89 female patients, 72.4 %) were included into the present study. All patients underwent CT-guided percutaneous bone biopsy with an 11 Gauge coaxial needle. Clinical parameters and quantitative imaging features were investigated. Random forest classifier was used to predict a positive biopsy result. Overall, 69 patients had osteolytic metastasis (56.1 %) and 54 had osteoblastic metastasis (43.9 %). The overall positive biopsy rate was 72 %. The developed radiomics model demonstrated a prediction accuracy of a positive biopsy result with an AUC of 0.75 [95 %CI 0.65 – 0.85]. In a subgroup of breast cancer patients, the model achieved an AUC of 0.85 [95 %CI 0.73 – 0.96]. In the subgroup of non-breast cancer patients, the signature achieved an AUC of 0.80 [95 %CI 0.60 – 0.99]. Quantitative CT imaging findings comprised of conventional and texture features can aid to predict the bioptic result of CT-guided bone biopsies. The developed radiomics signature aids in clinical decision-making, and could identify patients at risk for a negative biopsy.

Qualitative and quantitative assessment of non-clear cell renal cell carcinoma using contrast-enhanced ultrasound

BackgroundNon-clear cell renal cell carcinoma (nccRCC) represents a rare form of renal cell carcinoma (RCC) in the clinic. It is now understood that contrast-enhanced ultrasound (CEUS) exhibits diverse manifestations and can be prone to misdiagnosis. Therefore, summarizing the distinctive features of contrast-enhanced ultrasonography is essential for differentiation from ccRCC.ObjectiveThis study aims to evaluate the diagnostic efficacy of qualitative and quantitative CEUS in diagnosing nccRCC to enhance our understanding of this condition.MethodsWe conducted a retrospective analysis of 21 patients with confirmed nccRCC following surgery and assessed the characteristic conventional ultrasound and CEUS imaging features. The paired Wilcoxon signed-rank sum test was employed to compare differences in CEUS time-intensity curve (TIC) parameters between the lesions and the normal renal cortex.ResultsRoutine ultrasound revealed the following primary characteristics in the 21 nccRCC cases: hypoechoic appearance (10/21, 47.6%), absence of liquefaction (18/21, 66.7%), regular shape (19/21, 90.5%), clear boundaries (21/21, 100%), and absence of calcification (17/21, 81%). Color Doppler flow imaging (CDFI) indicated a low blood flow signal (only 1 case of grade III). Qualitative CEUS analysis demonstrated that nccRCC predominantly exhibited slow progression (76.1%), fast washout (57%), uniformity (61.9%), low enhancement (71.5%), and ring enhancement (61.9%). Quantitative CEUS analysis revealed that parameters such as PE, WiAUC, mTTI, WiR, WiPI, WoAUC, WiWoAUC, and WOR in the lesions were significantly lower than those in the normal renal cortex (Z=-3.980, -3.563, -2.427, -3.389, -3.980, -3.493, -3.528, -2.763, P < 0.001, < 0.001, = 0.015, = 0.001, < 0.001, < 0.001, < 0.001, = 0.006). However, there were no significant differences in RT, TTP, FT, or QOF (all P > 0.05).ConclusionnccRCC exhibits distinctive CEUS characteristics, including slow progression, fast washout, low homogeneity enhancement, and ring enhancement, which can aid in distinguishing nccRCC from ccRCC.

Predicting histological grade in pediatric glioma using multiparametric radiomics and conventional MRI features

Prediction of glioma is crucial to provide a precise treatment plan to optimize the prognosis of children with glioma. However, studies on the grading of pediatric gliomas using radiomics are limited. Meanwhile, existing methods are mainly based on only radiomics features, ignoring intuitive information about tumor morphology on traditional imaging features. This study aims to utilize multiparametric magnetic resonance imaging (MRI) to identify high-grade and low-grade gliomas in children and establish a classification model based on radiomics features and clinical features. A total of 85 children with gliomas underwent tumor resection, and part of the tumor tissue was examined pathologically. Patients were categorized into high-grade and low-grade groups according to World Health Organization guidelines. Preoperative multiparametric MRI data, including contrast-enhanced T1-weighted imaging, T2-weighted imaging, T2-weighted fluid-attenuated inversion recovery, diffusion-weighted images, and apparent diffusion coefficient sequences, were obtained and labeled by two radiologists. The images were preprocessed, and radiomics features were extracted for each MRI sequence. Feature selection methods were used to select radiomics features, and statistically significant clinical features were identified using t-tests. The selected radiomics features and conventional MRI features were used to train the AutoGluon models. The improved model, based on radiomics features and conventional MRI features, achieved a balanced classification accuracy of 66.59%. The cross-validated areas under the receiver operating characteristic curve for the classifier of AutoGluon frame were 0.8071 on the test dataset. The results indicate that the performance of AutoGluon models can be improved by incorporating conventional MRI features, highlighting the importance of the experience of radiologists in accurately grading pediatric gliomas. This method can help predict the grade of pediatric glioma before pathological examination and assist in determining the appropriate treatment plan, including radiotherapy, chemotherapy, drugs, and gene surgery.

