Hand-crafted Features Research Articles

Evaluating Methods for Violence Classification and Firearm Detection in Indoor CCTV Environment

The adoption of security systems based on computer vision for violence detection has the potential to significantly improve safety in various public and private properties. However, developing these systems can be extremely challenging.We can choose to use classification models to identify violence in images or also use object detection models to identify firearms, which may indicate robbery. Additionally, when developing such systems focused on private environments, we encounter specific challenges, such as obtaining appropriate datasets to train the algorithms. Many publicly available datasets for violence detection consist of outdoor images, with elements such as streets and cars, which do not adequately reflect the nuances and unique characteristics of private properties. In this work, we evaluate both learned and handcrafted features to classify videos as 'violence' or 'non-violence' across a variety of datasets, including a new dataset composed exclusively of closed-circuit television (CCTV) images. Additionally, we propose a new dataset for firearm detection in CCTV images and conduct some experiments using YoloV8. In this way, we hope to provide a clearer insight into the possible decisions when developing a security system for indoor environments.

Performance improvement of active sonar target classification using feature fusion

In this study, we propose a neural network architecture considering feature fusion to improve active sonar target classification, where training data is scarce. We use various hand-crafted features with different perspectives for the active sonar one- and two-dimention raw data to enhance target classification performance. We use two kinds of features, in 1D feature and 2D features, because different perspectives on the same sample complement effectively by exchanging. For this, we contrived attention based complementary learning module to extract comprehensive features using the two kinds of features. To verify the generalization performance of this model, two real-ocean datasets were used and the classification results are compared with existing DL models.

Environmental sound classification using raw-audio based ensemble framework

Environmental sound classification (ESC) aims to automatically recognize audio recordings from the underlying environment, such as "urban park" or "city centre". Most of the existing methods for ESC use hand-crafted time-frequency features such as log-mel spectrogram to represent audio recordings. However, the hand-crafted features rely on transformations that are defined beforehand and do not consider the variability in the environment due to differences in recording conditions or recording devices. To overcome this, we present an alternative representation framework by leveraging a pre-trained convolutional neural network, SoundNet, trained on a large-scale audio dataset to represent raw audio recordings. We observe that the representations obtained from the intermediate layers of SoundNet lie in low-dimensional subspace. However, the dimensionality of the low-dimensional subspace is not known. To address this, an automatic compact dictionary learning framework is utilized that gives the dimensionality of the underlying subspace. The low-dimensional embeddings are then aggregated in a late-fusion manner in the ensemble framework to incorporate hierarchical information learned at various intermediate layers of SoundNet. We perform experimental evaluation on publicly available DCASE 2017 and 2018 ASC datasets. The proposed ensemble framework improves performance between 1 and 4 percentage points compared to that of existing time-frequency representations.

Early Detection of Residual/Recurrent Lung Malignancies on Post-Radiation FDG PET/CT

Positron Emission Tomography/Computed Tomography (PET/CT) using Fluorodeoxyglucose (FDG) is an important imaging modality for assessing treatment outcomes in patients with pulmonary malignant neoplasms undergoing radiation therapy. However, distinguishing between benign post-radiation changes and residual or recurrent malignancies on PET/CT images is challenging. Leveraging the potential of artificial intelligence (AI), we aimed to develop a hybrid fusion model integrating radiomics and Convolutional Neural Network (CNN) architectures to improve differentiation between benign post-radiation changes and residual or recurrent malignancies on PET/CT images. We retrospectively collected post-radiation PET/CTs with identified labels for benign changes or residual/recurrent malignant lesions from 95 lung cancer patients who received radiation therapy. Firstly, we developed separate radiomics and CNN models using handcrafted and self-learning features, respectively. Then, to build a more reliable model, we fused the probabilities from the two models through an evidential reasoning approach to derive the final prediction probability. Five-folder cross-validation was performed to evaluate the proposed radiomics, CNN, and fusion models. Overall, the hybrid fusion model outperformed the other two models in terms of sensitivity, specificity, accuracy, and the area under the curve (AUC) with values of 0.67, 0.72, 0.69, and 0.72, respectively. Evaluation results on the three AI models we developed suggest that handcrafted features and learned features may provide complementary information for residual or recurrent malignancy identification in PET/CT.

