Fisher Discriminant Model Research Articles

Age and flexors as risk factors for cervical radiculopathy: A new machine learning method.

This study aimed to investigate the risk factors for cervical radiculopathy (CR) along with identifying the relationships between age, cervical flexors, and CR. This was a retrospective cohort study, including 60 patients with CR enrolled between December 2018 and June 2020. In this study, we measured C2 to C7 Cobb angle, disc degeneration, endplate degeneration, and morphology of paraspinal muscles and evaluated the value of predictive methods using receiver operating characteristic curves. Next, we established a diagnostic model for CR using Fisher discriminant model and compared different models by calculating the kappa value. Age and cervical flexor factors were used to construct clinical predictive models, which were further evaluated by C-index, receiver operating characteristic curve, calibration curve, and decision curve analysis. Multivariate analysis showed that age and cervical flexors were potential risk factors for CR, while the diagnostic model indicated that both exerted the best diagnostic effect. The obtained diagnostic equation was as follows: y1 = 0.33 × 1 + 10.302 × 2-24.139; y2 = 0.259 × 1 + 13.605 × 2-32.579. Both the C-index and AUC in the training set reached 0.939. Moreover, the C-index and AUC values in the external validation set reached 0.961. We developed 2 models for predicting CR and also confirmed their validity. Age and cervical flexors were considered potential risk factors for CR. Our noninvasive inspection method could provide clinicians with a more potential diagnostic value to detect CR accurately.

A comprehensive strategy integrating metabolomics with DNA barcoding for discovery of combinatorial discriminatory quality markers: A case of Cimicifuga foetida and Cimicifuga dahurica

The multiple species characteristics of traditional Chinese medicines (TCMs) are crucial for expanding TCMs sources, meeting the needs of the pharmaceutical industry and ensuring clinical requirements. It’s also one of the significant factors affecting the quality control of TCMs. Systematic differential analysis of original species in TCMs is an important link in achieving comprehensive quality control, ensuring the effectiveness and safety of clinical medication. The study aims to establish a reliable and efficient approach to screen combinatorial discriminatory quality markers for rapid differentiation of original species by metabolomics coupled with DNA barcoding as a case of Cimicifugae Rhizoma. DNA barcoding is used to identify the origin of Cimicifugae Rhizoma. The data-dependent acquisition mode integrated with the computerized intelligent filtering system was established for in-depth characterization of metabolites from Cimicifugae Rhizoma using ultra-high performance liquid chromatography to quadrupole time-of-flight mass spectrometry (UHPLC-Q-TOF-MS). The untargeted metabolomics combined with multivariate statistical analysis was performed to screen and identify the potential combinatorial discriminatory quality markers. Finally, quantitative analysis and predictive model of these markers were employed to validate the feasibility of this strategy to distinguish the original species. Based on the scores of variable importance in projection greater than 1.0 and t-test (p < 0.05) in chemometric analysis, caffeic acid, cimifugin, ferulic acid and isoferulic acid were authenticated as combinatorial discriminatory quality markers for the two original species of Cimicifugae Rhizoma. In addition, the Fisher discriminant model successfully classified 56 batches of Cimicifugae Rhizoma with an accuracy of 94.4 %, showcased the practicality and scientific validity of this method. This study has provided a comprehensive strategy for efficient discrimination of multiple species of medicinal materials.

Determination of Geographic Origin of Turquoise by Combining Laser Ablation Inductively Coupled Plasma Mass Spectrometry Analysis and Chemometrics

Microsampling elemental analysis is widely used for gemstone and mineralogy traceability. Using laser ablation inductively coupled plasma mass spectrometry combined with chemometrics, the contents of 56 elements in turquoise samples from 5 distinct producing areas in 3 nations were measured. An origin identification model for turquoise samples from various producing areas was established through random forest importance analysis, principal component analysis, and linear discriminant analysis. When combined with random forest importance screening, the traceability efficiency of principal component analysis is significantly improved. Moreover, by taking 48 elements as characteristic variables and introducing them into the discriminant model, a Fisher discriminant model for identifying the origin of turquoise was successfully established. The effective element fingerprint information of turquoise species is closely related to the species origin, and the accuracy of cross-validation reaches 99.5%, demonstrating the feasibility of the proposed model for the identification of the origin of turquoise samples.

Postmortem Interval Estimation Using Protein Chip Technology Combined with Multivariate Analysis Methods.