Differentiating small (< 2 cm) pancreatic ductal adenocarcinoma from neuroendocrine tumors with multiparametric MRI-based radiomic features.

To assess MR-based radiomic analysis in preoperatively discriminating small (< 2 cm) pancreatic ductal adenocarcinomas (PDACs) from neuroendocrine tumors (PNETs). A total of 197 patients (146 in the training cohort, 51 in the validation cohort) from two centers were retrospectively collected. A total of 7338 radiomics features were extracted from T2-weighted, diffusion-weighted, T1-weighted, arterial phase, portal venous phase and delayed phase imaging. The optimal features were selected by the Mann-Whitney U test, Spearman's rank correlation test and least absolute shrinkage and selection operator method and used to construct the radiomic score (Rad-score). Conventional radiological and clinical features were also assessed. Multivariable logistic regression was used to construct a radiological model, a radiomic model and a fusion model. Nine optimal features were identified and used to build the Rad-score. The radiomic model based on the Rad-score achieved satisfactory results with AUCs of 0.905 and 0.930, sensitivities of 0.780 and 0.800, specificities of 0.906 and 0.952 and accuracies of 0.836 and 0.863 for the training and validation cohorts, respectively. The fusion model, incorporating CA19-9, tumor margins, pancreatic duct dilatation and the Rad-score, exhibited the best performance with AUCs of 0.977 and 0.941, sensitivities of 0.914 and 0.852, specificities of 0.954 and 0.950, and accuracies of 0.932 and 0.894 for the training and validation cohorts, respectively. The MR-based Rad-score is a novel image biomarker for discriminating small PDACs from PNETs. A fusion model combining radiomic, radiological and clinical features performed very well in differentially diagnosing these two tumors. A fusion model combining MR-based radiomic, radiological, and clinical features could help differentiate between small pancreatic ductal adenocarcinomas and pancreatic neuroendocrine tumors. Preoperatively differentiating small pancreatic ductal adenocarcinomas (PDACs) and pancreatic neuroendocrine tumors (PNETs) is challenging. Multiparametric MRI-based Rad-score can be used for discriminating small PDACs from PNETs. A fusion model incorporating radiomic, radiological, and clinical features differentiated small PDACs from PNETs well.

SEP-AlgPro: An efficient allergen prediction tool utilizing traditional machine learning and deep learning techniques with protein language model features

Allergy is a hypersensitive condition in which individuals develop objective symptoms when exposed to harmless substances at a dose that would cause no harm to a “normal” person. Most current computational methods for allergen identification rely on homology or conventional machine learning using limited set of feature descriptors or validation on specific datasets, making them inefficient and inaccurate. Here, we propose SEP-AlgPro for the accurate identification of allergen protein from sequence information. We analyzed 10 conventional protein-based features and 14 different features derived from protein language models to gauge their effectiveness in differentiating allergens from non-allergens using 15 different classifiers. However, the final optimized model employs top 10 feature descriptors with top seven machine learning classifiers. Results show that the features derived from protein language models exhibit superior discriminative capabilities compared to traditional feature sets. This enabled us to select the most discriminatory baseline models, whose predicted outputs were aggregated and used as input to a deep neural network for the final allergen prediction. Extensive case studies showed that SEP-AlgPro outperforms state-of-the-art predictors in accurately identifying allergens. A user-friendly web server was developed and made freely available at https://balalab-skku.org/SEP-AlgPro/, making it a powerful tool for identifying potential allergens.