Alg-MFDL: A multi-feature deep learning framework for allergenic proteins prediction

The escalating global incidence of allergy patients illustrates the growing impact of allergic issues on global health. Allergens are small molecule antigens that trigger allergic reactions. A widely recognized strategy for allergy prevention involves identifying allergens and avoiding re-exposure. However, the laboratory methods to identify allergenic proteins are often time-consuming and resource-intensive. There is a crucial need to establish efficient and reliable computational approaches for the identification of allergenic proteins. In this study, we developed a novel allergenic proteins predictor named Alg-MFDL, which integrates pre-trained protein language models (PLMs) and traditional handcrafted features to achieve a more complete protein representation. First, we compared the performance of eight pre-trained PLMs from ProtTrans and ESM-2 and selected the best-performing one from each of the two groups. In addition, we evaluated the performance of three handcrafted features and different combinations of them to select the optimal feature or feature combination. Then, these three protein representations were fused and used as inputs to train the convolutional neural network (CNN). Finally, the independent validation was performed on benchmark datasets to evaluate the performance of Alg-MFDL. As a result, Alg-MFDL achieved an accuracy of 0.973, a precision of 0.996, a sensitivity of 0.951, and an F1 value of 0.973, outperforming the most of current state-of-the-art (SOTA) methods across all key metrics. We anticipated that the proposed model could be considered a useful tool for predicting allergen proteins. The datasets and code utilized in this study are freely available at https://github.com/Hupenpen/Alg-MFDL.

Surface quality prediction in-situ monitoring system: A deep transfer learning-based regression approach with audible signal

Surface roughness plays an indispensable and fundamental role as a leading indicator of the surface quality of machined parts in the manufacturing process. The precise and effective monitoring and prediction of surface roughness is crucial for surface quality control. In this regard, the development of an in-process surface quality monitoring system is necessary, which has the promising potential to achieve this goal. Such a system typically comprises data-driven models for decision-making and sensing techniques for detecting associated process information. However, some challenges still exist in building such systems. Firstly, the architecture design and deployment of data-driven models, specifically deep learning (DL)-based models, demand adequate domain knowledge. Secondly, most models trained on specific tasks with limited datasets are prone to suppressing their versatility and generalization across different machining conditions. Additionally, in most cases, reliance on handcrafted features to represent dynamic information on various signals during model training necessitates extensive expertise in selecting appropriate feature types. Furthermore, due to the nature of their low dimensionality, handcrafted features have difficulty in capturing of overall process-related underlying patterns from dynamics signatures, which is time-varying and often occurs in transient events. To address these challenges, this paper proposes the regression-based pre-trained convolutional neural network (pre-trained CNN) combined with Mel-spectrogram images based on the transfer learning method for surface roughness prediction. Within the context, the architecture of the transfer model is slightly adapted from already well-trained CNNs. Initial weights in each layer of the CNN model are directly inherited and then fine-tuned through the Bayesian optimization tuning method. Besides, the audible sound signals are captured and subsequently converted into 2D Mel-spectrogram images with variant time lengths, which are separately engaged to retrain and validate four existing pre-trained CNN models (VGG16, VGG19, ResNet50V2 and InceptionResNetV2). Eventually, the effectiveness of proposed models and comparison of their predictive capabilities are further validated through a case study in the turning process. The results demonstrate that each applied pre-trained CNN model is capable of effectively predicting surface quality with satisfactory prediction results. Therefore, the proposed method can facilitate the establishment of a machining monitoring system concerning its accuracy, reliability, and robustness.

A boundary-aware point clustering approach in Euclidean and embedding spaces for roof plane segmentation