To estimate postmortem interval (PMI) by analyzing the protein changes in skeletal muscle tissues with the protein chip technology combined with multivariate analysis methods. Rats were sacrificed for cervical dislocation and placed at 16 ℃. Water-soluble proteins in skeletal muscles were extracted at 10 time points (0 d, 1 d, 2 d, 3 d, 4 d, 5 d, 6 d, 7 d, 8 d and 9 d) after death. Protein expression profile data with relative molecular mass of 14 000-230 000 were obtained. Principal component analysis (PCA) and orthogonal partial least squares (OPLS) were used for data analysis. Fisher discriminant model and back propagation (BP) neural network model were constructed to classify and preliminarily estimate the PMI. In addition, the protein expression profiles data of human skeletal muscles at different time points after death were collected, and the relationship between them and PMI was analyzed by heat map and cluster analysis. The protein peak of rat skeletal muscle changed with PMI. The result of PCA combined with OPLS discriminant analysis showed statistical significance in groups with different time points (P<0.05) except 6 d, 7 d and 8 d after death. By Fisher discriminant analysis, the accuracy of internal cross-validation was 71.4% and the accuracy of external validation was 66.7%. The BP neural network model classification and preliminary estimation results showed the accuracy of internal cross-validation was 98.2%, and the accuracy of external validation was 95.8%. There was a significant difference in protein expression between 4 d and 25 h after death by the cluster analysis of the human skeletal muscle samples. The protein chip technology can quickly, accurately and repeatedly obtain water-soluble protein expression profiles in rats' and human skeletal muscles with the relative molecular mass of 14 000-230 000 at different time points postmortem. The establishment of multiple PMI estimation models based on multivariate analysis can provide a new idea and method for PMI estimation.

基于系统聚类算法和Fisher判别模型的古代玻璃成分分析与类别鉴定

基于系统聚类算法和Fisher判别模型的古代玻璃成分分析与类别鉴定

Fisher discriminant model based on LASSO logistic regression for computed tomography imaging diagnosis of pelvic rhabdomyosarcoma in children

Computed tomography (CT) has been widely used for the diagnosis of pelvic rhabdomyosarcoma (RMS) in children. However, it is difficult to differentiate pelvic RMS from other pelvic malignancies. This study aimed to analyze and select CT features by using least absolute shrinkage and selection operator (LASSO) logistic regression and established a Fisher discriminant analysis (FDA) model for the quantitative diagnosis of pediatric pelvic RMS. A total of 121 pediatric patients who were diagnosed with pelvic neoplasms were included in this study. The patients were assigned to an RMS group (n = 36) and a non-RMS group (n = 85) according to the pathological results. LASSO logistic regression was used to select characteristic features, and an FDA model was constructed for quantitative diagnosis. Leave-one-out cross-validation and receiver operating characteristic (ROC) curve analysis were used to evaluate the diagnostic ability of the FDA model. Six characteristic variables were selected by LASSO logistic regression, all of which were CT morphological features. Using these CT features, the following diagnostic models were established: (RMS group){G}_{1}=-14.283+6.613{x}_{1}+5.333{x}_{2}+5.753{x}_{3}+12.361{x}_{4}+8.095{x}_{5}-0.715{x}_{6}; (Non-RMS group){G}_{2}=-2.008+3.539{x}_{1}+1.080{x}_{2}+1.154{x}_{3}+2.307{x}_{4}+1.656{x}_{5}+1.380{x}_{6}, where {x}_{1}, {x}_{2}, … and {x}_{6} are lower than normal muscle density (1 = yes; 0 = no), multinodular fusion (1 = yes; 0 = no), enhancement at surrounding blood vessels (1 = yes; 0 = no), heterogeneous progressive centripetal enhancement (1 = yes; 0 = no), ring enhancement (1 = yes; 0 = no), and hemorrhage (1 = yes; 0 = no), respectively. The calculated area under the ROC curve (AUC) of the model was 0.992 (0.982–1.000), with a sensitivity of 94.4%, a specificity of 96.5%, and an accuracy of 95.9%. The calculated sensitivity, specificity and accuracy values were consistent with those from cross-validation. An FDA model based on the CT morphological features of pelvic RMS was established and could provide an easy and efficient method for the diagnosis and differential diagnosis of pelvic RMS in children.