Novel sound event and sound activity detection framework based on intrinsic mode functions and deep learning

AbstractThe detection of sound events has become increasingly important due to the development of signal processing methods, social media, and the need for automatic labeling methods in applications such as smart cities, navigation, and security systems. For example, in such applications, it is often important to detect sound events at different levels, such as the presence or absence of an event in the segment, or to specify the beginning and end of the sound event and its duration. This study proposes a method to reduce the feature dimensions of a Sound Event Detection (SED) system while maintaining the system’s efficiency. The proposed method, using Empirical Mode Decomposition (EMD), Intrinsic Mode Functions (IMFs), and extraction of locally regulated features from different IMFs of the signal, shows a promising performance relative to the conventional features of SED systems. In addition, the feature dimensions of the proposed method are much smaller than those of conventional methods. To prove the effectiveness of the proposed features in SED tasks, two segment-based approaches for event detection and sound activity detection were implemented using the suggested features, and their effectiveness was confirmed. Simulation results on the URBAN SED dataset showed that the proposed approach reduces the number of input features by more than 99% compared with state-of-the-art methods while maintaining accuracy. According to the obtained results, the proposed method is quite promising.

Feature selection based on contradictory state sequence for multi-scale interval valued decision table

In the context of selecting features for multi-scale interval valued decision table (MIVDT), conventional research approaches encounter difficulties including elevated data complexity, diminished computational efficiency, and limited model generalization capacity. To overcome these difficulties, feature selection methods based on contradictory state sequence (CSS) and fuzzy contradictory state sequence (FCSS) are proposed in this paper. Initially, MIVDT is established. According to the characteristics of interval value, a more thorough and impartial metric for assessing the inclusion degree is suggested, along with similarity relation based on the metric. Subsequently, the introduction of the first contradictory object allows for the delineation of contradictory state and fuzzy contradictory state, which serve to characterize the consistency of MIVDT. These two concepts are proposed to enhance the accuracy and efficiency of capturing key information in intricate data. Additionally, rapid feature selection algorithms are suggested, which rely on CSS and FCSS. In contrast to conventional feature selection approaches, the algorithms introduced in this study demonstrate superior computational efficiency and enhanced generalization capabilities when confronted with intricate datasets. In twelve open source datasets, the upper limit for the quantity of objects is 110341, and the average accuracy values of experiments under two parameter combinations are 99.54% and 97.08% respectively. The results of experiments demonstrate that the algorithms introduced are capable of efficiently identifying a novel feature subset from intricate datasets within a shorter timeframe.

A PRISMA-driven Review of Speech Recognition based on English, Mandarin Chinese, Hindi and Urdu Language

Urdu Language ranks ten and is continuously progressing. This unique PRISMA-Driven review deeply investigates Urdu speech recognition literature and adjoin it with English, Mandarin Chinese, and Hindi languages frame-works conceptualizing wider global perspective. The main objective is to unify progress on classical Artificially Intelligent (AI) and recent Deep Neural Networks (DNN) based speech recognition pipeline encompassing Dataset challenges, Feature extraction methods, Experimental design and the smooth integration with both Acoustic models (AM) and Language models (LM) using Transcriptions. A total of 176 articles were extracted from Google Scholar database for each language with custom query design. Inclusion criteria and quality assessment leads to end up with 5 review and 42 research articles. Comparative research questions have been addressed and findings were organized by four possible speech types: Isolated, connected, continuous and spontaneous. The finding shows that English, Mandarin, and Hindi languages used spontaneous speech size of 300, 200 and 1108 hours respectively which is quite remarkable as compared to Urdu spontaneous speech data size of only 9.5 hours. For the same data size reason, the Word Error Rate (WER) for English falls below 5% while for Mandarin Chinese the alternative metric Character Error Rate (CER) is mostly used that lies below 25%. The success of English and Chinese Speech recognition leads to incomparable accuracy due to wide use of DNNs like Conformer, Transformers, E2E-attention in comparison to conventional feature extraction and AI models LSTM, TDNN, RNN, HMM, GMM-HMM; used frequently by both Hindi and Urdu.

