A REGISTRATION FRAMEWORK FOR EVALUATION OF T1, T2

AND DWI SIGNAL INTENSITIES IN MULTIPLE MYELOMA

E. Montin

, P. Potepan

and L. T. Mainardi

Department of Bioengineering, Politecnico di Milano, Via Golgi 39, 20131 Milan, Italy

Dipartimento di Diagnostica per Immagini e Radioterapia, Fond. IRCCS Istituto Nazionale Tumori

Via Venezian 1, 20133, Milan, Italy

Keywords:

Whole Body Imaging, MRI, Myeloma.

Abstract:

Objective. In this study we point out the Diffusion-Weighted Imaging (DWI) role in the diagnosis of multiple

myeloma (MM), comparing its signal values (SV) (Sommer et al., 2010) with the standard imaging modalities

T1, T2. We further evaluate how SV change in relation with the percentage of plasma cells inﬁltration evalu-

ated through bone marrow biopsy (BMB).

Methods. Since March 2008 23 patients with average age of 61 (± 11) years old, 11 females and 12 males,

have been investigated before their own therapy with a whole body MRI protocol, concerning a whole body

T1, a whole body T2 and a whole body DWI and a BMB. An experienced radiologist deﬁned for each patient

two volume of interests (VOIs): onto the main lesions and on healthy bones (Femur and Homerus). After that,

we have subdivided the full population by a clinical threshold of 25% on cells inﬁltration percentage; then, we

analysed statistical differences in the 2 groups (A, B).

Results. We found out that DWI voxels intensities in group A (inﬁltration ≤ 25%) were higher than group B,

this gap had to be considered statistically different (P ≤ 0.05).

1 INTRODUCTION

Although conventional radiography is still the stan-

dard widely approved staging procedure for newly di-

agnosed and relapsed multiple myeloma (MM) (Som-

mer et al., 2010), whole-body MRI (WB-MRI) using

T1- and T2-weighted contrast images has proved evi-

dence of its advantages over conventional skeletal sur-

vey (Ghanem et al., 2006). In this scenario the whole-

body DWI imaging is being attracting interest as a

tool for the investigation of MM lesions (Sakurada

et al., 2009). This study analyses the great potential

of DWI in MM diagnosis, comparing T1, T2, DWI

and correlating their values with clonal cells inﬁltra-

tion in bone marrow. When different MRI method-

ologies have to be compared, relevant role is plaid

by registration which allows the comparative analysis

among modalities. In this work we present a registra-

tion framework to align T1, T2 and DWI MRI acqui-

sition. The framework was designed for the evalua-

tion of signal intensities in Multiple Myeloma lesions.

2 MATERIAL AND METHOD

2.1 Patients

The analysed patients form a subset of a study

on whole-body MRI for multiple myeloma lesions

started in May 2008 at the national cancer institute

of Milan (INT). The study considers only patients at

diagnosis or at relapse after disease response (CR or

PR) lasting at least 6 months. The study was ap-

proved by the local ethics committee and all patients

gave their written informed consent before being in-

cluded. Globally, we analysed 23 patients (12 males

and 11 females, median age 61 years, age range 44, 81

years). Patients underwent a whole-body MRI scan

consisting of whole-body standard MRI (T1 and T2)

and Diffusion weighted MRI. None of the patients

had major artifacts on DW imaging that warranted

their exclusion from the study. After or immediately

before the MRI exams the patients underwent to bone

marrow biopsy (BMB).

563

Montin E., Potepan P. and Mainardi L..

A REGISTRATION FRAMEWORK FOR EVALUATION OF T1, T2 AND DWI SIGNAL INTENSITIES IN MULTIPLE MYELOMA.