Roof plane segmentation from airborne light detection and ranging (LiDAR) point clouds is an important technology for three-dimensional (3D) building model reconstruction. One of the key issues of plane segmentation is how to design powerful features that can exactly distinguish adjacent planar patches. The quality of point feature directly determines the accuracy of roof plane segmentation. Most of existing approaches use handcrafted features, such as point-to-plane distance, normal vector, etc., to extract roof planes. However, the abilities of these features are relatively low, especially in boundary areas. To solve this problem, we propose a boundary-aware point clustering approach in Euclidean and embedding spaces constructed by a multi-task deep network for roof plane segmentation. We design a three-branch multi-task network to predict semantic labels, point offsets and extract deep embedding features. In the first branch, we classify the input data as non-roof, boundary and plane points. In the second branch, we predict point offsets for shifting each point towards its respective instance center. In the third branch, we constrain that points of the same plane instance should have the similar embeddings. We aim to ensure that points of the same plane instance are close as much as possible in both Euclidean and embedding spaces. However, although deep network has strong feature representative ability, it is still hard to accurately distinguish points near the plane instance boundary. Therefore, we first robustly group plane points into many clusters in Euclidean and embedding spaces to find candidate planes. Then, we assign the rest boundary points to their closest clusters to generate the final complete roof planes. In this way, we can effectively reduce the influence of unreliable boundary points. In addition, to train the network and evaluate the performance of our approach, we prepare a synthetic dataset and two real datasets. The experiments conducted on synthetic and real datasets show that the proposed approach significantly outperforms the existing state-of-the-art approaches in both qualitative evaluation and quantitative metrics. To facilitate future research, we will make datasets and source code of our approach publicly available at https://github.com/Li-Li-Whu/DeepRoofPlane.

Emotional parameter estimation from emo-soundscapes dataset using Deep Convolutional Autoencoders

AbstractPredicting the emotional responses of humans to acoustic features in the surrounding environment has a highly potential of applications in different fields, ranging from videogames, therapeutic use of virtual reality to the emotional design of spaces according to their expected use. In this paper we model the estimation process of the classical emotion characterization parameters (arousal and valence) from sounds. By means of convolutional neural networks and convolutional autoencoders, the model is adjusted for the prediction of these parameters from a standard dataset [1], improving the results obtained in previous literature. The relevance of the work, apart from improving the results due to the use of autoencoders, is that it eliminates the need to compute handcrafted features, thus demonstrating the ability of convolutional neural networks to treat raw information. Other main contributions of the paper is a new way to visualize the errors in the joint estimation of arousal and valence that facilitates the evaluation of the results obtained by the models. Finally, the use of bootstrap to estimate the confidence intervals of the MSE and r2 of Deep Learning models shows that in comparison to non-overlapping samples, overlapping samples introduces performance bias.

An approach for classification of breast cancer using lightweight deep convolution neural network

The rapid advancement of deep learning has generated considerable enthusiasm regarding its utilization in addressing medical imaging issues. Machine learning (ML) methods can help radiologists to diagnose breast cancer (BCs) barring invasive measures. Informative hand-crafted features are essential prerequisites for traditional machine learning classifiers to achieve accurate results, which are time-consuming to extract. In this paper, our deep learning algorithm is created to precisely identify breast cancers on screening mammograms, employing a training method that effectively utilizes training datasets with either full clinical annotation or solely the cancer status of the entire image. The proposed approach utilizes Lightweight Convolutional Neural Network (LWCNN) that allows automatic extraction features in an end-to-end manner. We have tested LWCNN model in two experiments. In the first experiment, the model was tested with two cases' original and enhancement datasets 1. It achieved 95 %, 93 %, 99 % and 98 % for training and testing accuracy respectively. In the second experiment, the model has been tested with two cases' original and enhancement datasets 2. It achieved 95 %, 91 %, 99 % and 92 % for training and testing accuracy respectively. Our proposed method, which uses various convolutional network to classify screening mammograms achieved exceptional performance when compared to other methods. The findings from these experiments clearly indicate that automatic deep learning techniques can be trained effectively to attain remarkable accuracy across a wide range of mammography datasets. This holds significant promise for improving clinical tools and reducing both false positive and false negative outcomes in screening mammography.

IACP-DFSRA: Identification of Anticancer Peptides Based on a Dual-channel Fusion Strategy of ResCNN and Attention

