Detection and classification performance levels of mammographic masses under different computer-aided detection cueing environments1

Zheng, Bin; Swensson, Richard G.; Golla, Sara; Hakim, Christiane M.; Shah, Rachit D.; Wallace, Luisa P.; Gur, David

doi:10.1016/s1076-6332(03)00677-9

Cited by 43 publications

(31 citation statements)

References 24 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Nonetheless, there is a hint that incorrect CAD prompts can lead to missed cancers in the clinical literature. Zheng et al (2004) reported that masses were more likely to be missed by radiologists if they appeared in a non-cued area.…”

Section: Implications For Cad In a Clinical Settingmentioning

confidence: 99%

“…Clinical studies have also indicated that CAD leads to differing levels of cancerous detection depending on prompt validity. For example, Zheng et al (2004) found that CAD cues might reduce cancer detection in non-cued areas (see also Samulski et al, 2010) 2 . We suggest that this could be because observers become over-reliant on CAD and so fail to detect a visible cancer if the CAD cue fails to prompt it.…”

Section: Introductionmentioning

confidence: 99%

“…We call this the over-reliance hypothesis. However, the study by Zheng et al (2004) again only had a limited number of radiologists and only used conditions in which the prevalence rate of the cancer was high.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Low prevalence search for cancers in mammograms: Evidence using laboratory experiments and computer aided detection.

Kunar

Watson

Taylor‐Phillips

et al. 2017

Journal of Experimental Psychology: Applied

View full text Add to dashboard Cite

People miss a large proportion of targets when they only appear rarely. This Low Prevalence (LP) Effect could lead to serious consequences if it occurred in the real world task of searching for cancers in mammograms. Using a novel mammogram search task, we asked participants to search for a pre-specified cancer (Experiments 1-2) or a range of masses (Experiments 3-5) under high or low prevalence conditions. Experiment 1 showed that an LP Effect occurred using these stimuli. Experiment 2 tested an over-reliance hypothesis and showed that the use of Computer Aided Detection (CAD) led to fewer missed cancers with a valid CAD prompt yet, a large proportion of cancers were missed when CAD was incorrect. Experiment 3 -5showed that false alarms also increased when searching for a range of masses and that CAD reduced miss errors when it correctly cued the target but increased miss errors and false alarms when it did not. Furthermore, when a mass fell outside the CAD prompt it was more likely to be misidentified. No LP Effect was observed with the addition of CAD when people were asked to search for a range of targets. Theories and implications for mammogram search are discussed. 3 Public Significance StatementThis study suggests that search for low prevalence cancers in mammograms can be aided by Computer Aided Detection (CAD) but there is also a cost. CAD highlights areas in mammograms that are likely to contain a mass. When CAD successfully prompts a mass, observers are better at detecting it. However, observers also show an over reliance on CAD, and are more likely to miss cancers if the cue is incorrect.4

show abstract

Section: Implications For Cad In a Clinical Settingmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Low prevalence search for cancers in mammograms: Evidence using laboratory experiments and computer aided detection.

Kunar

Watson

Taylor‐Phillips

et al. 2017

Journal of Experimental Psychology: Applied

View full text Add to dashboard Cite

show abstract

“…Diagnostic imaging, although inherently uncertain is often described to provide certainty [11,12]. However, there have been several occasions where image interpretation has led to clinical errors in diagnosis [13]. Furthermore, sociologists argue that the portrayal of medical images as certain is problematic [14] and argue that it could lead to increasing demand for medical imaging tests from patients and clinicians or disregard for other forms of information [14].…”

Section: Introductionmentioning

confidence: 99%

Do medical images aid understanding and recall of medical information? An experimental study comparing the experience of viewing no image, a 2D medical image and a 3D medical image alongside a diagnosis

