Biomedical subjects

Elizabeth S Burnside

Publications and source records attributed to Elizabeth S Burnside.

9 recordsLinked to original sources

PubMed2006

Bayesian network to predict breast cancer risk of mammographic microcalcifications and reduce number of benign biopsy results: initial experience.

PURPOSE: To retrospectively determine whether a Bayesian network (BN) computer model can accurately predict the probability of breast cancer on the basis of risk factors and mammographic appearance of microcalcifications, to improve the positive predictive value (PPV) of biopsy, with pathologic examination and follow-up as reference standards. MATERIALS AND METHODS: The institutional review board approved this HIPAA-compliant study; informed consent was not required. Results of 111 consecutive image-guided breast biopsies performed for microcalcifications deemed suspicious by radiologists were analyzed. Mammograms obtained before biopsy were analyzed in a blinded manner by a breast imager who recorded Breast Imaging Reporting and Data System (BI-RADS) descriptors and provided a probability of malignancy. The BN uses probabilistic relationships between breast disease and mammography findings to estimate the risk of malignancy. Probability estimates from the radiologist and the BN were used to create receiver operating characteristic (ROC) curves, and area under the ROC curve (A(z)) values were compared. PPV of biopsy was also evaluated on the basis of these probability estimates. RESULTS: The BN and the radiologist achieved A(z) values of 0.919 and 0.916, respectively, which were not significantly different. If the 34 patients estimated by the BN to have less than a 10% probability of malignancy had not undergone biopsy, the PPV of biopsy would have increased from 21.6% to 31.2% without missing a breast cancer (P < .001). At this level, the radiologist's probability estimation improved the PPV to 30.0% (P < .001). CONCLUSION: A probabilistic model that includes BI-RADS descriptors for microcalcifications can distinguish between benign and malignant abnormalities at mammography as well as a breast imaging specialist can and may be able to improve the PPV of image-guided breast biopsy.

Adult↗

PubMed2005

Bayesian networks: computer-assisted diagnosis support in radiology.

Medical knowledge is growing at an explosive rate. While the availability of pertinent data has the potential to make the task of diagnosis more accurate, it is also increasingly overwhelming for physicians to assimilate. Using artificial intelligence techniques, a computer can process large amounts of data to help physicians manage the growing body of medical knowledge and thereby make better decisions. Computer-assisted diagnosis support is of particular interest to the diagnostic imaging community because radiologists must integrate huge amounts of data in order to diagnose disease. Bayesian networks, among the most promising artificial intelligence techniques available, enable computers to store knowledge and estimate the probability of outcomes based on probability theory. The article describes what a Bayesian network is and how it works using a system in mammography for illustration. A comparison of Bayesian networks with other types of artificial intelligence methods, specifically neural networks and case-based reasoning, clarifies the unique features and the potential of these systems to aid radiologists in the decisions they make every day.

Bayes Theorem↗

PubMed2005

The use of batch reading to improve the performance of screening mammography.

OBJECTIVE: The objective of our study was to prove that batch reading of screening mammograms can reduce recall rates without sacrificing cancer detection. MATERIALS AND METHODS: We analyzed recall rate, cancer detection, minimal cancer detection, detection of low-stage cancer, and tumor size from consecutive screening mammography examinations from October 2001 to July 2003. The initial 7,984 mammograms were interpreted in the midst of a busy breast imaging practice. Although these studies were not read online, the interpretations were often interrupted for telephone calls, procedures, and diagnostic mammograms. The remaining 1,538 studies were interpreted after the institution of dedicated uninterrupted batch reading. RESULTS: Recall rates were 20.1% before and 16.2% after the introduction of batch reading (p < 0.001). Cancer detection rates were not significantly different: 5.6 cancers were detected per 1,000 examinations without and 7.2 were detected per 1,000 with batch reading. Prognostic factors for breast cancers diagnosed between these groups also were not significantly different. Of the screening-detected cancers diagnosed before batch reading, minimal cancers comprised 67% and low-stage cancers accounted for 76%. Of the cancers diagnosed using batch reading, 73% were minimal and 91% were low stage. The mean size of cancers, 11.7 mm without batch reading and 9.1 mm with batch reading, also showed no statistically significant difference. CONCLUSION: Our experience shows that batch reading can significantly reduce screening mammography recall rates without affecting the cancer detection rate or the proportion of cancers diagnosed with favorable prognostic indicators.

