This poster is published under an
open license. Please read the
disclaimer for further details.
Keywords:
Breast, Oncology, MR, MR-Diffusion/Perfusion, Computer Applications-Detection, diagnosis, Treatment effects, Cancer
Authors:
N. Michoux1, L. Bollondi2, A. Depeursinge3, A. Geissbuhler2, L. Fellah1, H. Müller3, I. Leconte1; 1Brussels/BE, 2Geneva/CH, 3Sierre/CH
DOI:
10.1594/ecr2015/B-1088
Methods and materials
This retrospective study was approved by our institutional ethical committee and included 70 patients.
All patients had an invasive breast carcinoma diagnosed on core-biopsy specimen.
To obtain a homogeneous histological sample for texture analysis,
only invasive ductal carcinomas with and without ductal carcinoma in situ were included in the study.
A baseline MRI as well as a pre-operative MRI to evaluate response to NAC was performed in all patients.
A pathological complete response (CR) was defined as the absence of invasive and in situ cancer in breast and nodes.
A partial response (PR) was defined as a decrease of invasive cancer exceeding 30%.
A non-response (NR) was defined as a decrease of invasive cancer lower than 30%.
At histological analysis,
15 patients were thus classified as CR,
36 as PR and 19 as NR.
MRI examinations were performed using a 1.5T whole-body imaging system (Gyroscan Intera,
Philips Medical System,
The Netherlands) and a breast coil.
Patients were imaged in the prone position with T2-weighted and diffusion-weighted imaging (DWI) (b0,
b600) sequences,
and a 3D gradient echo axial T1-weighted sequence with fat suppression (SPAIR).
Scan parameters were TR/TE = 4.8/2.4 ms,
flip angle = 10°,
FOV = 355x355 mm,
matrix 320x320,
slice thickness 2.5 mm,
voxel size 0.65x0.65x1.25 mm after reconstruction.
The anatomic study was followed by a dynamic study.
Patients received 0.1 mmol/kg of gadobenate dimeglumine (Multihance,
Bracco Imaging,
Germany) followed by 30 mL saline flush injected at a rate of 2 mL/s with an automated injector.
One pre- and five post-injection images were acquired with a temporal resolution of approximately 60 seconds.
Analyses were performed on subtracted images,
i.e.
the residual difference image obtained after the second post-contrast image has been subtracted from the pre-contrast image.
MR images were reviewed consensually by a trainee and two experienced radiologists without knowledge of the pathological findings or mammographic and sonographic data,
by using the BI-RADS MR lexicon 10 (Figure 1).
Breast lesions were segmented manually on each slice of the MRI volume then recontructed in 3D (Figure 2).
The intra-lesional texture was assessed as follows (Figure 3).
From the grey level co-occurrence matrix (GLCM),
11 texture parameters (i.e.
textons) describing the grey levels interdependence in the lesion were estimated 11.
From the run length matrix (RLM),
11 textons describing the distribution of runs of grey levels were estimated with the same computation parameters 12.
From the Riesz transform,
30 textons characterizing the important orientations and scale properties of grey levels were estimated 13.
3-D multiscale Riesz filterbanks are advantageous for texture characterization because they quantify the local amount of directional image patterns at multiple scales.
Second-order Riesz wavelets were used,
yielding 6 filters per image scale that are oriented along the main image directions X,
Y,
Z and three diagonals XY,
XZ and YZ.
We hypothesized that the local morphological properties of breast tissue can be expressed as the combinations of the responses of the oriented filters.
The mean value (over all voxels in the lesion) of the textons was estimated.
Then,
two multi-parametric classifiers were used to predict the non-responders to NAC: a logistic regression model 14 and a support vector machine (SVM) model 15.
As one cannot know a priori how many and which parameters are important to the classification,
all possible combinations of 2 to 5 parameters among 52 parameters) were submitted to the classifiers successively.
To estimate how accurately the predictive models would perform in practice,
a leave-one-out cross-validation was applied.