Combined Classification Method Research Articles

Application of time series analysis to classify therapeutic breathing patterns

ObjectiveCompare various methods for measuring time series similarity in order to classify referenced therapeutic breathing patterns (BP) used in respiratory disorder rehabilitation. MethodsThis experimental study involved the collection of respiratory signals during specified breathing exercises conducted with healthy volunteers. The study employed a screening phase using a k-NN classifier and eight distance measurement methods, including Minkowski Distance, Dynamic Time Warping-DTW (including FastDTW and constrained-cDTW variations), Longest Common Subsequence-LCSS, Edit Distance on Real Sequences-EDR, Time Warp Edit Distance-TWEED, and Minimum Jump Costs-MJC. Two distinct approaches were employed for classifying therapeutic BP based on time series similarity: (1) using the k-Shape algorithm for clustering, and 2) integrating methods to represent therapeutic BP and classify test curves using the most relevant measurement methods obtained from the first approach. ResultsAmong the two tested approaches, the combination of the cDTW algorithm and Minkowski distance (p = 2), using the 1-NN classifier, achieved the highest scores in this study, closely matching the metrics obtained from visual inspection conducted by human evaluators. ConclusionThe use of combined classification methods in the analysis of flow curves referring to therapeutic breathing patterns improves the classification results, with metrics closely aligned with those obtained through visual evaluation conducted by individuals. SignificanceTime series analysis methods proved to be sensitive to classify respiratory flow curves equivalent to therapeutic breathing patterns used in respiratory disorder rehabilitation. This methodology can be used to monitor respiratory curves in new applications and implementation in devices for evaluating and treating the ventilatory pattern.

Prommis Trial Prospectively Demonstrates the Efficacy of SKY92 Risk Stratification in Newly Diagnosed Multiple Myeloma Patients

Introduction: Multiple myeloma (MM) is a complex hematologic malignancy characterized by genetic instability, variable survival rates, and diverse treatment responses. SKY92 gene expression profiling (GEP) is a valuable tool in risk stratifying patients into high-risk and standard-risk groups for progression and survival. The PRospective Observational Multiple Myeloma Impact Study (PROMMIS; NCT02911571) trial, a prospective US multicenter study, was designed to validate the prognostic performance of SKY92 through real-world data. Methods: The GEP of 251 newly diagnosed MM patients was assessed using the SKY92 assay (SkylineDx, San Diego), which involved analyzing RNA extracted from CD138-positive plasma cells (≥80% purity) isolated through immunomagnetic separation. The test was performed at the enrolling institute, or by sending the samples into a central lab. The standard method for risk stratification by R-ISS stage was integrated with SKY92 into three distinct groups: Low-Risk (LR), defined by SKY92 standard-risk (SR) combined with R-ISS I; Intermediate Risk (IR), as SKY92 SR with R-ISS II/III, or SKY92 high-risk with R-ISS I; and High-Risk as SKY92 high-risk with R-ISS II/III. Progression-free survival (PFS) and overall survival (OS) were available for 221 patients, measured from the date of diagnosis of the patients. The median follow-up at the time of analysis is 38 months. Survival analyses were performed using Cox proportional hazards model. Results: Upon analysis, 179 patients (71.3%) were classified as SKY92 standard-risk, and 72 patients (28.7%) as SKY92 high-risk. Notably, the percentage of high-risk patients in this study was slightly higher than that observed in previous retrospective analyses (15-25%). Furthermore, the survival analysis revealed significant differences in both PFS (HR: 2.1, 95%CI 1.4-3.1, p<0.001) and OS (HR: 3.7, 95%CI 1.8-7.6, p<0.001) between the two risk categories. The median PFS from the time of the diagnosis for SKY92 standard-risk patients was 50 months, whereas for SKY92 high-risk patients, it was 26 months. The median OS was not reached within the 60-month timeframe for both the standard- and high-risk groups. R-ISS could be determined for 204 patients (81%), of which 29.9% (n=61) were classified as stage I, 59.3% (n=121) as stage II, and 10.8% (n=22) as stage III. We integrated R-ISS staging with SKY92 classification for these patients. The combination of these two classification systems yielded the following stratification ( Figure 1): LR n=53 (27%), IR n=96 (47%), and High-Risk n=55 (26%). Consistent with previous studies, R-ISS and SKY92 hold independent prognostic value: in particular, we found that this combined classification method identified a higher number of patients as high-risk compared to the R-ISS annotation (55 [27%] vs. 22 [11%]). The high-risk group exhibited significantly shorter OS and PFS compared to the LR group (OS: HR:18.7, 95% CI 2.5-140, p=0.004 and PFS: HR:2.6, 95% CI 1.4-4.7, p=0.002). Conclusions: This study represents the first US-based multicenter prospective evaluation of the GEP SKY92 in a real-world clinical setting. The results obtained align with the retrospective validations of SKY92, further confirming its reliability as a prognostic marker for MM patients. In addition, we observed that the SKY92 is able to identify a higher number of high-risk patients, with significantly shorter PFS and OS, compared to the conventional R-ISS staging system. Figure 1: PFS and OS based on SKY92+R-ISS risk classification

