Differentiation of Benign and Malignant Breast Tumors by Logistic Regression and a Classification Tree using Doppler flow signals

Abstract

In breast examinations with Doppler, an increased flow is found in malignant tumors. With the relatively new color Doppler, we measured different flow values in 133 cancer patients and in 325 women with benign disease. These measurements were used to develop diagnostic rules. For the highly correlated flow values, we used a stepwise procedure to select a final logistic regression model and a tree-based approach, which is a different way of modeling. With both approaches we developed simple diagnostic rules of which the sensitivity and the specificity exceeded 90%. There are no differences between the two approaches concerning discriminative ability. As complex statistical modeling leads to an overoptimism in the assessment of the error rates, we applied sensitivity analysis, investigated the stability of the selected logistic regression model, and estimated the magnitude of the overoptimism of the diagnostic rules with resampling methods. The results indicate that the estimates of sensitivity and specificity are probably close to realistic values for a clinical setting.

Keywords

Diagnostic Rules - Logistic Regression - Tree-based Approach - Breast Tumor - Resampling

• REFERENCES

• 1 Wells PNT, Halliwell M, Skidmore R, Webb AJ, Woodcock JP. Tumour detection by ultrasonic Doppler blood-flow signals. Ultrasonics 1977; 15: 231-2.
• 2 Cosgrove D, Bamber JC, Davey JB, McKinna JA, Sinnet HD. Color Doppler signals from breast tumors. Work in progress. Radiology 1990; 176: 175-80.
• 3 Madjar H. Breast examinations with continuous wave and color Doppler. Ultrasound Obstet Gynecol 1992; 2: 215-20.
• 4 Hosmer DW, Lemeshow S. Applied Logistic Regression. New York: Wiley; 1989
• 5 Efron B. Estimating the error rates of a prediction rule: Improvement of cross-validation. JASA 1983; 78: 316-31.
• 6 Gong G. Cross-validation, the Jackknife, and the bootstrap: Excess error estimation in forward logistic regression. JASA 1986; 81: 108-13.
• 7 Sauerbrei W, Schumacher M. A bootstrap resampling procedure for model building: Application to the Cox regression model. Statistics in Medicine 1992; 11: 2093-109.
• 8 Van Houwelingen JC, Le Cessie S. Predictive value of statistical models. Statistics in Medicine 1990; 9: 1303-25.
• 9 Efron B, Tibshirani RJ. An Introduction to the Bootstrap. New York: Chapman and Hall; 1993
• 10 Altman DG, Lausen B, Sauerbrei W, Schumacher M. Dangers of using "optimal" cutpoints in the evaluation of prognostic factors. J Nat Cancer Inst 1994; 86: 829-35.
• 11 Lausen B, Sauerbrei W, Schumacher M. Classification and regression trees (CART) used for the exploration of prognostic factors measured on different scales. In: Computational Statistics. Dirschedl P, Ostermann R. (eds) Berlin: Physica; 1994: 485-98.
• 12 Blettner M, Sauerbrei W. Influence of model-building strategies on the results of a case-control study. Statistics in Medicine 1993; 12: 1325-38.
• 13 Madjar H, Prompeler HJ, Sauerbrei W, Wolfarth R, Pfleiderer A. Color Doppler flow criteria of breast lesions. Ultrasound Med Biol 1994; 20: 849-58.
• 14 Parkin DM, Muir CS, Whelan S. et al. (eds) Cancer incidence in five continents. Lyon I ARC Scientific Publication 230; VI 1992
• 15 Breslow NE, Day NE. The analysis of case-control studies. Statistical Methods in Cancer research. Vol. 1: International Agency for Research on Cancer. Lyon: 1980
• 16 Altman DG. Practical statistics for medical research. London: Chapman and Hall; 1991
• 17 Breiman L, Friedman JH, Olshon RA, Stone CJ. Classification and regression trees. Montery: Wadsworth; 1984
• 18 Miller R, Siegmund D. Maximally selected chi-square statistics. Biometrics 1982; 38: 1011-6.
• 19 Lausen B, Schumacher M. Maximally selected rank statistics. Biometrics 1992; 48: 73-85.
• 20 Hanley JA, McNeil BJ. The meaning and use of the area under a receiver operating characteristic (ROC) curve. Radiology 1982; 143: 29-36.
• 21 Statistical Sciences, INC.. S-PLUS®, Reference Manual, Version 3.2. Seattle: 1993
• 22 SAS Institute Inc. SAS/STAT® User's Guide, Version 6. Cary (NC): 1989
• 23 Diepgen TL, Sauerbrei W, Fartasch M. Development and Validation of Diagnostic Scores for Atopie Dermatitis Incorporating Criteria of Data Quality and Practical Usefulness. J Clin Epidemiol 1996; 49: 1031-8.
• 24 Simel DL, Feussner JR, DeLong ER, Matchar DB. Intermediate, Indeterminat, and Uninterpretable Diagnostic Test results. Med Decis Making 1987; 7: 107-14.
• 25 Merler S, Furlanello C. Selection of tree-based classifiers with the bootstrap 632+ rule. Biometrical 1 1997; 39 (3) 369-82.
• 26 Hall FM, Storella JM, Silverstone DZ. et al. Nonpalpable breast lesions: Recommendations for biopsy based on suspicion of carcinoma at mammography. Radiology 1988; 167: 353-8.
• 27 Wu Y, Giger ML, Doi K. et al. Artificial neural networks in mammography: Application to decision making in the diagnosis of breast cancer. Radiology 1993; 187: 81-7.
• 28 Floyd CE, Lo JY, Yun AJ. et al. Prediction of breast cancer malignancy using an artificial neural network. Cancer 1994; 74: 2944-8.
• 29 Baker JA, Kornguth PJ, Lo JY. et al. Breast cancer: Prediction with artificial neural network based on BI-RADS Standardized Lexicon. Radiology 1995; 196: 817-22.
• 30 Kahn CE, Roberts LM, Shaffer KA. et al. Construction of a Bayesian network for mammographie diagnosis of breast cancer. Comput Biol Med 1997; 27: 19-29.