Support Vector Machine Classification Research Articles

Epilepsy lesion localization method based on brain function network.

In the past, the localization of seizure onset zone (SOZ) primarily relied on traditional EEG signal analysis methods. However, due to their limited spatial and temporal resolution, accurately pinpointing neural activity was challenging, thereby restricting their clinical applicability. Compared with traditional EEG signals, SEEG signals have superior spatial and temporal resolution, and can more accurately record neural activity near epileptic foci, making them better suited for studying SOZ. In addition, the traditional EEG signal analysis methods still have limitations, mainly focusing on the analysis of local signal features, while ignoring the complexity and interconnection of the overall brain network. How to more accurately locate SOZ is still not well resolved. The purpose of this study is to develop an effective positioning method for more accurate positioning. To overcome these limitations, this study proposed a model integrating brain functional network analysis with nonlinear dynamics. We utilized weighted phase lag index (WPLI) to construct brain functional network, epilepic network connectivity strength (ENCS) as the feature, and introduced persistence entropy (PE) for feature fusion, subsequently employing support vector machine (SVM) classification. The proposed method was verified on the HUP-iEEG dataset, our solution identified the SOZ with 0.9440 accuracy, 0.9848 precision, 0.8974 recall rate, 0.9340 F1 score and 0.9697 area under the ROC curve across patients, which outperforms the existing approaches. It exhibits a 2.30 percentage point enhancement in localisation accuracy along with a 2.97 percentage points in AUC compared to others. Our method consider the interactions between nodes in brain network connections, as well as the inherent nonlinear and non-stationary properties of neural signals, to be more robust.

Harmonizing the CBCL and SDQ ADHD scores by using linear equating, kernel equating, item response theory and machine learning methods.

A problem that applied researchers and practitioners often face is the fact that different institutions within research consortia use different scales to evaluate the same construct which makes comparison of the results and pooling challenging. In order to meaningfully pool and compare the scores, the scales should be harmonized. The aim of this paper is to use different test equating methods to harmonize the ADHD scores from Child Behavior Checklist (CBCL) and Strengths and Difficulties Questionnaire (SDQ) and to see which method leads to the result. Sample consists of 1551 parent reports of children aged 10-11.5 years from Raine study on both CBCL and SDQ (common persons design). We used linear equating, kernel equating, Item Response Theory (IRT), and the following machine learning methods: regression (linear and ordinal), random forest (regression and classification) and Support Vector Machine (regression and classification). Efficacy of the methods is operationalized in terms of the root-mean-square error (RMSE) of differences between predicted and observed scores in cross-validation. Results showed that with single group design, it is the best to use the methods that use item level information and that treat the outcome as interval measurement level (regression approach).

Residual Mulching Film Detection in Seed Cotton Using Line Laser Imaging

Due to the widespread use of mulching film in cotton planting in China, residual mulching film mixed with machine-picked cotton poses a significant hazard to cotton processing. Detecting residual mulching film in seed cotton has become particularly challenging due to the film’s semi-transparent nature. This study constructed an imaging system combining an area array camera and a line scan camera. A detection scheme was proposed that utilized features from both image types. To simulate online detection, samples were placed on a conveyor belt moving at 0.2 m/s, with line lasers at a wavelength of 650 nm as light sources. For area array images, feature extraction was performed to establish a partial least squares discriminant analysis (PLS-DA) model. For line scan images, texture feature analysis was used to build a support vector machine (SVM) classification model. Subsequently, image features from both cameras were merged to construct an SVM model. Experimental results indicated that detection methods based on area array and line scan images had accuracies of 75% and 79%, respectively, while the feature fusion method achieved an accuracy of 83%. This study demonstrated that the proposed method could effectively improve the accuracy of residual mulching film detection in seed cotton, providing a basis for reducing residual mulching film content during processing.

Classification and Analysis of HIV Neurocognitive MRI Images using Support Vector Machine

Medical imaging has expanded thanks to advances in processing power and advanced image analysis techniques, especially with magnetic resonance imaging (MRI), which offers comprehensive body scans for diagnosis. This work proposes a simple yet efficient method to use a support vector machine (SVM) to classify HIV neurocognitive MRI pictures into normal and pathological categories. The model consists of four steps: data pre-processing, feature extraction, SVM classification, and model evaluation. To separate desired and undesired elements, such as the scalp and skull, pre-processed images were converted from grayscale to colour using support vector machines. The discrete wavelet transform (DWT) was used in the feature extraction stage to extract image properties. Colour moments (CMs) were then used to optimize the feature collection. Afterwards, the SVM classifier was used to determine the ideal feature set to classify images. For example, a dataset is used for training and testing, with a split ratio of 75% to 25% respectively. The experimental results show that the proposed model has a high classification accuracy of 94.4%

