Malignant and benign clustered microcalcifications: automated feature analysis and classification.

Y Jiang; R M Nishikawa; D E Wolverton; C E Metz; M L Giger; R A Schmidt; C J Vyborny; K Doi

doi:10.1148/radiology.198.3.8628853

Home/Archive/PubMed

PubMed · 8628853

Malignant and benign clustered microcalcifications: automated feature analysis and classification.

Y Jiang·R M Nishikawa·D E Wolverton·C E Metz·M L Giger·R A Schmidt·C J Vyborny·K Doi

View original source ↗Export metadata ↓

Abstract

PURPOSE: To develop a method for differentiating malignant from benign clustered microcalcifications in which image features are both extracted and analyzed by a computer. MATERIALS AND METHODS: One hundred mammograms from 53 patients who had undergone biopsy for suspicious clustered microcalcifications were analyzed by a computer. Eight computer-extracted features of clustered microcalcifications were merged by an artificial neural network. Human input was limited to initial identification of the microcalcifications. RESULTS: Computer analysis allowed identification of 100% of the patients with breast cancer and 82% of the patients with benign conditions. The accuracy of computer analysis was statistically significantly better than that of five radiologists (P = .03). CONCLUSION: Quantitative features can be extracted and analyzed by a computer to distinguish malignant from benign clustered microcalcifications. This technique may help radiologists reduce the number of false-positive biopsy findings.

Explore related subjects

Breast Diseases Breast Neoplasms Calcinosis Diagnosis, Differential Female Humans Mammography Neural Networks, Computer ROC Curve Radiographic Image Interpretation, Computer-Assisted Sensitivity and Specificity Journal Article Research Support, Non-U.S. Gov't Research Support, U.S. Gov't, Non-P.H.S.Research Support, U.S. Gov't, P.H.S.

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Y Jiang, R M Nishikawa, D E Wolverton, C E Metz, M L Giger, R A Schmidt, C J Vyborny, K Doi. 1996. Malignant and benign clustered microcalcifications: automated feature analysis and classification.. https://doi.org/10.1148/radiology.198.3.8628853

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related citations

PubMed2007

Comparison of 4 MV photon surface dose among Varian, Siemens, and Elekta linear accelerators for tangential breast treatment: a phantom study.

PURPOSE: We have compared the differences in a 4-MV photon surface dose among Varian, Siemens, and Elekta linear accelerators (linacs) with wedges for tangential breast treatment. MATERIALS AND METHODS: The wedge factor and the surface dose were measured using a solid water phantom and an ion chamber for each linear accelerator with various field sizes and wedge angles. A tangential treatment plan was applied to an elliptical hollow cylinder water phantom with a radiochromic film placed thereon. A dose was delivered to a simulated target in the phantom, and the resulting dose distribution was analyzed using a film scanner. RESULTS: Varian's wedges resulted in the highest wedge factors, ranging from 0.37 to 0.75 depending on the wedge angles. Varian's wedges led to the highest normalized skin doses, ranging between 0.40 and 0.73 depending on the wedge angles and field sizes. In the cylinder phantom test with two tangential beams, the Varian linac provided a nearly 20% higher maximum dose than the Siemens and Elekta linacs. CONCLUSION: The Varian linac resulted in the highest surface doses, and the Elekta linac led to the lowest for nearly all the measurement conditions we employed, including open beams.

Breast Diseases↗

PubMed2007

Technique to prevent recurrence of lactiferous duct fistula.

Explore the source record for details and available documents.

Breast Diseases↗

PubMed2007

[Is the application of CAD useful for radiologists for passing the official tests required for outpatient breast imaging?].

PURPOSE: To analyze whether currently available CAD systems meet the diagnostic requirements for passing screening tests (first reader CAD) and to analyze whether the additional usage of CAD systems provides significant support for the diagnosing radiologist on the basis of official screening test cases (second reading by CAD). MATERIAL AND METHODS: 200 images of 100 mammographies of 50 patients of an official screening test case collection were analyzed double-blind with and without CAD printouts (iCAD, USA ) by three radiologists: one experienced in breast analysis and CAD application, one experienced in mammography analysis but inexperienced in CAD usage, one with minimal experience with breast analysis and CAD application. All radiologists measured the largest diameter of any malignant mass. The mean value of these calculations was correlated to the largest diameter given by CAD prompts. RESULTS: The mean sensitivity and specificity increased slightly as a result of the additional usage of CAD (1 and 0.6 %, resp.). Both values are not statistically significant. The highest effect was measured for the radiologist with CAD experience, while no effect was measured for the inexperienced radiologist. CAD met the sensitivity requirements but not the specificity criteria (96 and 20.3 %, resp.). The sizes given by CAD prompts corresponded significantly with the real sizes (r = 0.45, p < 0.05). CONCLUSION: A statistically significant effect of the use of CAD could not been detected dependent on the radiological experience or the CAD experience. The effect of CAD on radiologists is still subtle despite the highly sensitive CAD performance due to the high number of false positive prompts and thus does not reach statistical significance. Sizes given by CAD prompts correlate significantly with the real sizes of malignant lesions.

Breast Diseases↗