Research on Financial Risk Early Warning in the Online Game Industry Based on Random Forest and Fisher Discriminant Method

With the rapid rise of the online game industry and the intensification of market competition, online game enterprises face different degrees of financial risks. It is necessary to establish an effective financial risk early warning model. Based on the online game industry, this paper constructs an index system from six dimensions, and makes an empirical study on financial risk early warning. First of all, a financial risk early warning model based on the random forest is constructed, and the prediction accuracy is as high as 100%, but the machine learning model has poor explanatory ability. Therefore, according to the index of random forest screening, a financial risk warning model based on the Fisher discriminant method is established. This model optimized the prediction effect of the original fisher's discriminant model and the prediction accuracy reached 87.5%. Finally, suggestions are made from the three aspects of cash flow, solvency and profitability of online game enterprises, and national support policies.

Integrating generalized domain adaptation and Fisher discriminative learning: A unified framework for face recognition with single sample per person

In this paper, an enhanced discriminative feature learning (EDFL) method is proposed to address single sample per person (SSPP) face recognition. With a separate auxiliary dataset, EDFL integrates Fisher discriminative learning and domain adaptation into a unified framework. The separate auxiliary dataset and the gallery/probe dataset are from two different domains (named source and target domains respectively) and have different data distributions. EDFL is modeled to transfer the discriminative knowledge learned from the source domain to the target domain for classification. Since the gallery set with SSPP contains scarce number of samples, it is hard to accurately represent the data distribution of the target domain, which hinders the adaptation effect. To overcome this problem, the generalized domain adaption (GDA) method is proposed to realize good overall domain adaptation when one domain contains limited samples. GDA considers the both global and local domain adaptation effect at the same time. Further, to guarantee that the learned domain adaptation components are optimal for discriminative learning, the domain adaptation and Fisher discriminant model learning are unified into a single framework and an efficient algorithm is designed to optimize them. The effectiveness of the proposed approach is demonstrated by extensive evaluation and comparison with some state-of-the-art methods.

Characterization of Marine Oil Spills by Diagnostic Ratios, Wavelet Coefficients, and Ratio of Nickel to Vanadium with Chemometric Treatment and a Fisher Discriminant Model

Most identification methods for marine oil spills rely on single-dimensional chemical fingerprinting that is ineffective for weathered oil. A quantitative identification method is proposed in this study based upon multi-dimensional chemical fingerprinting composed of diagnostic ratios, wavelet coefficient, and the Ni/V. Twenty oil samples were used including 6 marine fuels, 7 Middle Eastern crude oils, and 7 non-Middle East crude oils. Diagnostic ratios were discussed using partial least square analysis, and fluorescence spectra were analyzed by db7 wavelet basis of 6-layers in the discrete wavelet transform. The first principal component and the wavelet coefficient at 354 ± 2 nm were combined with Ni/V and optimized by an exhaustive method as variables to establish the Fisher discriminant model. The developed model had 100% accuracy for modeled oil samples and for non-modeled crude oil before and after weathering, respectively, with 83.3% accuracy for non-modeled or short-term weathered marine fuel. The accuracy was 3% higher than reported in the literature.

Diagnosis of malignant versus tuberculous ascites using tumor markers and globulin ratios in serum and ascites: A Fisher discriminant model

Background and study aimsThis study was conducted to investigate the significance of tumor and biochemical markers in serum and ascitic fluid in the differential diagnosis of tuberculous and malignant ascites. Patients and methodsBased on findings from natural orifice transluminal endoscopic surgery and postoperative pathology or cytology of 63 patients, they were divided into the malignant group (31 patients) and the tuberculous group (32 patients). Levels of tumor markers, albumin, globulin, and lactate dehydrogenase were measured simultaneously. Data were statistically analyzed, and a Fisher discriminant model was established. The receiver operating characteristic curve was constructed to confirm the discriminant value. ResultsThe levels of carcinoembryonic antigen (CEA), cancer antigen 125 (CA125), cancer antigen 19-9 (CA 19-9), and globulin in serum and ascitic fluid were different between the tuberculous and malignant ascites groups (P < .05). The ratios of ascites-to-serum levels of CEA, CA125, and CA 19-9, as well as the ratio of serum-to-ascites of globulin levels, were different between the two groups (P < .05). The Fisher discriminant model was established based on the ascites-to-serum ratios of CEA, CA125, and CA 19-9 levels and the serum-to-ascites ratio of globulin levels. The area under the curve was 0.908, the sensitivity was 0.838 (26/31), and the specificity was 0.875 (28/32). ConclusionA Fisher discriminant model can be established using serum and ascites tumor markers and globulin ratios, which is valuable in the differential diagnosis of tuberculous versus malignant ascites.