Combination of Lexical Resources and Support Vector Machine for Film Sentiment Analysis

Text data generated by internet users holds potentially valuable information that can be researched for new insights. One strategy for obtaining information from a text data set is to classify text into predetermined categories based on existing data. Text classification is an aspect of Text Mining. One of the popular approaches in Text Mining uses the Support Vector Machine (SVM) classification algorithm, which aims to classify text and separate data into different classes. However, in some cases, SVM classification algorithms may face difficulties in understanding the context of the text properly due to unclear wording, varying sentence structures, or a lack of understanding of interpretation. To address this problem, applying SVM classification using lexical resources can be an effective solution. In this research framework, the first step is to obtain data, which in this case is a film review dataset taken from the kaggle.com site. After obtaining the data, the next step is preprocessing. The results of the preprocessing are then divided into 80:20 percentages. The 80% training data is used to search for the form of polarization, and this training data lexicon is used for training the SVM model. Based on the modeling results, the overall model accuracy is around 85%, calculated using the confusion matrix. The precision value, which shows the proportion of correct positive predictions, reached 88%. The precision for negative predictions reached 80%, and for neutral predictions, it reached 0%. These results show that the Lexicon+SVM model has good performance, with an accuracy of 85%.

Precision in practice: Exploring the impact of AI and machine learning on ultrasound guided regional anaesthesia.

Ultrasound Guided Regional Anaesthesia (UGRA) is a commonly utilized practice in both elective and emergency situations during surgical procedures and for pain management. Its benefits include, being non-invasive, cost-effective, readily accessible, and providing the anaesthetist with clear visualization of essential anatomical landmarks, needle progression, and the spread of local anaesthetic. Ultrasonography has been shown to increase success rates for regional anaesthesia and decrease complications. One of the critical steps during UGRA is identifying relevant anatomical structures like nerves or vertebras. However, this aspect can be hindered by external influences such as variations in nerve structure and position, interference from noise, and positional instability. Machine learning is a promising branch of artificial intelligence. It is used to conduct predictive tasks without programming instructions by creating algorithms. Extensive research has been conducted to evaluate the influence of machine learning on innovative anaesthesia methods. In 2023, Lopez et al. published a systematic review on how Artificial Intelligence could positively impact traditional anaesthesia practices.1 Various studies included in the review employed different models to achieve variable targets during the induction of anaesthesia. In one experiment, Alkhatib et al. used Convolutional neural network (CNN) based deep trackers to track the median and sciatic nerve with a surprising accuracy of 0.87.2 Another study employed the same CNN model to locate and discriminate accurate images of sacrum, vertebral levels and intervertebral gaps during percutaneous spinal needle insertion.3 Another study used a different AI model called SVM (support vector machine) classification, image processing, and template matching to locate lumbar level L3-L4 and the ideal puncture site for epidural anaesthesia in real-time. In this experiment, anaesthetists with minimal experience in ultrasonography were able to successfully determine needle puncture sites accurately.4 All these studies utilizing AI models not only yielded beneficial results but also led to significant time savings. Considering this, anaesthesiologists nationwide must integrate effective AI models to improve their clinical practice. This would minimize adverse outcomes in regional anaesthesia and ultimately enhance patient care and satisfaction.

Enhancing microalgae classification accuracy in marine ecosystems through convolutional neural networks and support vector machines

Accurately classifying microalgae species is vital for monitoring marine ecosystems and managing the emergence of marine mucilage, which is crucial for monitoring mucilage phenomena in marine environments. Traditional methods have been inadequate due to time-consuming processes and the need for expert knowledge. The purpose of this article is to employ convolutional neural networks (CNNs) and support vector machines (SVMs) to improve classification accuracy and efficiency. By employing advanced computational techniques, including MobileNet and GoogleNet models, alongside SVM classification, the study demonstrates significant advancements over conventional identification methods. In the classification of a dataset consisting of 7820 images using four different SVM kernel functions, the linear kernel achieved the highest success rate at 98.79 %. It is followed by the RBF kernel at 98.73 %, the polynomial kernel at 97.84 %, and the sigmoid kernel at 97.20 %. This research not only provides a methodological framework for future studies in marine biodiversity monitoring but also highlights the potential for real-time applications in ecological conservation and understanding mucilage dynamics amidst climate change and environmental pollution.