Breast Neoplasms↗

PubMed2005

Knowledge discovery from structured mammography reports using inductive logic programming.

The development of large mammography databases provides an opportunity for knowledge discovery and data mining techniques to recognize patterns not previously appreciated. Using a database from a breast imaging practice containing patient risk factors, imaging findings, and biopsy results, we tested whether inductive logic programming (ILP) could discover interesting hypotheses that could subsequently be tested and validated. The ILP algorithm discovered two hypotheses from the data that were 1) judged as interesting by a subspecialty trained mammographer and 2) validated by analysis of the data itself.

Algorithms↗

PubMed2004

Patient, faculty, and self-assessment of radiology resident performance: a 360-degree method of measuring professionalism and interpersonal/communication skills.

RATIONALE AND OBJECTIVES: To develop and test the reliability, validity, and feasibility of a 360-degree evaluation to measure radiology resident competence in professionalism and interpersonal/communication skills. MATERIALS AND METHODS: An evaluation form with 10 Likert-type items related to professionalism and interpersonal/communication skills was completed by a resident, supervising radiologist and patient after resident-patient interactions related to breast biopsy procedures. Residents were also evaluated by faculty, using an end-of-rotation global rating form. Residents, faculty, and technologists were queried regarding their reaction to the assessments after a 7-month period. RESULTS: Fifty-six complete 360-degree data sets (range, 2-14 per resident) and seven rotational evaluations for seven residents were analyzed and compared. Internal consistency reliability estimates were 0.85, 0.86, and 0.87 for resident, patient, and faculty 360-degree evaluations, respectively. Correlations between resident-versus-patient, resident-versus-faculty, and patient-versus-faculty ratings for the 56 interactions were -0.06 (P =.64), 0.31 (P <.02), and 0.45 (P <.0006), respectively. Pearson correlation coefficients approached significant correlation (0.70) between the faculty global rating and patient 360-degree scores (P =.08) but not with faculty 360-degree scores. Residents and faculty felt that completing the 360-degree forms was easy, but the requirement for faculty presence during the consent process was burdensome. CONCLUSION: Results from this pilot study suggest that self, faculty, and patient evaluations of resident performance constitutes a valid and reliable assessment of resident competence. Additional data are needed to determine whether the 360-degree assessment should be incorporated into residency programs and how frequently the assessment should be performed. Requiring only a specified number of assessments per rotation would make the process less burdensome for residents and faculty.

Breast↗

PubMed2004

A probabilistic expert system that provides automated mammographic-histologic correlation: initial experience.

OBJECTIVE: We sought to determine whether a probabilistic expert system can provide accurate automated imaging-histologic correlations to aid radiologists in assessing the concordance of mammographic findings with the results of imaging-guided breast biopsies. MATERIALS AND METHODS: We created a Bayesian network in which Breast Imaging Reporting and Data System (BI-RADS) descriptors are used to convey the level of suspicion of mammographic abnormalities. Our system is a computer model that links BI-RADS descriptors with diseases of the breast using probabilities derived from the literature. Mammographic findings are used to update pretest probabilities (prevalence of disease) into posttest probabilities applying Bayes' theorem. We evaluated the histologic results of 92 consecutive imaging-guided breast biopsies for concordance with the mammographic findings during radiology-pathology review sessions. First, radiologists with no knowledge of the biopsy results chose BI-RADS descriptors for the mammographic findings. After the histologic diagnosis was revealed, the radiologists assessed concordance between the pathologic results and the mammographic findings. We then input the information gathered from these sessions into the Bayesian network to produce an automated mammographic-histologic correlation. RESULTS: We had a sampling error rate of 1.1% (1/92 biopsies). Our expert system was able to integrate pathologic diagnoses and mammographic findings to obtain probabilities of sampling error, thereby enabling us to identify the incorrect pathologic diagnosis with 100% sensitivity while maintaining a specificity of 91%. CONCLUSION: Our probabilistic expert system has the potential to help radiologists in identifying breast biopsy results that are discordant with mammographic findings and discovering cases in which biopsy sampling errors may have occurred.

Adult↗

PubMed2004

Using a Bayesian network to predict the probability and type of breast cancer represented by microcalcifications on mammography.

