Archana Vadiraj Malagi1,2, Devasenathipathy Kandasamy3, Deepam Pushpam4, Kedar Khare5, Raju Sharma3, Rakesh Kumar6, Sameer Bakhshi4, and Amit Mehndiratta1,7
1Centre for Biomedical Engineering, Indian Institute of Technology Delhi (IIT Delhi), New Delhi, India, 2Biomedical Imaging Research Institute, Cedars-Sinai Medical Center, Los Angeles, CA, United States, 3Department of Radiodiagnosis, All India Institute of Medical Sciences Delhi, New Delhi, India, 4Department of Medical Oncology, Dr. B.R. Ambedkar Institute-Rotary Cancer Hospital (IRCH), All India Institute of Medical Sciences Delhi, New Delhi, India, 5Department of Physics, Indian Institute of Technology Delhi (IIT Delhi), New Delhi, India, 6Department of Nuclear Medicine, All India Institute of Medical Sciences Delhi, New Delhi, India, 7Department of Biomedical Engineering, All India Institute of Medical Sciences Delhi, New Delhi, India
Synopsis
Keywords: Cancer, Diffusion/other diffusion imaging techniques
Intravoxel Incoherent motion-diffusion kurtosis imaging (IVIM-DKI) was
used for evaluation and characterization of malignant lymph nodes in lymphoma. A total of twenty-one (n=21) patients diagnosed
with biopsy proven Hodgkin lymphoma(HL: n=13) or non-Hodgkin lymphoma(NHL: n=8)
were prospectively evaluated. IVIM-DKI parameters were estimated using standard
IVIM-DKI model (standard model) and IVIM-DKI model with total variation(TV)
penalty function method(IDTV model). Perfusion fraction (f) and kurtosis (k)
estimated using IDTV model, and apparent diffusion coefficient(ADC) were
significantly(p<0.05) lower in malignant lymph nodes than benign lymph
nodes. f and k showed high AUC of 0.88 and 0.83, respectively for malignant vs.
benign lymph nodes.
Introduction
Intravoxel incoherent motion (IVIM) and
diffusion kurtosis imaging (DKI) analysis in oncological imaging have shown
promising results in detection, characterization, evaluation, and the
prediction of treatment response for various cancers1–3. A non-Gaussian model of water
diffusion has been shown to accurately detect tumor heterogeneity such as IVIM
with DKI (IVIM-DKI)4. There is limited literature on the
characterization of benign and malignant lymph nodes in Hodgkin lymphoma (HL)
and non-Hodgkin lymphoma (NHL) utilizing IVIM-DKI, mostly on head and neck
cancer4,5. Thus, this study aimed to
investigate the role of quantitative parameters such as diffusion coefficient
(D), pseudo-diffusion coefficient (D*), perfusion fraction (f), and kurtosis
(k) obtained from the multi-b-values IVIM-DKI model with the TV method6–8 for the characterization of benign
and malignant lymph nodes in lymphoma.Methods
Subjects: A total of 21 patients were included
in the analysis under this investigation (Age (mean±SD):34±11.4 years old,
Female/Male:4/17). As indicated in figure 1,
histopathological results revealed three different subsets of lymphoma with HL
(n=13) and NHL comprising of Diffuse large B-cell lymphoma (DLBCL:n=5), and
follicular lymphoma (n=3).
ROI
localization: The whole volume ROI was drawn
manually. Figure 2(a-l) demonstrates the ROI placement for malignant lymph
nodes in HL and NHL. Benign lymph node ROI was drawn on IVIM-DKI dataset at
b-values=0 s/mm2 where small lymph nodes displayed hyperintensity
area and SUV map indicated no FDG uptake.
IVIM-DKI acquisition and Analysis: All patients were scanned in 1.5T MRI
(Ingenia; Philips Healthcare, Netherlands) with a STIR (Thoracic:TR=1.503s and
TE=0.09s; Abdomen: TR=1.503s and TE=0.06s), including IVIM-DKI with 9b-values= 0,35,50,100,175,300,500,1500,2000 s/mm2 using
phased-array surface coil (Thoracic:TR=12.44s,TE=0.081s; Abdomen:
TR=12.44s,TE=0.081s) for thoracic and abdominal area.
The ADC of each ROI was estimated
voxelwise using a monoexponential model with the three different b-values
(0,500,1500 s/mm2).
The IVIM-DKI model for
multi-b-values:
$$ \frac{S I}{S I_0}=f e^{-b D^*}+(1-f) e^{-b D+\frac{1}{6} b^2 D^2 k} -(1) $$
where, SI and SI0 are diffusion signals with and without a diffusion gradient b in s/mm2. The IVIM-DKI images was analyzed using two
models: (1) standard model as shown in equation 1; and (2) IDTV model, which is
an IVIM-DKI model with the TV penalty function method which minimizing the
total variation of a whole 3D-parametric map6–8.
Statistics: Coefficient of variation (CV) was
calculated to compare the model performance of standard and IDTV models. Kolmogorov-Smirnov
test was used to calculate any significant differences between the IVIM-DKI
parameters of benign and malignant lymph nodes in lymphoma. Diagnostic performance of IVIM-DKI parameters were
measured using receiver operating characteristics (ROC) analysis.Results
Qualitative and Quantitative characterization of
benign and malignant lymph nodes in lymphoma using IVIM-DKI
Figure 3 shows a
representative image of a 32 year old male patient with DLBCL, stage IV, with a tumor in the
anterior chest wall. In the tumor region, a
high b-value map,
and STIR image showed a hyperintense region, as shown in Figure 3.