A two-stage clonal selection algorithm for local feature selection on high-dimensional data

Various evolutionary computation algorithms have shown their excellent performance for high-dimensional feature selection (FS). However, most of current FS methods choose a global feature subset and ignore the correlations between feature subsets and sample subspaces, which limits the performance of methods in the sample space with different probability distributions. To solve the issue, we propose a two-stage clonal selection algorithm for filter-based local feature selection (TSCSA-LFS) that integrates symmetric uncertainty and a discrete clonal selection algorithm with three contributions. First, unlike conventional feature selection methods, TSCSA-LFS introduces local sample behaviors and assigns subsets of features for different sample regions. Second, an improved discrete clonal selection algorithm is developed for searching relevant features, which contains mutual information-based individual initialization, a differential evolution-based mutation strategy pool and a local search technique. Third, a two-part antibody representation is employed for automatical adjustment of the weight-related parameter. Our method shows the promising experimental results compared with well-known global FS and clonal selection-based local FS methods on fourteen high-dimensional datasets. For instance, our method can obviously outperform local FS, filter-based and hybrid methods on at least nine datasets

Breast Cancer: Multi-b-Value Diffusion Weighted Habitat Imaging in Predicting Pathologic Complete Response to Neoadjuvant Chemotherapy

Rationale and ObjectivesThe aim of this study was to ascertain whether the utilization of multiple b-value diffusion-weighted habitat imaging, a technique that depicts tumor heterogeneity, could aid in identifying breast cancer patients who would derive substantial benefit from neoadjuvant chemotherapy (NAC). Materials and MethodsThis prospective study enrolled 143 women (II-III breast cancer), who underwent multi-b-value diffusion-weighted imaging (DWI) in 3-T magnetic resonance (MR) before NAC. The patient cohort was partitioned into a training set (consisting of 100 patients, of which 36 demonstrated a pathologic complete response [pCR]) and a test set (featuring 43 patients, 16 of whom exhibited pCR). Utilizing the training set, predictive models for pCR, were constructed using different parameters: whole-tumor radiomics (ModelWH), diffusion-weighted habitat-imaging (ModelHabitats), conventional MRI features (ModelCF), along with combined models ModelHabitats+CF. The performance of these models was assessed based on the area under the receiver operating characteristic curve (AUC) and calibration slope. ResultsIn the prediction of pCR, ModelWH, ModelHabitats, ModelCF, and ModelHabitats+CF achieved AUCs of 0.733, 0.722, 0.705, and 0.756 respectively, within the training set. These scores corresponded to AUCs of 0.625, 0.801, 0.700, and 0.824 respectively in the test set. The DeLong test revealed no significant difference between ModelWH and ModelHabitats (P = 0.182), between ModelHabitats and ModelHabitats+CF (P = 0.113). ConclusionThe habitat model we developed, incorporating first-order features along with conventional MRI features, has demonstrated accurate predication of pCR prior to NAC. This model holds the potential to augment decision-making processes in personalized treatment strategies for breast cancer.

Discriminating atypical parotid carcinoma and pleomorphic adenoma utilizing extracellular volume fraction and arterial enhancement fraction derived from contrast-enhanced CT imaging: A multicenter study.

To investigate the added value of extracellular volume fraction (ECV) and arterial enhancement fraction (AEF) derived from enhanced CT to conventional image and clinical features for differentiating between pleomorphic adenoma (PA) and atypical parotid adenocarcinoma (PCA) pre-operation. From January 2010 to October 2023, a total of 187 cases of parotid tumors were recruited, and divided into training cohort (102 PAs and 51 PCAs) and testing cohort (24 PAs and 10 atypical PCAs). Clinical and CT image features of tumor were assessed. Both enhanced CT-derived ECV and AEF were calculated. Univariate analysis identified variables with statistically significant differences between the two subgroups in the training cohort. Multivariate logistic regression analysis with the forward variable selection method was used to build four models (clinical model, clinical model+ECV, clinical model+AEF, and combined model). Diagnostic performances were evaluated using receiver operating characteristic (ROC) curve analyses. Delong's test compared model differences, and calibration curve and decision curve analysis (DCA) assessed calibration and clinical application. Age and boundary were chosen to build clinical model, and to construct its ROC curve. Amalgamating the clinical model, ECV, and AEF to establish a combined model demonstrated superior diagnostic effectiveness compared to the clinical model in both the training and test cohorts (AUC = 0.888, 0.867). There was a significant statistical difference between the combined model and the clinical model in the training cohort (p = 0.0145). ECV and AEF are helpful in differentiating PA and atypical PCA, and integrating clinical and CT image features can further improve the diagnostic performance.

Improving Prediction of Survival and Progression in Metastatic Non-Small Cell Lung Cancer After Immunotherapy Through Machine Learning of Circulating Tumor DNA.