Model and Data Driven Machine Learning Approach for Analyzing the Vulnerability to Cascading Outages With Random Initial States in Power Systems

In this paper, a hybrid machine learning model is applied to evaluate the relationship between random initial states and the power system’s vulnerability to cascading outages. A cascading outage simulator (CS), which uses off-line AC power flows, is proposed for generating training data. The initial states are randomly selected and the CS model is deployed for each initial state, where power system generation and loads are adjusted dynamically and power flows are redistributed to quantify the vulnerability metric. Furthermore, the proposed hybrid machine learning model deploys a combined Support Vector Machine (SVM) classification and Gradient Boosting Regression (GBR) to improve the learning precision. The classification model is trained by SVM, which divides the data into two categories with and without load shedding. Then, GBR is adopted only for the data with load shedding to determine the relationship between input power outage states and the vulnerability metric. The proposed vulnerability analysis approach is applied to several test systems and the results are analyzed. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —The power system vulnerability can be quantified by cascading outage simulations. However, there are two challenges: i) there are a huge number of possible initial states and we cannot enumerate all these initial states for the cascading outage simulation. Neither can we precisely quantify the bus vulnerability. ii) The cascading outage simulation may be time-consuming for large-scale power systems, which is challenging for the online application. To address the above challenges, we expect to design a machine learning technique to predict the power system vulnerability, which can train the model in an offline way and then use it for the online application. Firstly, since there is not enough operation data from practical power systems, we develop a cascading outage simulator, using off-line AC power flows, for generating synthetic training data. Secondly, we observe that the training precision by directly applying the regression model may be very poor because the output of the machine learning model may take on an uneven distribution concerning input parameters. Thus, we propose a hybrid machine learning model with a combined classification and regression method, where the classification model is employed to remove the data without the load shedding, and the regression model then determines the relationship between input power outage states and the vulnerability metric. The proposed model and method have been tested on several systems including a practical large-scale Polish power system to show the effectiveness.

BGRF: A broad granular random forest algorithm

The random forest is a combined classification method belonging to ensemble learning. The random forest is also an important machine learning algorithm. The random forest is universally applicable to most data sets. However, the random forest is difficult to deal with uncertain data, resulting in poor classification results. To overcome these shortcomings, a broad granular random forest algorithm is proposed by studying the theory of granular computing and the idea of breadth. First, we granulate the breadth of the relationship between the features of the data sets samples and then form a broad granular vector. In addition, the operation rules of the granular vector are defined, and the granular decision tree model is proposed. Finally, the multiple granular decision tree voting method is adopted to obtain the result of the granular random forest. Some experiments are carried out on several UCI data sets, and the results show that the classification performance of the broad granular random forest algorithm is better than that of the traditional random forest algorithm.