A rapid and non-destructive method for detecting the water-holding capacity of pork using composite film

Abstract Water-holding capacity (WHC) is an important indicator of pork quality, but the existing detection methods of WHC are either expensive or time-consuming. In this study, a new method of pork WHC detection was developed by a composite film. The preparation method, mechanical properties and service life of the composite film were studied. The result showed that composite film was 0.46 ± 0.06 mm thick and had a service life of 21 days, tensile strength of 7.72 ± 0.11 MPa and the elongation at break of 28.54 ± 0.15%. Thirty groups of pork samples were randomly selected to build the model and another twenty groups were used to verify the model accuracy. Results showed that the accuracy of composite film coupled with Fisher discriminant model to detect the WHC of pork is 90%. This study demonstrates the high value of composite film as a detection tool to classify WHC of pork.

Discriminant analysis of mine water inrush sources with multi-aquifer based on multivariate statistical analysis

Accurate and effective identification of the source mine water inrush water is vital for warning system implementation and post-disaster rescue decision-making and is crucial for mine water disaster prevention plans. To fully excavate the hydrogeological information carried by the water samples of different water sources, and fully consider the influence of the correlation between the ions and the existence of multi-collinearity on the discrimination results, so as to improve the accuracy of the inrush water source discrimination, this study conducted multivariate statistical analysis and discriminant analysis on 37 water samples of three types of water sources in Xutuan coal mine, and established the Fisher discriminant model for mine water inrush sources based on fuzzy cluster analysis and factor analysis. The discriminant accuracy of the model was tested using re-substitution and cross-validation, and compared with the discriminant result of traditional Fisher discriminant model. The results showed that the discriminant accuracies of the re-substitution and cross-validation were 91.9% and 89.2%, respectively, while the discrimination accuracy of cross-validation of the traditional Fisher discrimination model was 86.5%. The discrimination accuracy of this model was higher than that of the traditional Fisher discrimination model. Therefore, the Fisher discriminant model for mine water inrush sources based on fuzzy cluster analysis and factor analysis established in this study can improve the accuracy of a water source discrimination model, and can provide a useful reference for mine water disaster prevention.

Fisher Discriminant Model Based on LASSO Logistic Regression for the Computed Tomography Imaging Diagnosis of Pelvic Rhabdomyosarcoma in Children

Fisher Discriminant Model Based on LASSO Logistic Regression for the Computed Tomography Imaging Diagnosis of Pelvic Rhabdomyosarcoma in Children

Radiomic analysis on magnetic resonance diffusion weighted image in distinguishing triple-negative breast cancer from other subtypes: a feasibility study

PurposeThis work aimed to explore whether radiomic features on magnetic resonance diffusion weighted image (MR DWI) can be used to identify triple-negative breast cancer (TNBC) and other subtypes (non-TNBC). Materials and methodsThis retrospective study included 221 unilateral patients who underwent breast MR imaging prior to neoadjuvant chemotherapy. The subtypes of breast cancer include luminal A (n = 63), luminal B (n = 103), human epidermal growth factor receptor-2 (HER2) overexpressing (n = 30), and triple negative (n = 25). Radiomic features were extracted using Omini-Kinetic software on DWI. Student's t-test and Mann–Whitney U test were used to compare the features between TNBC and non-TNBC patients. Logistic regression analysis and receiver operating characteristic (ROC) curve were used to evaluate the diagnostic efficiency of radiomic features. The Fisher discriminant model was employed to distinguish TNBC and non-TNBC patients automatically. An additional validation dataset with 169 patients was utilized to validate the model. ResultsA total of 76 imaging features were extracted from each lesion on DWI images, and 12 radiomic features were statistically significant between TNBC and non-TNBC patients (P < 0.05). The area of receiver operating characteristic curve (AUC) was 0.817 to apply logistic regression analysis. The accuracy of Fisher discriminant model in distinguishing TNBC and non-TNBC patients was 95.4%, and leave-one-out cross validation was achieved with an accuracy of 83.7%. The same classification analysis of the validation dataset showed an accuracy of 83.4% and an AUC of 0.804. ConclusionBreast lesions exhibit differences in radiomic features from DWI, enabling good discrimination between TNBC and non-TNBC tumors.

