Effect of correlation on combining diagnostic information from two images of the same patient

Liu, Bei; Metz, Charles E.; Jiang, Yulei

doi:10.1118/1.2064787

Cited by 19 publications

(15 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Fourth, the fixed weighting method has been used in previous studies to combine the CAD scores of different view images (Wang et al , 2011; Liu et al , 2005), in this study we also tested the fixed weighting method to combine two ANN-generated classification scores computed from the CC and MLO view images. Comparing with the result of our previous study (Wang et al , 2011), we found that the optimal weighting factors on the two detection or classification scores computed from the CC and MLO view images were dataset dependent.…”

Section: Discussionmentioning

confidence: 99%

“…Thus, in order to improve CAD performance, many investigators have also developed and tested different methods of combining multiple view images into the classification process to improve CAD performance in detection and classification of mammographic abnormalities. These efforts include more accurately locating the breast masses and/or FP detections on two ipsilateral (CC and MLO) views (Zheng et al , 2006; Zheng et al , 2009; van Engeland and Karssemeijer, 2007; Wei et al , 2009; Yuan et al , 2008) and more effectively combining the information detected from different image views (Liu et al , 2005; Wang et al , 2011; Wei et al , 2011b). Despite these research efforts, multi-view based CAD schemes have not been accepted and used in the clinical practice to date.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Reduction of false-positive recalls using a computerized mammographic image feature analysis scheme

Tan¹,

Pu²

2014

Phys. Med. Biol.

View full text Add to dashboard Cite

The high false-positive recall rate is one of the major dilemmas that significantly reduce the efficacy of screening mammography, which harms a large fraction of women and increases healthcare cost. This study aims to investigate the feasibility of helping reduce false-positive recalls by developing a new computer-aided diagnosis (CAD) scheme based on the analysis of global mammographic texture and density features computed from four-view images. Our database includes full-field digital mammography (FFDM) images acquired from 1052 recalled women (669 positive for cancer and 383 benign). Each case has four images: two craniocaudal (CC) and two mediolateral oblique (MLO) views. Our CAD scheme first computed global texture features related to the mammographic density distribution on the segmented breast regions of four images. Second, the computed features were provided to two artificial neural network (ANN) classifiers that were separately trained and tested in a ten-fold cross-validation scheme on CC and MLO view images, respectively. Finally, two ANN classification scores were combined using a new adaptive scoring fusion method that automatically determined the optimal weights to assign to both views. CAD performance was tested using the area under a receiver operating characteristic curve (AUC). The AUC=0.793±0.026 was obtained for this four-view CAD scheme, which was significantly higher at the 5% significance level than the AUCs achieved when using only CC (p = 0.025) or MLO (p = 0.0004) view images, respectively. This study demonstrates that a quantitative assessment of global mammographic image texture and density features could provide useful and/or supplementary information to classify between malignant and benign cases among the recalled cases, which may eventually help reduce the false-positive recall rate in screening mammography.

show abstract

Section: Discussionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Reduction of false-positive recalls using a computerized mammographic image feature analysis scheme

Tan¹,

Pu²

2014

Phys. Med. Biol.

View full text Add to dashboard Cite

show abstract

“…The degree of improvement is inversely related to the correlation coefficients of the two sets of detection scores [27]. Although previous study [28] has shown that averaging two lower correlated detection scores usually achieves the best performance than other fusion methods, we tested different methods to combine the ANN and KNN generated detection scores in this study. First, we computed the final detection score by averaging ANN and KNN generated detection scores.…”

Section: Methodsmentioning

confidence: 99%

Improving Performance of Computer-aided Detection Scheme by Combining Results From Two Machine Learning Classifiers