Https://journals.scholarpublishing.org/index.php/AIVP/article/view/13922

Land use system is the interface between society and the biophysical environment; it is concern how humans use ecosystems services provide by natural resources.&nbsp; Land-use and land-cover changes have great effects for the environmental and socioeconomic sustainability of rural communities. A capacity for detecting and reporting land use change is critical to evaluating and monitoring trends in natural resource conditions and the effectiveness of public investment in their management. In Mexico, despite ongoing mapping land use - land cover efforts; there remains a need for development of basic datasets providing quantitative and spatial land use/land cover information, mainly a detailed scale. Recent availability of high-resolution satellite remote-sensing images coupling with advances digital image processing techniques offers an improved opportunity to map in a more detailed scale the land use&nbsp; and land cover, which will be an important tool to monitor the environment at local scales. It is the objective of this study, conduct a detailed mapping of land use-land cover in&nbsp; Usumacinta watershed, using high-resolution satellite images and analyze&nbsp; changes occurred when&nbsp; is compared with&nbsp; National Institute of Statistic and geography map - Series VII,2016. The basis for the land use and vegetation cover presented in this study is the digital land cover and land use classification elaborated with SPOT imagery, spatial resolution 1.5 meters, using &nbsp;an object-based classification method&nbsp; in combination with an image classification based on expert knowledge. The resulting map was &nbsp;updated using Sentinel images&nbsp; 2021 and 2022 , by means of on screen digital interpretation. The result is a detail map of land use land cover, where new classes were generated as association and consociation units. As a conclusion is established that the resulting map is more stable in time, and it can be used as indicative to implement restoration measures on natural vegetation and as input to calculate a more precise estimations of agriculture statistic’s. All of this aspects can guide the design and implementation of public policies about land use.

An overview of methods of left and right foot motor imagery based on Tikhonov regularisation common spatial pattern.

The motor imagery brain-computer interface (MI-BCI) provides an interactive control channel for spinal cord injury patients. However, the limitations of feature extraction algorithms may lead to low accuracy and instability in decoding electroencephalogram (EEG) signals. In this study, we examined the classification performance of an MI-BCI system by focusing on the distinction of the left and right foot kinaesthetic motor imagery tasks in five subjects. Feature extraction was performed using the common space pattern (CSP) and the Tikhonov regularisation CSP (TRCSP) spatial filters. TRCSP overcomes the CSP problems of noise sensitivity and overfitting. Moreover, support vector machine (SVM) and linear discriminant analysis (LDA) were used for classification and recognition. We constructed four combined classification methods (TRCSP-SVM, TRCSP-LDA, CSP-SVM, and CSP-LDA) and evaluated them by comparing their accuracies, kappa coefficients, and receiver operating characteristic (ROC) curves. The results showed that the TRCSP-SVM method performed significantly better than others (average accuracy 97%, average kappa coefficient 0.91, and average area under ROC curve (AUC) 0.98). Using TRCSP instead of standard CSP improved accuracy by up to 10%. This study provides insights into the classification of EEG signals. The results of this study can aid lower limb MI-BCI systems in rehabilitation training.

Offshore ship recognition based on center frequency projection of improved EMD and KNN algorithm

Ship-radiated underwater noises are essential in marine target identification. However, it is very difficult to extract valuable target information from ocean ambient noises. Here, a combined classification method is proposed based on the improved empirical mode decomposition to process the underwater signals of offshore ships. The signals are separated into a series of samples with a signal duration of 100 ms and a set of intrinsic mode functions (IMFs) is generated for each sample. The feature extraction of each IMF is realized by defining and calculating its central frequency and energy intensity. Then, the central frequency of each IMF is projected into a three-dimensional space according to the three top-ranked IMFs based on energy intensity, and the K nearest neighbor (KNN) algorithm is employed to identify the different offshore ships. It was observed that the proposed method can efficiently distinguish the three ships when 80 % of the samples are randomly selected as the training set and the remaining 20 % are used as the test set. The overall recognition rate after 1000 random tests was greater than 95 % for the short-time signals of approximately 100 ms and completely unknown types of ships, which indicates that the proposed method is applicable to offshore ship classification and identification.

ProtRe-CN: Protein Remote Homology Detection by Combining Classification Methods and Network Methods via Learning to Rank.

Protein remote homology detection is one of fundamental research tasks for downstream analysis (i.e., protein structure and function prediction). Many advanced methods are proposed from different views with complementary detection ability, such as the classification method, the network method, and the ranking method. A framework integrating these heterogeneous methods is urgently desired to reduce the false positive rate and predictive bias. We propose a novel ranking method called ProtRe-CN by fusing the classification methods and network methods via Learning to Rank. Experimental results on the benchmark dataset and the independent dataset show that ProtRe-CN outperforms other existing state-of-the-art predictors. ProtRe-CN improves the detective performance via correcting the false positives in the ranking list by combining the heterogeneous methods. The web server of ProtRe-CN can be accessed at http://bliulab.net/ProtRe-CN.