Anticancer peptides (ACPs) have been widely applied in the treatment of cancer owing to good safety, rational side effects, and high selectivity. However, the number of ACPs that have been experimentally validated is limited as identification of ACPs is extremely expensive. Hence, accurate and cost-effective identification methods for ACPs are urgently needed. In this work, we proposed a deep learning-based model, named iACP-DFSRA, for ACPs identification. Specifically, we adopted two kinds of sequence embedding technologies, ProtBert_BFD pre-training language model and handcrafted features to encode protein sequences. Then, the LightGBM was used for feature selection, and the selected features were input into ResCNN and Attention mechanism, respectively, to extract local and global features. Finally, the concatenate features were deeply fused by using the Attention mechanism to allow key features to be paid more attention to by the model and make predictions by fully connected layer. The results of 10-fold cross-validation demonstrated that the iACP-DFSRA model delivered improved results in most metrics with Sp of 94.15%, Sn of 95.32%, Acc of 94.74% and MCC of 89.48% compared to the latest AACFlow model. Indeed, the iACP-DFSRA model is the only model with Acc > 90% and MCC > 80% on this independent test dataset. Furthermore, we have further demonstrated the superiority of our model on additional datasets. In addition, t-SNE and SHAP interpretation analysis demonstrated that it is crucial to use two channels for feature extraction and use the Attention mechanism for deep fusion, which helps the iACP-DFSRA to predict ACPs more effectively.

A Distance Transformation Deep Forest Framework With Hybrid-Feature Fusion for CXR Image Classification.

Detecting pneumonia, especially coronavirus disease 2019 (COVID-19), from chest X-ray (CXR) images is one of the most effective ways for disease diagnosis and patient triage. The application of deep neural networks (DNNs) for CXR image classification is limited due to the small sample size of the well-curated data. To tackle this problem, this article proposes a distance transformation-based deep forest framework with hybrid-feature fusion (DTDF-HFF) for accurate CXR image classification. In our proposed method, hybrid features of CXR images are extracted in two ways: hand-crafted feature extraction and multigrained scanning. Different types of features are fed into different classifiers in the same layer of the deep forest (DF), and the prediction vector obtained at each layer is transformed to form distance vector based on a self-adaptive scheme. The distance vectors obtained by different classifiers are fused and concatenated with the original features, then input into the corresponding classifier at the next layer. The cascade grows until DTDF-HFF can no longer gain benefits from the new layer. We compare the proposed method with other methods on the public CXR datasets, and the experimental results show that the proposed method can achieve state-of-the art (SOTA) performance. The code will be made publicly available at https://github.com/hongqq/DTDF-HFF.

HemoFuse: multi-feature fusion based on multi-head cross-attention for identification of hemolytic peptides.

Hemolytic peptides are therapeutic peptides that damage red blood cells. However, therapeutic peptides used in medical treatment must exhibit low toxicity to red blood cells to achieve the desired therapeutic effect. Therefore, accurate prediction of the hemolytic activity of therapeutic peptides is essential for the development of peptide therapies. In this study, a multi-feature cross-fusion model, HemoFuse, for hemolytic peptide identification is proposed. The feature vectors of peptide sequences are transformed by word embedding technique and four hand-crafted feature extraction methods. We apply multi-head cross-attention mechanism to hemolytic peptide identification for the first time. It captures the interaction between word embedding features and hand-crafted features by calculating the attention of all positions in them, so that multiple features can be deeply fused. Moreover, we visualize the features obtained by this module to enhance its interpretability. On the comprehensive integrated dataset, HemoFuse achieves ideal results, with ACC, SP, SN, MCC, F1, AUC, and AP of 0.7575, 0.8814, 0.5793, 0.4909, 0.6620, 0.8387, and 0.7118, respectively. Compared with HemoDL proposed by Yang et al., it is 3.32%, 3.89%, 5.93%, 10.6%, 8.17%, 5.88%, and 2.72% higher. Other ablation experiments also prove that our model is reasonable and efficient. The codes and datasets are accessible at https://github.com/z11code/Hemo .

Subword recognition in historical Arabic manuscripts using handcrafted features and deep learning approaches

Subword recognition in historical Arabic manuscripts using handcrafted features and deep learning approaches

Enhancing Automative Safety System for Driver Fatigue Detection by Using Mediapipe’s Framework

Traffic accidents caused by tiredness, exhaustion, and distracted driving are a big concern on a global scale. As per global Road Safety reports drowsy driving was reason for 91,000 road accidents in Tamil Nadu in the year 2023. In this study, I suggest a collective deep learning architecture for automatic driver sleepiness detection while some earlier works are suggested extracting the lips and eyes movements to identify sleepiness, more contemporary computer vision-based systems have performed only somewhat well. This is because they either use extremely large deep learning models with still-poor performance or they use hand-crafted features with traditional methods like K-NN Algotithm and openCV. To overcome such issues in this project I am planing to create a Driver Drowsiness Detection and Alerting System using Mediapipe‘s Face Mesh solution in Python. These systems assess the driver‘s alertness and warn the driver if needed. Continuous driving can be tedious and exhausting. A motorist may get droopy and perhaps nod off due to inactivity. we will create a drowsy driver detection system to address such an issue. For this, we will use Mediapipe‘s Face Mesh solution in python and the Eye Aspect ratio formula using by Navie Bayes Algorithm. My goal is to create a robust and easy-to-use application that detects and alerts users if their eyes are closed for a long time. Keyword: Navie Baye’s, Mediapipe Framework.