DOI: 10.5220/0003777505630566

In Proceedings of the International Conference on Bio-inspired Systems and Signal Processing (MIAD-2012), pages 563-566

ISBN: 978-989-8425-89-8

 2012 SCITEPRESS (Science and Technology Publications, Lda.)

2.2 Imaging Protocol

MRI examinations were performed on a 1.5 T

MR imaging scanner Siemens Avanto (Erlangen-

Germany) using 6 array coils (head, neck, abdominal,

torso, pelvic and legs). Whole body images were cre-

ated composing different volumes: T1 and T2 were

acquired coronally in 3 - 5 steps, EPI DWI images

were acquired axially in 8 - 10 steps depending on

patients’ height.

The coronal T1-weighted tse2d1rr2 sequence was

performed with TR 550, TE 9.5, voxels size ∼ 1 x

5 x 1 mm, ∼ 1 mm gap, a ﬁeld of view (FoV) of 500

x 500 mm, a matrix of ∼ 500 x 500 , two NEX, TA

2.40 minutes.

Coronal T2-weighted spcir3d1345ns sequence was

performed with TR 4000, TE 367, voxels size ∼ 1

x 5 x 1 mm, ∼ 1 mm gap, a FoV of 500 x 500 mm, a

matrix of ∼ 500 x 500 , two NEX, TA 4 minutes.

Axial Echo-planar DW imaging was performed with

4 b-values (50, 400, 800, 1000

), ep2ddiff : TR

7900, TE 81, voxels size ∼ 2.5 x 2.5 x 6 mm, ∼ 1

mm gap, a FoV of 400 x 400 mm, a matrix of ∼ 160

x 160, two NEX, TA 3 minutes (table 1).

The total acquisition time was 30 min, chest and

abdominal T1 and T2 sequences were acquired dur-

ing breath-hold; all others sequences were acquired

during free breathing. No contrast agent was applied.

Table 1: Principal sequence features parameters.

Features T1 T2 DWI

Orientation Coronal Coronal Axial

TR 550 4000 7900

TE 9.5 367 81

Voxels size 1x5x1 1x5x1 2.5x2.5x6

2.3 Image Data Analysis

To evaluate different tissues property on different

MRI modalities we need to create 3D volumes and

whole-body volume has many challenges such as the

huge size of the matrix, the anisotropy of the volume,

the breathing of the patients etc . . . . We propose a

registration framework which is explained in the next

section.

2.3.1 3D Reconstruction and Registration

Whole-body examination was obtained by partial

MRI scans of sub-volumes; the ﬁrst step was to re-

build a unique 3D whole-body volume from the sub-

volumes acquired by T1 and T2 sequences. The sub-

Figure 1: Reconstruction of whole-body DWI, each DWI

sub-volumes is aligned on the reconstructed whole-body

T2.

volumes were spatial combined and voxels averaged

in the overlapped areas. In addiction, to compensate

possible deformation due to patient breathing a non-

rigid registration between T1 and T2 was performed

using the T2 as reference imaging.

Conversely the 3D Volume reconstruction for the

DWI was obtained by the afﬁne registration of each

DWI sub-volume with the reconstructed whole-body

T2. An afﬁne registration was used in this purpose

and overlapped voxels were averaged. The proce-

dure is described by ﬁgure 1. To optimize compu-

tation time, the registration parameters were calcu-

lated for b50 volumes and then applied to the other

volume acquired with different b-value. Finally to re-

move possible local misalignment, a non-rigid regis-

tration was performed between whole-body T2 and

DWI at lowest b-value (50

). All non-rigid reg-

istration were performed with IRTK software (Schn-

abel et al., 2001) using a multi-resolution optimiza-

tion with free-form deformations based on multi-level

B-splines (Lee et al., 1996) (Lee et al., 1997).

The total time for reconstruction and registration

was roughly 15 minutes for each patient. The accu-

racy of the registration was visually scored by an ex-

pert radiologist. None of the registered volume was

classiﬁed as non acceptable or erroneous by the radi-

ologist.

Our framework subdivides the registration task in

two parts: a global transformation and a local one.