Estimating Postmortem Interval by Protein Chip Detection Technology Combined with Multidimensional Statistics

Objective To obtain the protein expression profile of rat liver tissue after death by the 2100 bioanalyzer combined with protein chip, and infer the relationship between protein expression profile and postmortem interval. Methods Rats were killed by abdominal anesthesia and placed at 16 ℃. Water-soluble proteins in liver tissues were extracted at 14 time points after death. The expression profile data of proteins with relative molecular weight of 14 000-230 000 were obtained using protein chip, and principal component analysis （PCA）, partial least squares-discriminant analysis （PLS-DA） and Fisher discriminant were used to analyze the data. Results According to the changes of protein expression profile, the postmortem interval was divided into group A （0 d）, group B （1-9 d）, group C （12-30 d） according to the result of PLS-DA. The prediction accuracy of the training set and test set of the model were all 100.0%, and the internal cross-validation of the training set was 100.0% according to Fisher discriminant. The Fisher discriminant model at each time point of group B and C was established to narrow the time window of postmortem interval estimation. The prediction accuracy of the training set and test set were all 100.0%, and the internal cross-validation accuracy of the training set was 100.0% in group B. The prediction accuracy of the training set and test set were respectively 95.2% and 78.6% in group C, and the internal cross-validation of the training set was 88.1%. Conclusion Protein chip detection technology can quickly and easily obtain the expression profile of water-soluble proteins of rat liver tissue with a relative molecular weight of 14 000-230 000 at different time points after death. PLS-DA and Fisher discriminant models are established to classify and predict the postmortem interval, in order to provide new ideas and methods for postmortem interval estimation.

Evaluation of taste changes of Scutellariae Radix before and after wine-frying based on electronic ongue technology and its application in identification of Scutellariae Radix pieces

Scutellariae Radix(Huangqin) is a well-known traditional Chinese medicine(TCM) used for the treatment of clearing heat in clinical application. It is bitter-cold by using directly, but the bitter-cold property can be relieved after wine-frying. The study of taste changes before and after wine-frying of Scutellariae Radix is of great significance in identifying Scutellariae Radix and wine-processed Scutellariae Radix and clarifying the traditional theory of wine-processing. In this experiment, 10 batches of Scutellariae Radix and wine-processed Scutellariae Radix were prepared. The contents of 5 flavonoids were determined by high performance liquid chromatography(HPLC), and principal component analysis(PCA) was performed with 5 flavonoids as variables. As a result, the contents were different in different batches of Scutellariae Radix, but Scutellariae Radix and wine-processed Scutellariae Radix could not be distinguished. Five sensory attributes(sour, salty, fresh, sweet, and bitter) were evaluated by artificial tasting, and the response values of 7 sensors(AHS, AHS, PKS, CTS, NMS, CPS, ANS, SCS) representing the taste of pieces were detected by electronic tongue. The correlation between sensory evaluation and response values of the electronic tongue were analyzed, and the results showed that the sensory evaluation of sour, salty, fresh, sweet, bitter and AHS, CTS, NMS, ANS, SCS sensors had different degrees of correlation, indicating that the electronic tongue technology can be used as an alternative to artificial taste and can serve as a means for quantifying the taste, and it can be used to evaluate the taste of TCM pieces. The taste method was used to analyze the response values of the electronic tongue, and the results showed that the bitterness of wine-processed Scutellariae Radix decreased and the salty taste increased. PCA was used to analyze taste changes before and after wine-processed Scutellariae Radix, and the results showed that taste differences between 2 pieces were divided into 2 categories. PCA loading scattering plots showed that response of saltiness and bitterness were the major factors to affect overall taste in Scutellariae Radix and wine-processed Scutellariae Radix. Based on electronic tongue response values, the Fisher discriminant model for Scutellariae Radix and wine-processed Scutellariae Radix was established, which showed that it could effectively discriminate them with a recognition rate of 100%. The experimental results showed that the electronic tongue combined with multivariate statistical analysis can be used to evaluate taste of TCM, at the same time, it could provide a fast and simple method for identifying different processed products.