MFOC-CliqueNet: A CliqueNet-Based Optimal Combination of Multidimensional Features Classification Method for Large-Scale Laser Point Clouds

As large-scale laser 3D point clouds data contains massive and complex data, it faces great challenges in the automatic intelligent processing and classification of large-scale 3D point clouds. Aiming at the problem that 3D point clouds in complex scenes are self-occluded or occluded, which could reduce the object classification accuracy, we propose a multidimension feature optimal combination classification method named MFOC-CliqueNet based on CliqueNet for large-scale laser point clouds. The optimal combination matrix of multidimension features is constructed by extracting the three-dimensional features and multidirectional two-dimension features of 3D point cloud. This is the first time that multidimensional optimal combination features are introduced into cyclic convolutional networks CliqueNet. It is important for large-scale 3D point cloud classification. The experimental results show that the MFOC-CliqueNet framework can realize the latest level with fewer parameters. The experiments on the Large-Scale Scene Point Cloud Oakland dataset show that the classification accuracy of our method is 98.9%, which is better than other classification algorithms mentioned in this paper.

The Identification of ECG Signals Using Wavelet Transform and WOA-PNN.

Electrocardiogram (ECG) signal identification technology is rapidly replacing traditional fingerprint, face, iris and other recognition technologies, avoiding the vulnerability of traditional recognition technologies. This paper proposes an ECG signal identification method based on the wavelet transform algorithm and the probabilistic neural network by whale optimization algorithm (WOA-PNN). Firstly, Q, R and S waves are detected by wavelet transform, and the P and T waves are detected by local windowed wavelet transform. The characteristic values are constructed by the detected time points, and the ECG data dimension is smaller than that of the non-reference detection. Secondly, combined with the probabilistic neural network, the mean impact value algorithm is used to screen the characteristic values, the characteristic values with low influence are eliminated, and the input and complexity of the model are simplified. Finally, a WOA-PNN combined classification method is proposed to intelligently optimize the hyper parameters in the probabilistic neural network algorithm to improve the model accuracy. According to the simulation verification on three databases, ECG-ID, MIT-BIH Normal Sinus Rhythm and MIT-BIH Arrhythmia, the identification accuracy of a single ECG cycle is 96.97%, and the identification accuracy of three ECG cycles is 99.43%.

Real-time evaluation method of flight mission load based on sensitivity analysis of physiological factors

As the complexity of flight missions continues to increase, sending a timely warning or providing assistance to pilots helps to reduce the probability of operational errors and flight accidents. Monitoring pilots’ physiological data, real-time evaluation of mission load is a feasible technical way to achieve this. In this paper, a set of flight tasks including aircraft control, human-computer interaction and mental arithmetic tests are designed to simulate five mission loads at different flight difficulty levels. A sensitivity analysis method based on a comprehensive test is proposed to select a set of sensitive physiological factors. Then, based on the SVM hierarchical combination classification method, the pilot mission load real-time evaluation model is established. The test results show significant differences in EMG, respiration rate (abdomen), heart rate, blood oxygen saturation, pupil area, fixation duration, number of fixations, and saccades. The high accuracy obtained from experiments proved that the proposed real-time evaluation model is applicable to meet the requirements of real working environments. The findings can provide methodological references for mission load evaluation research in other fields.

Automated classification of heat sources detected using SWIR remote sensing

The potential of shortwave infrared (SWIR) remote sensing to detect hotspots has been investigated using satellite data for decades. The hotspots detected by satellite SWIR sensors include very high-temperature heat sources such as wildfires, volcanoes, industrial activity, or open burning. This study proposes an automated classification method of heat source detected utilizing Landsat 8 and Sentinel-2 data. We created training data of heat sources via visual inspection of hotspots detected by Landsat 8. A scheme to classify heat sources for daytime data was developed by combining classification methods based on a Convolutional Neural Network (CNN) algorithm utilizing spatial features and a decision tree algorithm based on thematic land-cover information and our time series detection record. Validation work using 10,959 classification results corresponding to hotspots acquired from May 2017 to July 2019 indicated that the two classification results were in 79.7% agreement. For hotspots where the two classification schemes agreed, the classification was 97.9% accurate. Even when the results of the two classification schemes conflicted, either was correct in 73% of the samples. To improve the accuracy, the heat source category was re-allocated to the most probable category corresponding to the combination of the results from the two methods. Integrating the two approaches achieved an overall accuracy of 92.8%. In contrast, the overall accuracy for heat source classification during nighttime reached 79.3% because only the decision tree-based classification was applicable to limited available data. Comparison with the Visible Infrared Imaging Radiometer Suite (VIIRS) fire product revealed that, despite the limited data acquisition frequency of Landsat 8, regional tendencies in hotspot occurrence were qualitatively appropriate for an annual period on a global scale.