Phelps

Wellings

Griffiths

et al. 2017

Patient Education and Counseling

View full text Add to dashboard Cite

Copies of full items can be used for personal research or study, educational, or not-for-profit purposes without prior permission or charge. Provided that the authors, title and full bibliographic details are credited, a hyperlink and/or URL is given for the original metadata page and the content is not changed in any way. A note on versions:The version presented here may differ from the published version or, version of record, if you wish to cite this item you are advised to consult the publisher's version. Please see the 'permanent WRAP url' above for details on accessing the published version and note that access may require a subscription. (0) 2476 574 880 Abstract Objective: This study compared the experience of viewing 3D medical images, 2D medical images and no image presented alongside a diagnosis. Methods: We conducted two laboratory experiments, each with 126 healthy participants. Participants heard three diagnoses; one accompanied by 3D medical images, one accompanied by 2D medical images and one with no image. Participants completed a questionnaire after each diagnosis rating their experience. In Experiment 2, half of the participants were informed that image interpretation can be susceptible to errors. Results: Participants preferred to view 3D images alongside a diagnosis (p<.001) and reported greater understanding (p<.001), perceived accuracy (p<.001) and increased trust (p<.001) when the diagnosis was accompanied by an image compared to no image. There was no significant difference in trust between participants who were informed of errors within image interpretation and those who were not. Conclusion: When presented alongside a diagnosis, medical images may aid patient understanding, recall and trust in medical information. Practical Considerations: Medical images may be a powerful resource for patients that could be utilised by clinicians during consultations. KeywordsDoctor-patient communication; patient understanding; recall of medical information; patient trust in medical information.

show abstract

“…16,17 Therefore, a key step in the development of ICAD schemes for mammography is to improve their performance in classifying between malignant and benign mass regions. We show that the choice of distance metric affects the performance of an ICAD system and that machine learning enables the construction of effective domain-specific distance metrics.…”

Section: Introductionmentioning

confidence: 99%

Learning distance metrics for interactive search-assisted diagnosis of mammograms

et al. 2007

Self Cite

View full text Add to dashboard Cite

The goal of interactive search-assisted diagnosis (ISAD) is to enable doctors to make more informed decisions about a given case by providing a selection of similar annotated cases. For instance, a radiologist examining a suspicious mass could study labeled mammograms with similar conditions and weigh the outcome of their biopsy results before determining whether to recommend a biopsy. The fundamental challenge in developing ISAD systems is the identification of similar cases, not simply in terms of superficial image characteristics, but in a medically-relevant sense. This task involves three aspects: extraction of a representative set of features, identifying an appropriate measure of similarity in the high-dimensional feature space, and return the most similar matches at interactive speed. The first has been an active research area for several decades. The second has largely been ignored by the medical imaging community. The third can be achieved using the Diamond framework, an open-source platform that enables efficient exploration of large distributed complex data repositories. This paper focuses on the second aspect. We show that the choice of distance metric affects the accuracy of an ISAD system and that machine learning enables the construction of effective domain-specific distance metrics. In the learned distance, data points with the same labels (e.g., malignant masses) are closer than data points with different labels (e.g., malignant vs. benign). Thus, the labels of the near neighbors of a new case are likely to be informative. We present and evaluate several novel methods for distance metric learning and evaluate them on a database involving 2522 mass regions of interest (ROI) extracted from digital mammograms, with ground truth defined by biopsy results (1800 malignant, 722 benign). Our results show that learned distance metrics improve both classification (ROC curve) and retrieval performance.

show abstract

Detection and classification performance levels of mammographic masses under different computer-aided detection cueing environments1

Cited by 43 publications

References 24 publications

Low prevalence search for cancers in mammograms: Evidence using laboratory experiments and computer aided detection.

Low prevalence search for cancers in mammograms: Evidence using laboratory experiments and computer aided detection.

Do medical images aid understanding and recall of medical information? An experimental study comparing the experience of viewing no image, a 2D medical image and a 3D medical image alongside a diagnosis

Learning distance metrics for interactive search-assisted diagnosis of mammograms

Contact Info

Product

Resources

About