A novel bounded loss framework for support vector machines

This paper introduces a novel bounded loss framework for SVM and SVR. Specifically, using the Pinball loss as an illustration, we devise a novel bounded exponential quantile loss (Leq-loss) for both support vector machine classification and regression tasks. For Leq-loss, it not only enhances the robustness of SVM and SVR against outliers but also improves the robustness of SVM to resampling from a different perspective. Furthermore, EQSVM and EQSVR were constructed based on Leq-loss, and the influence functions and breakdown point lower bounds of their estimators are derived. It is proved that the influence functions are bounded, and the breakdown point lower bounds can reach the highest asymptotic breakdown point of 1/2. Additionally, we demonstrated the robustness of EQSVM to resampling and derived its generalization error bound based on Rademacher complexity. Due to the Leq-loss being non-convex, we can use the concave–convex procedure (CCCP) technique to transform the problem into a series of convex optimization problems and use the ClipDCD algorithm to solve these convex optimization problems. Numerous experiments have been conducted to confirm the effectiveness of the proposed EQSVM and EQSVR.

Diagnostic value of regional homogeneity and fractional amplitude of low-frequency fluctuations in the classification of schizophrenia and bipolar disorders.

Schizophrenia (SCZ) and bipolar disorders (BD) show significant neurobiological and clinical overlap. In this study, we wanted to identify indexes of intrinsic brain activity that could differentiate these disorders. We compared the diagnostic value of the fractional amplitude of low-frequency fluctuations (fALFF) and regional homogeneity (ReHo) estimated from resting-state functional magnetic resonance imaging in a support vector machine classification of 59 healthy controls (HC), 40 individuals with SCZ, and 43 individuals with BD type I. The best performance, measured by balanced accuracy (BAC) for binary classification relative to HC was achieved by a stacking model (87.4% and 90.6% for SCZ and BD, respectively), with ReHo performing better than fALFF, both in SCZ (86.2% vs. 79.4%) and BD (89.9% vs. 76.9%). BD were better differentiated from HC by fronto-temporal ReHo and striato-temporo-thalamic fALFF. SCZ were better classified from HC using fronto-temporal-cerebellar ReHo and insulo-tempo-parietal-cerebellar fALFF. In conclusion, we provided evidence of widespread aberrancies of spontaneous activity and local connectivity in SCZ and BD, demonstrating that ReHo features exhibited superior discriminatory power compared to fALFF and achieved higher classification accuracies. Our results support the complementarity of these measures in the classification of SCZ and BD and suggest the potential for multivariate integration to improve diagnostic precision.

Discrimination of Fe-Ni-Laterites from Bauxites Using a Novel Support Vector Machines-Based Methodology on Sentinel-2 Data

Currently, the global shift towards green energy is at the forefront of efforts introducing a new era, thus rendering exploration for critical raw materials essential. To this purpose, the utilization of advanced machine learning methods in remote sensing has emerged as a rapid and cost-effective approach. This study proposes a new methodology, utilizing Sentinel-2 satellite data, to distinguish ferronickel (Fe-Ni-) laterite from bauxite across pre-mining, mining, and post-mining occurrences worldwide. Both ores contain mineral raw materials such as nickel, iron, cobalt, and alumina and their discrimination is generally macroscopically challenging, especially when their locations are often in geographical proximity. The proposed method is based on Support Vector Machines (SVM) classification using spectral signatures of known Fe-Ni-laterite and bauxite-bearing pixels in Greece, Cuba, and Jamaica. The highest classification accuracies are obtained by combining 𝑏12 with 𝑏6 or 𝑏7 spectral bands. Comparisons with specific ore mineralogies show that 𝑏6 and 𝑏7 are strongly linked to the ferric phase, while 𝑏12 is mainly associated with the argillic mineralogies, the latter probably being the key discriminating factor between the two ores. From laboratory chemical analyses, we also establish that 𝑏12 and 𝑏6 or 𝑏7 are strongly associated with Al2O3 and Fe2O3 content correspondingly. The proposed method is accurate, it has reduced prospection costs, and it can facilitate the initial screening of broad areas by automatically characterizing whether an ore is bauxite or Fe-Ni-laterite. This underscores the methodology’s significance in ore differentiation and exploration within the context of green energy endeavors.

Altered Intrinsic Brain Activity in Ischemic Stroke Patients Assessed Using the Percent Amplitude of a Fluctuation Method.