Hyperspectral image spectral-spatial classification via weighted Laplacian smoothing constraint-based sparse representation.

As a powerful tool in hyperspectral image (HSI) classification, sparse representation has gained much attention in recent years owing to its detailed representation of features. In particular, the results of the joint use of spatial and spectral information has been widely applied to HSI classification. However, dealing with the spatial relationship between pixels is a nontrivial task. This paper proposes a new spatial-spectral combined classification method that considers the boundaries of adjacent features in the HSI. Based on the proposed method, a smoothing-constraint Laplacian vector is constructed, which consists of the interest pixel and its four nearest neighbors through their weighting factor. Then, a novel large-block sparse dictionary is developed for simultaneous orthogonal matching pursuit. Our proposed method can obtain a better accuracy of HSI classification on three real HSI datasets than the existing spectral-spatial HSI classifiers. Finally, the experimental results are presented to verify the effectiveness and superiority of the proposed method.

Assessing Stream Thermal Heterogeneity and Cold-Water Patches from UAV-Based Imagery: A Matter of Classification Methods and Metrics

Understanding stream thermal heterogeneity patterns is crucial to assess and manage river resilience in light of climate change. The dual acquisition of high-resolution thermal infrared (TIR) and red–green–blue-band (RGB) imagery from unmanned aerial vehicles (UAVs) allows for the identification and characterization of thermally differentiated patches (e.g., cold-water patches—CWPs). However, a lack of harmonized CWP classification metrics (patch size and temperature thresholds) makes comparisons across studies almost impossible. Based on an existing dual UAV imagery dataset (River Ovens, Australia), we present a semi-automatic supervised approach to classify key riverscape habitats and associated thermal properties at a pixel-scale accuracy, based on spectral properties. We selected five morphologically representative reaches to (i) illustrate and test our combined classification and thermal heterogeneity assessment method, (ii) assess the changes in CWP numbers and distribution with different metric definitions, and (iii) model how climatic predictions will affect thermal habitat suitability and connectivity of a cold-adapted fish species. Our method was successfully tested, showing mean thermal differences between shaded and sun-exposed fluvial mesohabitats of up to 0.62 °C. CWP metric definitions substantially changed the number and distance between identified CWPs, and they were strongly dependent on reach morphology. Warmer scenarios illustrated a decrease in suitable fish habitats, but reach-scale morphological complexity helped sustain such habitats. Overall, this study demonstrates the importance of method and metric definitions to enable spatio-temporal comparisons between stream thermal heterogeneity studies.

A combination classification method based on Ripper and Adaboost

With the growing demand for data analysis, machine learning technology has been widely used in many applications, such as mass data summarising rules, predicting behaviours and dividing characteristics. The Ripper algorithm presents better pruning and stopping criteria than the traditional decision tree algorithm (C4.5), while its error rate less than or equal to C4.5 by O(nlog2n) time complexity. As a result of that, Ripper can maintain high efficiency even on the massive dataset which contains lots of noise. Adaboost is one of iterative algorithms, which combines a group of weak classifiers together to set up a strong classifier. In order to improve the accuracy of Ripper classification algorithm and reduce the computational complexity, this paper proposes a Ripper-Adaboost combined classification method (Ripper-ADB). The experiment result shows Ripper-ADB could improve the classifier and get higher classification accuracy than decision tree and SVM.

Combined Non-parametric and Parametric Classification Method Depending on Normality of PDF of Training Samples

Classification method with combined nonparametric and parametric classifications which depends on the normality of Probability Density Function of training samples is proposed. The proposed classification method is also based on spatial information for high spatial resolution of satellite based optical sensor images is proposed. Also, a classification method which takes into account not only spectral but also spatial features for LANDSAT-4 and 5 Thematic Mapper (TM) data is proposed. Treatment of the spatial-spectral variability existing within a region is more important for such high spatial resolution of satellite imagery data. Standard deviations in small cells, such as 2x2, 3x3 and 4x4 pixels, were used as measures to represent the spatial-spectral variabilities. This information can be used together with conventional spectral features in a unified way, for the traditional classifier such as the pixelwise Maximum Likelihood Decision Rule (MLHDR). The classification performance of new clear cuts and alpine meadows which are very close in spectral space characteristics and difficult to distinguish them by conventional methods are focused. Through experiments, it is found that there is a substantial improvement in overall classification accuracy for TM forestry data. The Probability of Correct Classification (PCC) for the new clear cuts and the alpine meadows classes rose by 7% to 97% correct. The confusion between alpine meadows and new clear cuts was reduced from 9% to 3%.