Park

2009

Academic Radiology

View full text Add to dashboard Cite

Rationale and Objectives-Global data and local instance based machine learning methods and classifiers have been widely used to optimize computer-aided detection (CAD) schemes to classify between true-positive and false-positive detections. In this study the authors investigated the correlation between these two types of classifiers using a new independent testing dataset and assessed the potential improvement of a CAD scheme performance by combining the results of the two classifiers in detecting breast masses.Materials and Methods-The CAD scheme first used image filtering and a multi-layer topographic region growth algorithm to detect and segment suspicious mass regions. The scheme then used an image feature based classifier to classify these regions into true-positive and falsepositive regions. Two classifiers were used in this study. One was a global data based machine learning classifier, an artificial neural network (ANN), and the other one was a local instance based machine learning classifier, a k-nearest neighbor (KNN) algorithm. An independent image database involving 400 mammography examinations was used in this study. Among them, 200 were cancer cases and 200 were negative cases. The pre-optimized CAD scheme was applied twice to the database using the two different classifiers. The correlation between the two sets of classification results was analyzed. Three sets of CAD performances using the ANN, KNN, and average detection scores from both classifiers were assessed and compared using the free-response receiver operating characteristics (FROC) method. Results-The results showed that the ANN achieved higher performance than the KNN with a normalized area under the performance curve (AUC) of 0.891 versus 0.845. The correlation coefficients between the detection scores generated by the two classifiers were 0.436 and 0.161 for the true-positive and false-positive detections, respectively. The average detection scores of the two classifiers improved CAD performance and reliability by increasing AUC to 0.912 and reducing the standard error of the estimated AUC by 14.4%. The detection sensitivity was also increased from 75.8% (ANN) and 65.9% (KNN) to 80.3% at the false-positive detection rate of 0.3 per image.Conclusions-The study demonstrates that the global and local data based machine learning classifiers (ANN and KNN) generate low correlated detection results and combining the detection scores of these two classifiers can significantly improve overall CAD performance (p < 0.01) and reduce standard error in CAD performance assessment.

show abstract

“…A comparison of the performance differences between the three classification protocols (Table 4) does in fact show a statistically significant increase in performance when some form of averaging is used (note that multiple comparisons are not strictly necessary here because we are looking at the trend in the whole set of comparisons —A and B vs. C— and not selecting one or two out). This is not entirely surprising as previous studies have shown that combining image information via averaging tends to be helpful in non-medical image science [25] and it yields a positive improvement in diagnostic performance in medical imaging as well [26, 27]. It is likely that averaging is able to filter out some of the noise associated with outliers that are generated during the feature extraction or lesion classification process.…”

Section: Discussionmentioning

confidence: 86%

Performance of Breast Ultrasound Computer-aided Diagnosis

et al. 2008

View full text Add to dashboard Cite

Rationale and Objectives-The automated classification of sonographic breast lesions is generally accomplished by extracting and quantifying various features from the lesions. The selection of images to be analyzed, however, is usually left to the radiologist. Here we present an analysis of the effect that image selection can have on the performance of a breast ultrasound computer-aided diagnosis system. Materials and Methods-A database of 344 different sonographic lesions was analyzed for this study (219 cysts/benign processes, 125 malignant lesions). The database was collected in an IRBapproved, HIPAA-compliant manner. Three different image selection protocols were used in the automated classification of each lesion: all images, first image only, and randomly selected images. After image selection, two different protocols were used to classify the lesions: A) the average feature values were input to the classifier, or B) the classifier outputs were averaged together. Both protocols generated an estimated probability of malignancy. Round-robin analysis was performed using a Bayesian neural network-based classifier. Receiver operating characteristic analysis was used to evaluate the performance of each protocol. Significance testing of the performance differences was performed via 95% confidence intervals and non-inferiority tests. Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain. Results-The differences in the area under the ROC curves were never more than 0.02 for the primary protocols. Non-inferiority was demonstrated between these protocols with respect to standard input techniques (all images selected and feature averaging). NIH Public AccessConclusion-We have proven that our automated lesion classification scheme is robust and can perform well when subjected to variations in user input.

show abstract

Effect of correlation on combining diagnostic information from two images of the same patient

Cited by 19 publications

References 19 publications

Reduction of false-positive recalls using a computerized mammographic image feature analysis scheme

Reduction of false-positive recalls using a computerized mammographic image feature analysis scheme

Improving Performance of Computer-aided Detection Scheme by Combining Results From Two Machine Learning Classifiers

Performance of Breast Ultrasound Computer-aided Diagnosis

Contact Info

Product

Resources

About