Ischemic stroke is a vascular disease that may cause cognitive and behavioral abnormalities. This study aims to assess abnormal brain function in ischemic stroke patients using the percent amplitude of fluctuation (PerAF) method and further explore the feasibility of PerAF as an imaging biomarker for investigating ischemic stroke pathophysiology mechanisms. Sixteen ischemic stroke patients and 22 healthy controls (HCs) underwent resting state functional magnetic resonance imaging (rs-fMRI) scanning, and the resulting data were analyzed using PerAF. Then a correlation analysis was conducted between PerAF values and Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA) scores. Finally, the abnormal PerAF values were extracted and defined as features for support vector machine (SVM) analysis. Compared with HCs, ischemic stroke patients showed decreased PerAF in the bilateral cuneus, left middle frontal gyrus, precuneus and right inferior temporal gyrus, and increased PerAF in the bilateral orbital part of middle frontal gyrus and right orbital part of superior frontal gyrus. Correlation analyses revealed that PerAF values in the left orbital part of middle frontal gyrus was negatively correlated with the MoCA scores. The SVM classification of the PerAF values achieved an area under the curve (AUC) of 0.98 and an accuracy of 94.74%. Abnormal brain function has been found among ischemic stroke patients, which may be correlated with visual impairment, attention deficits, and dysregulation of negative emotions following a stroke. Our findings may support the potential of PerAF as a sensitive biomarker for investigating the underlying mechanisms of ischemic stroke.

A proposed two-level classification approach for forensic detection of diesel adulteration using NMR spectroscopy and machine learning.

Adulteration of diesel fuel poses a major concern in most developing countries including Ghana despite the many regulatory schemes adopted. The solvent tracer analysis approach currently used in Ghana has over the years suffered several limitations which affect the overall implementation of the scheme. There is therefore a need for alternative or supplementary tools to help detect adulteration of automotive fuel. Herein we describe a two-level classification method that combines NMR spectroscopy and machine learning algorithms to detect adulteration in diesel fuel. The training sets used in training the machine learning algorithms contained 20-40% w/w adulterant, a level typically found in Ghana. At the first level, a classification model is built to classify diesel samples as neat or adulterated. Adulterated samples are passed on to the second stage where a second classification model identifies the type of adulterant (kerosene, naphtha, or premix) present. Samples were analyzed by 1H NMR spectroscopy and the data obtained were used to build and validate support vector machine (SVM) classification models at both levels. At level 1, the SVM model classified all 200 samples with only 2.5% classification errors after validation. The level 2 classification model developed had no classification errors for kerosene and premix in diesel. However, 2.5% classification error was recorded for samples adulterated with naphtha. Despite the great performance of the proposed schemes, it showed significantly erratic predictions with adulterant levels below 20% w/w as the training sets for both models contained adulterants above 20% w/w. The proposed method, nevertheless, proved to be a potential tool that could serve as an alternative to the marking system in Ghana for the fast detection of adulterants in diesel.

Associations between Radiomics and Genomics in Non-Small Cell Lung Cancer Utilizing Computed Tomography and Next-Generation Sequencing: An Exploratory Study

Background: Radiomics, an evolving paradigm in medical imaging, involves the quantitative analysis of tumor features and demonstrates promise in predicting treatment responses and outcomes. This study aims to investigate the predictive capacity of radiomics for genetic alterations in non-small cell lung cancer (NSCLC). Methods: This exploratory, observational study integrated radiomic perspectives using computed tomography (CT) and genomic perspectives through next-generation sequencing (NGS) applied to liquid biopsies. Associations between radiomic features and genetic mutations were established using the Area Under the Receiver Operating Characteristic curve (AUC-ROC). Machine learning techniques, including Support Vector Machine (SVM) classification, aim to predict genetic mutations based on radiomic features. The prognostic impact of selected gene variants was assessed using Kaplan–Meier curves and Log-rank tests. Results: Sixty-six patients underwent screening, with fifty-seven being comprehensively characterized radiomically and genomically. Predominantly males (68.4%), adenocarcinoma was the prevalent histological type (73.7%). Disease staging is distributed across I/II (38.6%), III (31.6%), and IV (29.8%). Significant correlations were identified with mutations of ROS1 p.Thr145Pro (shape_Sphericity), ROS1 p.Arg167Gln (glszm_ZoneEntropy, firstorder_TotalEnergy), ROS1 p.Asp2213Asn (glszm_GrayLevelVariance, firstorder_RootMeanSquared), and ALK p.Asp1529Glu (glcm_Imc1). Patients with the ROS1 p.Thr145Pro variant demonstrated markedly shorter median survival compared to the wild-type group (9.7 months vs. not reached, p = 0.0143; HR: 5.35; 95% CI: 1.39–20.48). Conclusions: The exploration of the intersection between radiomics and cancer genetics in NSCLC is not only feasible but also holds the potential to improve genetic predictions and enhance prognostic accuracy.