Since the widespread adoption of mammographic screening in the 1980's there has been a significant increase in the detection and biopsy of both benign and malignant microcalcifications. Though current practice standards recommend that the positive predictive value (PPV) of breast biopsy should be in the range of 25-40%, there exists significant variability in practice. Microcalcifications, if malignant, can represent either a non-invasive or an invasive form of breast cancer. The distinction is critical because distinct surgical therapies are indicated. Unfortunately, this information is not always available at the time of surgery due to limited sampling at image-guided biopsy. For these reasons we conducted an experiment to determine whether a previously created Bayesian network for mammography could predict the significance of microcalcifications. In this experiment we aim to test whether the system is able to perform two related tasks in this domain: 1) to predict the likelihood that microcalcifications are malignant and 2) to predict the likelihood that a malignancy is invasive to help guide the choice of appropriate surgical therapy.

Bayes Theorem↗

PubMed2002

Interpreting data from audits when screening and diagnostic mammography outcomes are combined.

OBJECTIVE: The objective of this study was to use mathematic models to aid mammography practices in interpreting outcomes data derived from a combination of screening and diagnostic examinations, and in interpreting diagnostic mammography outcomes data that are not segregated by indication for examination. MATERIALS AND METHODS: We analyzed outcomes from 51,805 consecutive mammography examinations. Screening and diagnostic examinations were audited separately. Diagnostic examinations were audited by indication for examination. Extrapolating from our known mix of screening (79%) and diagnostic (21%) examinations, we determined expected combined outcomes for various mixes that might be encountered in clinical practice. Similarly, we determined the expected overall diagnostic mammography outcomes for various clinically relevant mixes of indications for examination. RESULTS: Outcomes vary substantially depending on the mix of screening and diagnostic examinations performed. For example, expected outcomes for practices with screening-diagnostic mixes of 90-10% and 50-50% are, respectively: rate of abnormal findings, 6% versus 11%; rate of positive biopsy findings, 38% versus 42%; cancer detection rate, 10 per 1,000 versus 30 per 1,000; mean invasive cancer size, 14.4 mm versus 16.0 mm; nodal metastasis rate, 8% versus 11%; and rate of stage 0 and stage I cancers, 87% versus 82%. Diagnostic outcomes also vary substantially according to indication for examination, with a higher rate of abnormal findings, a higher rate of positive biopsy findings, and a larger mean invasive cancer size expected for mixes involving a high percentage of workups for palpable lesions. CONCLUSION: When screening and diagnostic mammography outcomes are not segregated during auditing, and when diagnostic outcomes are not segregated by indication for examination, analysis of combined audit data should be based on extrapolations from known outcomes.

Biopsy, Needle↗

PubMed2002

Differential value of comparison with previous examinations in diagnostic versus screening mammography.

OBJECTIVE: The purpose of our study was to analyze the differences in clinical outcomes of diagnostic and screening mammography depending on whether comparison is made with previous examinations. MATERIALS AND METHODS: We analyzed 48,281 consecutive mammography examinations for which previous mammography (9825 diagnostic, 38,456 screening) had been performed between 1997 and 2001, collecting data on demographics, whether comparison actually was made with previous examinations, abnormal findings (recall for screening mammography or biopsy recommendation for diagnostic mammography), biopsy yield of cancer, cancer detection rate, size of invasive cancers, axillary nodal status, and cancer stage. RESULTS: Comparison with previous examinations in the incidence screening setting decreases the recall rate from 4.9% to 3.8% (p < 0.0001) but does not significantly affect the biopsy yield (40-44%, p = 0.56) or the cancer detection rate (5.5-5.2/1000, p = 0.87). In the diagnostic setting, comparison with previous examinations increases the biopsy-recommended rate from 4.3% to 9.4% (p < 0.0001), the biopsy yield from 38% to 51% (p = 0.12), and the overall cancer detection rate from 11/1000 to 39/1000 (p < 0.0001). Comparison with previous examinations is not associated with a significant difference in mean tumor size. However, it is associated with a significant decrease in the frequency of axillary node metastasis and the cancer stage for screening mammography, but not for diagnostic mammography. CONCLUSION: For screening mammography, comparison with previous examinations significantly decreases false-positive but not true-positive findings and permits detection of cancers at an earlier stage. For diagnostic mammography, comparison with previous examinations increases true-positive findings.

Biopsy, Needle↗