To use modern machine learning approaches to enhance and automate the feature extraction from the longitudinal circulating tumor DNA (ctDNA) data and to improve the prediction of survival and disease progression, risk stratification, and treatment strategies for patients with 1L non-small cell lung cancer (NSCLC). Using IMpower150 trial data on patients with untreated metastatic NSCLC treated with atezolizumab and chemotherapies, we developed a machine learning algorithm to extract predictive features from ctDNA kinetics, improving survival and progression prediction. We analyzed kinetic data from 17 ctDNA summary markers, including cell-free DNA concentration, allele frequency, tumor molecules in plasma, and mutation counts. Three hundred and ninety-eight patients with ctDNA data (206 in training and 192 in validation) were analyzed. Our models outperformed existing workflow using conventional temporal ctDNA features, raising overall survival (OS) concordance index to 0.72 and 0.71 from 0.67 and 0.63 for C3D1 and C4D1, respectively, and substantially improving progression-free survival (PFS) to approximately 0.65 from the previous 0.54-0.58, a 12%-20% increase. Additionally, they enhanced risk stratification for patients with NSCLC, achieving clear OS and PFS separation. Distinct patterns of ctDNA kinetic characteristics (eg, baseline ctDNA markers, depth of ctDNA responses, and timing of ctDNA clearance, etc) were revealed across the risk groups. Rapid and complete ctDNA clearance appears essential for long-term clinical benefit. Our machine learning approach offers a novel tool for analyzing ctDNA kinetics, extracting critical features from longitudinal data, improving our understanding of the link between ctDNA kinetics and progression/mortality risks, and optimizing personalized immunotherapies for 1L NSCLC.

Privacy-Preserving Synthetic Continual Semantic Segmentation for Robotic Surgery.

Deep Neural Networks (DNNs) based semantic segmentation of the robotic instruments and tissues can enhance the precision of surgical activities in robot-assisted surgery. However, in biological learning, DNNs cannot learn incremental tasks over time and exhibit catastrophic forgetting, which refers to the sharp decline in performance on previously learned tasks after learning a new one. Specifically, when data scarcity is the issue, the model shows a rapid drop in performance on previously learned instruments after learning new data with new instruments. The problem becomes worse when it limits releasing the dataset of the old instruments for the old model due to privacy concerns and the unavailability of the data for the new or updated version of the instruments for the continual learning model. For this purpose, we develop a privacy-preserving synthetic continual semantic segmentation framework by blending and harmonizing (i) open-source old instruments foreground to the synthesized background without revealing real patient data in public and (ii) new instruments foreground to extensively augmented real background. To boost the balanced logit distillation from the old model to the continual learning model, we design overlapping class-aware temperature normalization (CAT) by controlling model learning utility. We also introduce multi-scale shifted-feature distillation (SD) to maintain long and short-range spatial relationships among the semantic objects where conventional short-range spatial features with limited information reduce the power of feature distillation. We demonstrate the effectiveness of our framework on the EndoVis 2017 and 2018 instrument segmentation dataset with a generalized continual learning setting. Code is available at https://github.com/XuMengyaAmy/Synthetic_CAT_SD.

Cartoon copyright recognition method based on character personality action

Aiming at the problem of cartoon piracy and plagiarism, this paper proposes a method of cartoon copyright recognition based on character personality actions. This method can be used to compare the original cartoon actions with the action characteristics of pirated or copied cartoons to identify whether there is piracy or plagiarism. Firstly, an image preprocessing scheme for character extraction is designed. GrabCut interactive image segmentation algorithm was used to obtain cartoon characters, and then binarization and morphological processing were performed on the results. Secondly, a feature extraction scheme based on character profile, moving character and character pose is designed. By extracting the perimeter and area of the character contour, the length-to-width ratio of the smallest rectangle and the inclination angle of the contour, the character contour features are obtained. The three-dimensional coordinates are established by the central point position of the cartoon character in the two-dimensional image and the change of the character's zoom in and out, and the character's motion angle characteristics are calculated. By skeletonizing a character to obtain the pose characteristics, and using deburring operation to remove redundant branches, then extract the skeleton joint angle information. Finally, feature fusion is performed on the extracted features. The experimental results show that the proposed method breaks the limitation of the conventional single feature based recognition, and can better extract the character features including contour feature, motion feature and pose feature through multi-feature based extraction, so as to protect the cartoon copyright.