Identification of Bloodstains by Species Using Extreme Learning Machine and Hyperspectral Imaging Technology.

How to identify bloodstains and obtain some potential evidence is of great significance for solving criminal cases. First, the spectral data of different species of bloodstain samples (human blood and animal blood) were acquired by using a hyperspectral imager. Then, an extreme learning machine (ELM) algorithm was used to build the training models of different species of bloodstain samples. Meanwhile, two traditional support vector machine and random forest classification algorithms were also compared with the ELM algorithm. The prediction results showed that the precision, sensitivity, specificity, and F1 score of the ELM algorithm were the highest. This indicates that hyperspectral technology, together with an ELM algorithm, could identify bloodstain species rapidly, non-destructively, and accurately. It has provided a new technical reference for bloodstain detection and identification.

Lithological Discrimination of Khyber Range Using Remote Sensing and Machine Learning Algorithms

In this study, the satellite data of ASTER and Landsat 8 OLI were used for the discrimination of lithological units covering the Khyber range. Of the 24 tested band combinations, the most suitable include 632 and 468 of ASTER and 754 and 147 of OLI in the RGB sequence. The data were also tested with two conventional machine learning algorithms (MLAs), namely maximum likelihood classification (MLC) and support vector machine (SVM), for lithological mapping. Principal component analysis (PCA), minimum noise fraction (MNF), band ratios, and color composites in combination with available lithological maps and field data were utilized for training sample collection for the MLC and SVM models to classify the lithological units. The accuracy assessment of SVM and MLC was performed using a confusion matrix, which revealed a higher accuracy of 74.8419% and 72.1217% for ASTER and an accuracy of 58.4833% and 60.0257% for OLI, respectively. The results indicate that ASTER imagery is more suitable for lithological discrimination in the study area due to its high spectral resolution in the VNIR to SWIR range. The experiment revealed that the SVM classification offered the highest overall accuracy of nearly 75% and the kappa coefficient value of 0.7 on ASTER data. This demonstrates the effectiveness of SVM classification in exploring lithological mapping in dry to semi-arid regions.

WORLD GREENHOUSE GAS EMISSION CLASSIFICATION USING SUPPORT VECTOR MACHINE (SVM) METHOD

The phenomenon of Heatwaves has struck several countries across the globe due to climate change. This climate change has led to an increase in greenhouse gas emissions surpassing the limits set by the IPCC Fourth Assessment Report GWPs. This study utilizes the Support Vector Machine (SVM) classification method to identify and categorize greenhouse gas emission data from 1990 to 2020 using four kernels function such as linear, polynomial, radial basis function (RBF), and sigmoid. The SVM method demonstrates excellent performance in constructing classification models with a polynomial kernel function. This is evidenced by high values of training accuracy, testing accuracy, and F1-score, accompanied by short training and testing analysis times. Successively, these values are 97.39%, 97.69%, 96.82%, 0.59 seconds, and 0.22 seconds.

Unveiling Neurophysiological Markers of Consciousness Levels through EEG Exploration

The concept of consciousness levels typically refers to various aspects and tiers related to an individual’s cognition, perception, thinking, and awareness. Although neurophysiological markers have not yet been definitively identified to distinguish between these nuanced levels, this paper introduces a robust marker, the Approximate Entropy (ApEn), which quantifies the complexity of EEG signals to differentiate states of altered consciousness. Utilizing ApEn, we analyze EEG data from the frontal lobe—a region closely associated with consciousness—in states indicative of severely altered conditions, specifically anesthesia, coma, and brain death. To enhance the precision of consciousness levelassessment, we employ a Support Vector Machine (SVM) model, which classifies the states based on EEG complexity measures. This approach not only provides valuable insights into the neural correlations associated with changes in these critical states but also underscores the potential of combining quantitative EEG analysis with machine learning techniques to advance our understanding of consciousness. The findings demonstrate that EEG complexity, when analyzed using ApEn coupled with SVM classification, offers a novel and effective method for assessing and distinguishing between degrees of consciousness. This approach promises significant implications for clinical diagnostics and patient monitoring.