A combination classification method based on Ripper and Adaboost

A combination classification method based on Ripper and Adaboost

Fuzzy characterization and classification of bacteria species detected at single-cell level by surface-enhanced Raman scattering

Advanced chemometric methods, such as fuzzy c-means, a semi-supervised clustering method, and fuzzy linear discriminant analysis (FLDA), a new robust supervised classification method in combination with principal component analysis (PCA), namely PCA-FLDA, have been successfully applied for characterization and classification of bacterial species detected at single-cell level by surface-enhanced Raman scattering (SERS) spectroscopy. SERS spectra of three species (S. aureus, E. faecalis and P. aeruginosa) were recorded in an original fashion, using in situ laser induced silver spot as metallic substrate. The detection process of bacteria was isolated inside a hermetically sealed in-house built microfluidic device, connected to a syringe pump for injecting the analytes and a portable Raman spectrometer as detection tool. The obtained results (fuzzy partitions) and spectra of the prototypes (robust fuzzy spectra mean corresponding to each fuzzy partition) clearly demonstrated the efficiency and information power of the advanced fuzzy methods in bacteria characterization and classification based on SERS spectra, and allowed a rationale assigning to a specific group. Also, this powerful detection and classification methodology generates the premises for future investigations of Raman and other spectroscopic data obtained for various samples.

Land cover change in Tashkent province during 1992 – 2018

In this research, a specific period in time (from 1992 – 2018) with 8 -10 years’ timescales of Landsat satellite images were used for land cover classification. To detect land cover change combined classification method with clustering study area into 120 classes and further validation by Google Earth was performed. Our findings show significant land cover changes during the whole study period. Especially this related to grassland/scrubland/meadow/agriculture class, which was around 7000 km2 in 1992, and it dramatically fell to 3000 km2 in 2008 and reached 6000 km2 after ten years. Forest/parks/woodlands land cover class shows a tendency to increase and consisted of its maximum area 4000 km2, but then rapidly fell to 1600 km2 within the next ten years. Glaciers, in contrast, covered area of 1825 km2, but then starts steadily decline and finally reaches only 263 km2 in 2018. Bare soil and barren land starts from 3111 km2 and expands to 5640 km2 area in 2018. The same pattern shows urban/asphalt/rocks class started from 1089 km2 and ends with 2677 km2. Only water bodies do not represent any significant change in Tashkent province during the study period.

Diagnosis and treatment of congenital nasal dermoid and sinus cysts in 11 infants: A consort compliant study.

There have been few studies on congenital nasal dermoid and sinus cysts (NDSCs) in infants.This study was performed to obtain clinical data for the diagnosis and treatment of NDSCs in infants.We performed a retrospective analysis of 11 infants admitted with NDSCs between 2014 and 2019. Patient demographics, lesion site, preoperative radiological findings, surgical technique, intraoperative findings, and postoperative sequelae were analyzed.In total, 11 infants (average age, 19 months; lowest age, 10 months) were included in this study. All patients presented with a nasal root mass, 2 patients also had nasal tip fistula, and only 1 case had a history of preoperative infection. Preoperative enhanced computed tomography (CT) examination showed nasal surface lesion (type I) in 3, nasal intraosseous (type II) in 5, intracranial epidural (type III) in 2, and intracranial dural (type IV) in 1 patient. The main surgical methods included direct resection with a vertical midline incision (9 patients), vertical incision + transnasal endoscopic resection + skull base repair (1 patient), and transverse incision of the lower margin of the left eyebrow (1 patient). All wounds healed well without serious complications.Using the 4-type classification method in combination with the preoperative CT findings to analyze the extent of NDSC in infants is helpful for formulating the surgical plan. Using vertical incision approach alone or combined with nasal endoscopy for minimally invasive surgery can meet the needs of complete resection and reconstruction.Our results provide clinical data that can help establish standardized criteria for the diagnosis and treatment of NDSCs in infants.