This kind of solution is usually adopted in breast MR

Images to model the movement of the tissues (Rueck-

BIOSIGNALS 2012 - International Conference on Bio-inspired Systems and Signal Processing

564

ert et al., 1999). The global motion model describes

the overall motion of the sub-volume and is com-

pensated by an afﬁne transformation, which has 12

degrees of freedom, describing rotations and transla-

tions scaling and shearing.

For the local registration, through which we try to

compensate the local deformation of the sub-volume

(i.e. breathing movement on rib cage), we selected

a free form Deformations model (FFD), based on B-

splines (Rueckert et al., 1999). The basic idea of FFD

is to deform an object by manipulating an underlying

mesh of control points. The resulting deformation is

applied to the entire 3-D object and produces the ﬁnal

transformation.

For both the models we used normalised mutual

information as voxels similarity measure and a work-

ing resolution of 5 x 5 x 5 mm for the global registra-

tion while a thinner mesh for the local model 2 x 2 x

2 mm.

2.3.2 Image Quantiﬁcation

An experienced radiologist identiﬁes for each patients

2 volumes of interest: within the main lesions (in the

surrounding of the BMB) and on healthy bone (Femur

and Homerus). The selection was performed in one of

the image modalities, usually on b1000 DWI images

as suggested by the literature (Khoo et al., 2011) and

easily transferred to the other as the volume was regis-

tered previously, averaged values inside the ROI were

computed for each image modality.

2.3.3 Statistical Analysis

Comparison between intensity of healthy bone and le-

sions was performed using a Wilcoxon’s signed-rank

test, the level of signiﬁcance was set to P = 0.01.

In addition, the patients were categorised in two

groups based on the percentage of clonal cells in bone

marrow resulting from the BMB exam; group A had

an inﬁltration percentage ≤ 25% and group B > 25%.

An inﬁltration greater than 25% is for oncologist clin-

icians a typical threshold that decides for treatments

on patients. Mean values of the voxels intensities

were compared using Mann-Whitney rank-sum test,

to reveal the differences between the two groups, the

level of signiﬁcance was set P = 0.05.

3 RESULTS

Median signal intensity and 25

and 75

percentile

are reported in table 2, the same data are shown in

box-plot of ﬁgure 2. In all the modalities a statisti-

cal signiﬁcance P < 0.001 was observed between the

voxels values in the healthy ROI vs the lesions ROI.

Figure 2: Voxels intensity of the 2 ROI healthy bone mar-

row (H) and lesions (L) in T1, T2 and DWI (b50 ,b1000).

The second comparison between group A and

group B is shown in ﬁgure 3, it shows the difference

between the voxels values of the 2 groups (BMB≤

25% and BMB > 25%), signiﬁcant difference is ob-

served between group A and group B in the all the

DWI imaging in particular in all the cases signal in-

tensity was lower in the group B (Koh and Collins,

2007) (Khoo et al., 2011). Conversely no signiﬁcant

differences were observed in T1 and T2 imaging even

if a trend for lower value of lesions is observed.

4 CONCLUSIONS

Whole-body DWI imaging is a powerful tool for the

staging of patients with multiple myeloma (Padhani

et al., 2009) (Sommer et al., 2010). The new tech-

nique allows for fast whole-body imaging with low

technical and operational efforts. DWI sequences can

improve the accuracy of focal MM lesions identiﬁca-

tion in patients newly diagnosed. A decrease of DWI

voxels values evaluated at the same b-value can be po-

tentially related to an high grade myeloma with cells

inﬁltration percentage higher than 25% (P ≤ 0.05).

A REGISTRATION FRAMEWORK FOR EVALUATION OF T1, T2 AND DWI SIGNAL INTENSITIES IN MULTIPLE

MYELOMA

565

WBT1 WBT2 DWITB50 DWITB400 DWITB800 DWITB1000

100

150

200

250

300

350

400

450

500

voxel value group A n 6

WBT1 WBT2 DWITB50 DWITB400 DWITB800 DWITB1000

100

150

200

250

300

350

400

450

500

voxel group B n 17

Figure 3: Ranksum test on T1, T2 and DWI (b50, b1000) voxels values in patient from group A and B.