Identifying human elephant conflict hotspots through satellite remote sensing and GIS to support conflict mitigation

Human-elephant conflict (HEC) is a significant issue in Sri Lanka and many parts of the world where elephants and humans coexist. To address HEC, this study integrates remote sensing and GIS analysis, focusing on monitoring changes in greenery. The study prepared the latest land cover and land use (LCLU) maps with Sentinel-2 satellite data, correlating them with reported HEC incidents reported in 2021 and 2022 to identify HEC hotspots in two forest-dominated regions of Southeast Sri Lanka. High-resolution sentinel-2 satellite imagery were used to detect areas of human activities and elephant habitats in proximity to each other. Random Forest (RF) and Support Vector Machine (SVM) classification methods were used for LCLU classification. The overall accuracy of the classification was 97.31 and 94.62, and kappa was 0.95 and 0.90 for RF and SVM, respectively. Multi-temporal normalised difference vegetation index (NDVI) analysis provided insights into vegetation health and coverage, offering a clear picture of greenery changes. Monthly changes in vegetation cover readings were quantified using NDVI values derived from MODIS data, identifying suitable regions for elephants to forage frequently. Furthermore, Kernel density estimation identified high-density areas for reported incidents of human and elephant deaths. This process involved assigning weight to conflict incidents within a 5 km radius, considering the proximity to the forest, and evaluating greenery changes using NDVI values, revealing varying levels of HEC risk, ranging from very high to low. The LCLU map, created using the RF classifier, indicates that all potential HEC hotspots for very high and high HEC risks are closely aligned with forest boundaries. The findings support HEC mitigation strategies through community awareness, HEC hotspots mapping and restoration practices to ensure a sustainable human-elephant coexistence. This method will help policymakers in wildlife conservation to identify high risk HEC zones to support HEC mitigation. In conclusion, this study highlights the potential of integrating remote sensing and GIS techniques in demarcating HEC hotspots in Sri Lanka to support conflict mitigation efforts.

Analysis of gasoline quality by ATR-FTIR spectroscopy with multivariate techniques

In this paper, chemometric methods were used for exploratory analysis, categorization, and quantification of gasoline fuel using Fourier Transform Infrared Spectroscopy (FTIR) data. During exploratory analysis, Principal Component Analysis (PCA) was employed, and to categorise the gasoline samples, Support Vector Machine Classification (SVMC), Linear Discrimination Analysis (LDA), and Partial Least Squares Discriminant Analysis (PLS-DA) were used. The concentration of Benzene, Methyl Tert-butyl Ether (MTBE), and Public Distribution System (PDS) Kerosene were determined using the Partial Least Squares Regression (PLSR), Principal Component Regression (PCR), and Support Vector Machine Regression (SVMR). All of the chemometric models had 100% accuracy and high R-square and RMSEC significance values. The SVM classification techniques were identified as a suitable choice for both classifying oxygenated and adulterated samples among all approaches. Both PLSR and PCR can also be suitable choices for quantification when dealing with high dimensional data. As a result, the research reported in this study has the potential to become a significant tool for improving gasoline sample discrimination and as a resource for quality control in the petroleum industry.

Endovascular Detection of Catheter-Thrombus Contact by Vacuum Excitation.

The objective of this work is to introduce and demonstrate the effectiveness of a novel sensing modality for contact detection between an off-the-shelf aspiration catheter and a thrombus. A custom robotic actuator with a pressure sensor was used to generate an oscillatory vacuum excitation and sense the pressure inside the extracorporeal portion of the catheter. Vacuum pressure profiles and robotic motion data were used to train a support vector machine (SVM) classification model to detect contact between the aspiration catheter tip and a mock thrombus. Validation consisted of benchtop accuracy verification, as well as user study comparison to the current standard of angiographic presentation. Benchtop accuracy of the sensing modality was shown to be 99.67%. The user study demonstrated statistically significant improvement in identifying catheter-thrombus contact compared to the current standard. The odds ratio of successful detection of clot contact was 2.86 (p = 0.03) when using the proposed sensory method compared to without it. The results of this work indicate that the proposed sensing modality can offer intraoperative feedback to interventionalists that can improve their ability to detect contact between the distal tip of a catheter and a thrombus. By offering a relatively low-cost technology that affords off-the-shelf aspiration catheters as clot-detecting sensors, interventionalists can improve the first-pass effect of the mechanical thrombectomy procedure while reducing procedural times and mental burden.