IDTV model showed significantly
(p<0.001) lower CV than the standard model by 42-59% in both benign and
malignant lymph nodes for all IVIM-DKI parameters.
ADC and IVIM-DKI
parameters estimated using the IDTV model were evaluated to characterize benign
and malignant lymph nodes. Figure 4(a, d, e) shows that ADC, f, and k values were
significantly (p<0.05) lower in malignant lymph nodes than in benign lymph
nodes in lymphoma. D and D* did not show significant differences in the benign and malignant lymph nodes.
Differential
diagnosis of benign and malignant lymph nodes using IVIM-DKI and ROC analysis
High diagnostic
performance was obtained for the f parameter, which showed the highest area under
curve (AUC), accuracy, and F1 score of 0.88, 86%, and 86%, with a cut-off value of 0.223
to differentiate between nodes, as shown in figure 5.
Even the k parameter showed high diagnostic
performance, with AUC, accuracy, and F1 score of 0.83, 81%, and 79%, with a cut-off value of 0.993.Discussion and Conclusion
In the present study, benign and malignant lymph
nodes in lymphoma were characterized using ADC and IVIM-DKI parameters. In
benign lymph nodes, ADC was higher than in malignant lymph nodes in the
lymphoma. ADC is measured using a conventional monoexponential function and has
shown its potential in the differentiation of benign and malignant lymph nodes3,9. However, IVIM-DKI
can produce quantitative diffusion and perfusion information with additional
kurtosis information on tumor heterogeneity at low b-values (100 s/mm2)
and high b-values (>1500 s/mm2)4,10,11. In this study, a
parametric reconstruction method such as TV was employed with the IVIM-DKI
model, which has been shown to remove any spurious values while estimating
parameters and improve the characterization of prostatic lesions6. f and k estimated
using the IDTV model were able to differentiate benign and malignant lymph
nodes, same as also seen in previous studies12. f represents blood
volume fraction, and high perfusion in benign nodes may indicate the presence
of vasodilation, increased blood flow, and high permeability13. However, k was
lower in malignant than benign lymph nodes due to the presence of inflammation
or fibrous tissues14. While f and k
showed higher AUC and accuracy compared to ADC in differentiation between
benign and malignant lymph nodes. This demonstrates that the IDTV model may
provide high diagnostic performance even when parameters are employed
individually.Acknowledgements
The authors
would like to thank SERB, the Department of Science and Technology, the
Government of India (CRG/2021/005342) for the funding support.References
- Wu, R. et al. Assessment of chemotherapy
response in non-Hodgkin lymphoma involving the neck utilizing diffusion
kurtosis imaging: a preliminary study. Diagnostic and Interventional Radiology
23, 245–249 (2017).
-
Beyhan, M., Sade, R., Koc, E., Adanur, S.
& Kantarci, M. The evaluation of prostate lesions with IVIM DWI and MR
perfusion parameters at 3T MRI. La radiologia medica 124, 87–93 (2019).
-
Lecler, A. et al. Intravoxel incoherent
motion (IVIM) 3 T MRI for orbital lesion characterization. Eur Radiol 31, 14–23
(2021).
-
Lu, Y. et al. Extension of the intravoxel
incoherent motion model to non-gaussian diffusion in head and neck cancer.
Journal of Magnetic Resonance Imaging 36, 1088–1096 (2012).
-
Sijtsema, N. D. et al. An optimal
acquisition and post‐processing pipeline for hybrid IVIM‐DKI in head and neck.
Magnetic Resonance in Medicine (2020) doi:https://doi.org/10.1002/mrm.28461.
-
Malagi, A. V. et al. IVIM–DKI for
differentiation between prostate cancer and benign prostatic hyperplasia:
comparison of 1.5 T vs. 3 T MRI. Magnetic Resonance Materials in Physics,
Biology and Medicine (2021) doi:10.1007/s10334-021-00932-1.
-
Kayal, E. B. et al. Quantitative Analysis
of Intravoxel Incoherent Motion (IVIM) Diffusion MRI using Total Variation and
Huber Penalty Function. Medical Physics 44, 5849–5858 (2017).
-
Rudin, L. I., Osher, S. & Fatemi, E.
Nonlinear total variation based noise removal algorithms. Physica D: Nonlinear
Phenomena 60, 259–268 (1992).
-
De Paepe, K. N. et al. Improving lymph node
characterization in staging malignant lymphoma using first-order ADC texture
analysis from whole-body diffusion-weighted MRI. Journal of Magnetic Resonance
Imaging 48, 897–906 (2018).
-
Le Bihan, D. et al. Separation of diffusion
and perfusion in intravoxel incoherent motion MR imaging. Radiology 168,
497–505 (1988).
- Jensen, J. H., Helpern, J. A., Ramani, A.,
Lu, H. & Kaczynski, K. Diffusional kurtosis imaging: The quantification of
non-Gaussian water diffusion by means of magnetic resonance imaging. Magnetic
Resonance in Medicine 53, 1432–1440 (2005).
-
Yu, J. et al. Diffusion kurtosis imaging in
identifying the malignancy of lymph nodes during the primary staging of rectal
cancer. Colorectal Disease 20, 116–125 (2018).
-
Xu, C. et al. Value of integrated PET-IVIM
MR in assessing metastases in hypermetabolic pelvic lymph nodes in cervical
cancer: a multi-parameter study. Eur Radiol 30, 2483–2492 (2020).
-
Liu, Y. et al. The diagnostic accuracy of
intravoxel incoherent motion and diffusion kurtosis imaging in the
differentiation of malignant and benign soft-tissue masses: which is better?
Acta Radiol 63, 785–793 (2022).