Improved vision-only localization method for mobile robots in indoor environments

To solve the problem of mobile robots needing to adjust their pose for accurate operation after reaching the target point in the indoor environment, a localization method based on scene modeling and recognition has been designed. Firstly, the offline scene model is created by both handcrafted feature and semantic feature. Then, the scene recognition and location calculation are performed online based on the offline scene model. To improve the accuracy of recognition and location calculation, this paper proposes a method that integrates both semantic features matching and handcrafted features matching. Based on the results of scene recognition, the accurate location is obtained through metric calculation with 3D information. The experimental results show that the accuracy of scene recognition is over 90%, and the average localization error is less than 1 meter. Experimental results demonstrate that the localization has a better performance after using the proposed improved method.

Specific Emitter Identification Algorithm Based on Time–Frequency Sequence Multimodal Feature Fusion Network

Specific emitter identification is a challenge in the field of radar signal processing. Its aims to extract individual fingerprint features of the signal. However, early works are all designed using either signal or time–frequency image and heavily rely on the calculation of hand-crafted features or complex interactions in high-dimensional feature space. This paper introduces the time–frequency multimodal feature fusion network, a novel architecture based on multimodal feature interaction. Specifically, we designed a time–frequency signal feature encoding module, a wvd image feature encoding module, and a multimodal feature fusion module. Additionally, we propose a feature point filtering mechanism named FMM for signal embedding. Our algorithm demonstrates high performance in comparison with the state-of-the-art mainstream identification methods. The results indicate that our algorithm outperforms others, achieving the highest accuracy, precision, recall, and F1-score, surpassing the second-best by 9.3%, 8.2%, 9.2%, and 9%. Notably, the visual results show that the proposed method aligns with the signal generation mechanism, effectively capturing the distinctive fingerprint features of radar data. This paper establishes a foundational architecture for the subsequent multimodal research in SEI tasks.

ConvexAdam: Self-Configuring Dual-Optimisation-Based 3D Multitask Medical Image Registration.

Registration of medical image data requires methods that can align anatomical structures precisely while applying smooth and plausible transformations. Ideally, these methods should furthermore operate quickly and apply to a wide variety of tasks. Deep learning-based image registration methods usually entail an elaborate learning procedure with the need for extensive training data. However, they often struggle with versatility when aiming to apply the same approach across various anatomical regions and different imaging modalities. In this work, we present a method that extracts semantic or hand-crafted image features and uses a coupled convex optimisation followed by Adam-based instance optimisation for multitask medical image registration. We make use of pre-trained semantic feature extraction models for the individual datasets and combine them with our fast dual optimisation procedure for deformation field computation. Furthermore, we propose a very fast automatic hyperparameter selection procedure that explores many settings and ranks them on validation data to provide a self-configuring image registration framework. With our approach, we can align image data for various tasks with little learning. We conduct experiments on all available Learn2Reg challenge datasets and obtain results that are to be positioned in the upper ranks of the challenge leaderboards. github.com/multimodallearning/convexAdam.

The impact of radiomics libraries and gray level discretization on the discovery of immunotherapy biomarkers in NSCLC patients.