Prediction of pathological grading of Henoch-Schönlein purpura nephritis in children based on Fisher stepwise discriminant analysis

Objective To establish the pathological grades of Henoch-Schonlein purpura nephritis(HSPN) in children with diagnostic prediction models by stepwise Fisher discriminant in children. Methods Based on the investigation of 28 clinical indicators from 144 cases with HSPN came from Children′s Hospital of Chongqing Medical University, the sensitive indicators were found and stepwise Fisher discriminant model was established and its accuracy in predicting the pathological classification of HSPN was tested. Results There were 5 laboratory indicators and clinical manifestations with different pathological grades of HSPN.In children with pathological grade Ⅱ, Ⅲ and Ⅳ, 5 indicators were screened (P<0.05) and stepwise Fisher discriminant models were established.And the correct rate of comprehensive diagnosis was (61.371±8.740)% in 100 random sampling diagnostic simulations; in children with pathological grade Ⅲa and Ⅲb, 5 indicators were also screened (P<0.05) and stepwise Fisher discriminant models were established.And the correct rate of comprehensive diagnosis was (68.015±5.736)% in 100 random sampling diagnostic simulations. Conclusions The stepwise Fisher discriminant models established in this research have a better diagnostic accuracy in forecasting for pathological grade of HSPN, and have a certain guiding value on early treatment and prognosis evaluation of children with newly diagnosed HSPN. Key words: Stepwise Fisher discriminant model; Henoch-Schonlein purpura nephritis; Pathology; Child

The Stability Evaluation of Shaft during Drastic Drawdown Dewatering of Alluvium

The hydrophobicity of the aquifer at the bottom of the porous alluvium will affect the stability of the shaft. According to the changes of water level and the compressive amount of alluvium, we can evaluate the shaft stability and predict the shaft failure. In this work, the simulation model of the auxiliary shaft in the Zhuxianzhuang Coal Mine is generated by using the Nsdc software to evaluate the stability of the shaft during drastic drawdown dewatering. Based on the measured hydrophobic compression ratio in an adjacent coal mine, the compressive amounts of the strata near the main and auxiliary shafts in the Zhuxianzhuang Coal Mine are predicted under the condition of drastic drawdown dewatering, which will be 249.69 mm and 302.75 mm, respectively. It is more likely that the shaft wall may fracture in the 15th day (fourth load level) under the condition of drastic drawdown dewatering. The formation compressive amount near the auxiliary shaft is approximately 320 mm, which is close to the measured predicted value. At the same time, the Fisher discriminant model is established, and it is predicted that the state of the main and auxiliary shafts will be failure under the conditions of drastic drawdown dewatering in the Zhuxianzhuang Coal Mine. Based on the simulating results, the technical means of using the ground grouting for early prevention and control is proposed.

High-throughput foodomics strategy for screening flavor components in dairy products using multiple mass spectrometry

A reliable Fisher discriminant model was established which was able to analyze the aroma component in milk, dairy products, flavors and fragrance, and applied on its variety identification. Foodomics was applied on screening of flavor components in 1093 dairy products and flavor samples in this study. Stepwise discrimination was used to screen the components of the dairy products and flavor samples that had a significant effect on the classification results, and discriminant function analysis. Then nine principal components were used for established the Fisher discriminant model. The three-dimensional coordinate distance of the sample was calculated and as the gist. The result showed that samples and flavors were distributed in eight different sites. The separation and clustering effects are better. The objective of the present study was to effectively determine whether or not flavors were added to dairy products.

Latent Discriminant Analysis with Representative Feature Discovery

Linear Discriminant Analysis (LDA) is a well-known method for dimension reduction and classification with focus on discriminative feature selection. However, how to discover discriminative as well as representative features in LDA model has not been explored. In this paper, we propose a latent Fisher discriminant model with representative feature discovery in an semi-supervised manner. Specifically, our model leverages advantages of both discriminative and generative models by generalizing LDA with data-driven prior over the latent variables. Thus, our method combines multi-class, latent variables and dimension reduction in an unified Bayesian framework. We test our method on MUSK and Corel datasets and yield competitive results compared to baselines. We also demonstrate its capacity on the challenging TRECVID MED11 dataset for semantic keyframe extraction and conduct a human-factors ranking-based experimental evaluation, which clearly demonstrates our proposed method consistently extracts more semantically meaningful keyframes than challenging baselines.