Table 2: Median and quartile of lesion (L) and healthy bone marrow (H) pixel values of the different techniques.

techniques median

q 25

q 75

median

q 25

q 75

T1 397 287 470 517 412 596

T2 28.3 20.8 34.4 40.2 35.3 48.3

DWIb50 49.2 30.3 68.2 14.1 10.5 15.4

DWIb1000 23.7 13.5 36.2 7.31 5.36 8.49

REFERENCES

Ghanem, N., Lohrmann, C., Monika Engelhardtand Gre-

gor Pache, M. U., Saueressig, U., Kotter, E., and

Langer, M. (2006). Whole-body mri in the de-

tection of bone marrow inﬁltration in patients with

plasma cell neoplasms in comparison to the radiolog-

ical skeletal survey. Eur Radiol, 16:1005–1014.

Khoo, M., Tyler, P., Saifuddin, A., and Padhani, A. (2011).

Diffusion-weighted imaging (DWI) in musculoskele-

tal MRI: a critical review. Skeletal Radiology, pages

1–17.

Koh, D.-M. and Collins, D. J. (2007). Diffusion-Weighted

MRI in the body: Applications and challenges in on-

cology. Am. J. Roentgenol., 188(6):1622–1635.

Lee, S., Wolberg, G., and Shin, S. Y. (1997). Scattered

data interpolation with multilevel b-splines. IEEE

Transactions on Visualization and Computer Graph-

ics, 3:228–244.

Lee, S., Wolberg, G., yong Chwa, K., and Shin, S. Y.

(1996). Image metamorphosis with scattered feature

constraints. IEEE Transactions on Visualization and

Computer Graphics, 2:337–354.

Padhani, A. R., Liu, G., Koh, D. M. M., Chenevert,

T. L., Thoeny, H. C., Takahara, T., Dzik-Jurasz, A.,

Ross, B. D., Van Cauteren, M., Collins, D., Ham-

moud, D. A., Rustin, G. J., Taouli, B., and Choyke,

P. L. (2009). Diffusion-weighted magnetic resonance

imaging as a cancer biomarker: consensus and recom-

mendations. Neoplasia (New York, N.Y.), 11(2):102–

125.

Rueckert, D., Sonoda, L., Hayes, C., Hill, D., Leach, M.,

and Hawkes, D. (1999). Nonrigid registration us-

ing free-form deformations: application to breast mr

images. IEEE Transactions on Medical Imaging,

18(8):712–721.

Sakurada, A., Takahara, T., Kwee, T. C., Yamashita, T.,

Nasu, S., Horie, T., Cauteren, M. V., and Imai,

Y. (2009). Diagnostic performance of diffusion-

weighted magnetic resonance imaging in esophageal

cancer. Eur Radiol, 19:1461–1469.

Schnabel, J. A., Rueckert, D., Quist, M., Blackall, J.,

Castellano-Smith, A., Hartkens, T., Penney, G., Hall,

W., Liu, H., Truwit, C., Gerritsen, F., Hill, D.,

and Hawkes, D. (2001). A generic framework for

non-rigid registration based on non-uniform multi-

level Free-Form deformations. In Niessen, W. and

Viergever, M., editors, Medical Image Computing and

Computer-Assisted Intervention

a MICCAI 2001, vol-

ume 2208 of Lecture Notes in Computer Science,

chapter 69, pages 573–581. Springer Berlin / Heidel-

berg, Berlin, Heidelberg.

Sommer, G., Klarh

A¶fer, M., Lenz, C., Schefﬂer, K., Bon-

gartz, G., and Winter, L. (2010). Signal characteristics

of focal bone marrow lesions in patients with multi-

ple myeloma using whole body T1w-TSE, T2w-STIR

and diffusion-weighted imaging with background sup-

pression. European Radiology, pages 1–6–6.

BIOSIGNALS 2012 - International Conference on Bio-inspired Systems and Signal Processing

566