The influence of radiomics pipeline and the grey-level discretization on the discovery of immunotherapy biomarkers is still a poorly understood topic. This study is aimed at identifying robust features by comparing two radiomics libraries and their association with clinical outcomes in non-small cell lung cancer (NSCLC) patients treated with immune checkpoint inhibitors (ICIs). A retrospective cohort of 164 NSCLC patients administered with ICIs was used in this study. Radiomic features were extracted from the pre-treatment CT scans. Univariate models were used to assess the association of radiomics features with progression free survival (PFS), PD-L1 and CD8 cell counts. We also examined the impact of gray-level discretization on feature robustness by evaluating the association of features with clinical endpoints. We extracted 1224, 441 radiomic features using Pyradiomics and RaCat, respectively, out of which 75 were common between them. We showed that the directionality of association between features and clinical endpoints is specific to the radiomic library used. Overall, more Pyradiomics and RaCat features were statistically associated with PFS, and PD-L1, respectively. We found intensity-based features to be more agnostic to the gray-level discretization parameters. Among features that showed significant correlation with PFS with varying gray-level discretization parameters, 45% were intensity-based, compared to PD-L1, and CD8. This study highlights the heterogeneity of radiomics libraries and the gray level discretization parameters that will impact the feature selection and predictive model development. Importantly, our work highlights the significance of selecting features that are agnostic to radiomics libraries for clinical translation. Our study emphasizes the need to select stable CT-derived handcrafted features to build immunotherapy biomarkers, which is a necessary precursor for multi-institutional validation of imaging biomarkers.

Diagnose Alzheimer’s disease and mild cognitive impairment using deep CascadeNet and handcrafted features from EEG signals

Alzheimer’s disease (AD) is the most prevalent clinically diagnosed neurodegenerative disorder. Early detection of mild cognitive impairment (MCI) is crucial for implementing effective interventions and potentially preventing further cognitive decline. Due to its efficiency, the electroencephalogram (EEG) is a promising tool for AD diagnosis. This paper proposes a computer-aided diagnostic model for identifying AD using EEG data analysis. The proposed approach comprises two key steps: signal processing and classification. Initially, the EEG signal is decomposed into sub-bands using the discrete wavelet transform (DWT), followed by windowing for data augmentation. Subsequently, an improved CascadeNet model is employed for feature extraction and classification from the windowed EEG data. CascadeNet’s architecture was specifically chosen for its ability to achieve high accuracy in AD and MCI detection, even with relatively small EEG datasets. Moreover, the paper explores the potential of deep Cascade learning for AD prediction. The effectiveness of the proposed strategy is evaluated using metrics such as F-measures, specificity, sensitivity, and overall detection accuracy. The developed model attains impressive accuracies of 98.84 % and 97.78 % on the Figshare and Brazilian datasets, respectively, significantly outperforming existing methods. To validate the model’s efficacy, it was applied to the Figshare and Brazilian datasets encompassing control (CO), MCI, and AD classes. The results confirm the proposed method as a valuable tool for identifying potential biomarkers aiding in the clinical diagnosis of AD. Notably, unlike previous methods limited to two-class identification, this technique effectively distinguishes between MCI and AD.

Development and validation of a CT-based deep learning radiomics signature to predict lymph node metastasis in oropharyngeal squamous cell carcinoma: a multicenter study.

Lymph node metastasis (LNM) is a pivotal determinant that influences the treatment strategies and prognosis for oropharyngeal squamous cell carcinoma (OPSCC) patients. This study aims to establish and verify a deep learning (DL) radiomics model for the prediction of LNM in OPSCCs using contrast-enhanced computed tomography (CECT). A retrospective analysis included 279 OPSCC patients from three institutions. CECT images were used for handcrafted (HCR) and DL feature extraction. Dimensionality reduction for HCR features used recursive feature elimination and least absolute shrinkage and selection operator algorithms, whereas DL feature dimensionality reduction used variance-threshold and recursive feature elimination algorithms. Radiomics signatures were constructed using support vector machine, decision tree, random forest, k-nearest neighbor, gaussian naive bayes classifiers and light gradient boosting machine. A combined model was then constructed using the screened DL, HCR, and clinical features. The area under the receiver operating characteristic curve (AUC) served to quantify the model's performance, and calibration curves were utilized to assess its calibration. The combined model exhibited robust performance, achieving AUC values of 0.909 (95% CI: 0.861-0.957) in the training cohort, 0.884 (95% CI: 0.800-0.968) in the internal validation cohort, and 0.865 (95% CI: 0.791-0.939) in the external validation cohort. It outperformed both the clinical model and best-performing radiomics model. Moreover, calibration was deemed satisfactory. The combined model based on CECT demonstrates the potential to predict LNM in OPSCCs preoperatively, offering a valuable tool for more precise and tailored treatment strategies. This study presents a novel combined model integrating clinical factors with deep learning radiomics, significantly enhancing preoperative LNM prediction in OPSCC.