Selection of diagnostic features on breast MRI to differentiate between malignant and benign lesions using computer-aided diagnosis: differences in lesions presenting as mass and non-mass-like enhancement

• DOI: 10.1007/s00330-009-1616-y
• Publication Date: 2009-09-29T06:23:50Z
• Authors: Dustin Newell; Ke Nie; Jeon-Hor Chen; Chieh-Chih Hsu; Hon J. Yu; Orhan Nalcioglu; Min-Ying Su
• Keywords: Neuroradiology

To investigate methods developed for the characterisation of morphology and enhancement kinetic features both mass non-mass lesions, to determine their diagnostic performance differentiate between malignant benign lesions that present as versus types.Quantitative analysis morphological parameters breast were used among four groups lesions: 88 (43 mass, 45 non-mass) 28 (19 9 non-mass). The kinetics was measured analysed obtain transfer constant (K(trans)) rate (k(ep)). For each eight shape/margin 10 texture obtained. presenting nonmass-like enhancement, only An artificial neural network (ANN) build model.For selected could reach an area under ROC curve (AUC) 0.87 in differentiating lesions. parameter (k(ep)) from hot spot tumour reached a comparable AUC 0.88. combined improved 0.93, with sensitivity 0.97 specificity 0.80. non-mass-like by ANN achieved 0.76. k(ep) 0.59, low added value.The results suggest quantitative can be developing automated CAD (computer-aided diagnosis) achieve high performance, but more advanced algorithms are needed diagnosis enhancement.