Dharmesh Singh1, Saumya Diwan2, Virendra Kumar3, Chandan J Das4, Anup Singh1,5, and Amit Mehndiratta1,5
1Centre for Biomedical Engineering, Indian Institute of Technology (IIT) Delhi, New Delhi, India, 2Delhi Technological University, Delhi, India, 3Department of NMR, All India Institute of Medical Sciences Delhi, New Delhi, India, 4Department of Radiology, All India Institute of Medical Sciences Delhi, New Delhi, India, 5Department of Biomedical Engineering, All India Institute of Medical Sciences Delhi, New Delhi, India
Synopsis
Accurate estimation of tumor size is a
challenging task due to the large variation in shape of prostatic tumors. Measurement
of diameter and volume of tumor are useful for the staging of prostate cancer(PCa). Therefore, an appropriate method for tumor assessment is essential to assist
clinical management of PCa. The goal of this study was to investigate the role
of tumor size estimation methods (2D vs. 3D) for the Prostate Imaging Reporting
and Data System version-2(PI-RADS v2) assessment of PCa using multiparametric
MRI. Automatic tumor size assessment using ellipse-fitting approach showed relatively better performance of PCa assessment
compared to radiologist-assessment.
Introduction
Prostate cancer (PCa) is the fifth
leading cause of cancer death among men worldwide.1 Multiparametric-MRI(mpMRI) [T2-weighted(T2w) imaging,diffusion-weighted
imaging (DWI),its derivative apparent-diffusion-coefficient (ADC),and dynamic contrast-enhanced(DCE)
MRI, etc.] have significantly improved the diagnostic accuracy of PCa.2 PI-RADS v2 helps to identify
clinically significant PCa using mpMRI and therefore,can aid in clinical
decisions regarding treatment.3 Tumor estimation plays an essential role in automated diagnosis of PCa.4 Based on the current
uses of mpMRI,any tumor having longest-diameter>1.5cm,volume>0.5cc,and PI-RADS v2 score>3, is a high-grade cancer.4 PI-RADS v2 follows the
2D-approach for tumor marking and estimation. The creation of 3D prostate tumor
mask could assist with radiotherapy treatment planning and for correlative pathology-imaging research.5 The objective of this study was to develop method
for automated estimation of maximum tumor-dimension in both 2D and 3D;where in
3D the maximum-diameter could be in an oblique plane also. The major-axis of
best-fit ellipse in 2D and ellipsoid in 3D was used as the longest-dimension of
tumor and evaluated for automated assessment of PI-RADS v2. Methods
MRI data acquisition:
A retrospective mpMRI dataset from 18 patients of PCa was used in this study. All prostate mpMRI examinations were acquired using a 3T MRI system (Ingenia,Philips,The Netherlands)
using an external phased-array body coil. MRI sequences
included axial turbo-spin-echo(TSE) T2w(TR/TE=3715/100ms;slice-thickness=3mm;field-of-view(FOV)=160×160mm2;acquisition matrix=400×400;number-of-slices=30) and echo-planar DWI(TR/TE=5521/75
ms; FOV=177.6×177mm2; acquisition-matrix=176×176;slice-thickness=4mm; number-of-slice
=30;with four b-values of 0,500,800 and 1500s/mm2).
Methodology:
Data were processed using in-house
developed codes in MATLAB-R2018a. In the pre-processing step,prostate-gland
and zonal segmentation,6 ADC calculation, and 3D-affine registration were performed. Region of interest (ROI) of segmented prostate
zones for T2w, high b-value DWI and ADC was extracted for tumors-marking.
Tumors-marking was performed as per PI-RADS
v2 guidelines with the help of expert radiologist(>10 years of experience in
prostate imaging)
Tumor
estimation methods:
Based on the qualitative-classification,tumor
configuration was divided into two categories: ellipsoid and non-ellipsoid
shape. The tumor-size and volume were measured from 2D and 3D methods for
estimation purpose.
Radiologist-measurements:
The
size and volume were measured within ROIs outlined by the expert
radiologist on DWI and ADC maps. The analysis was done using image processing
software (ImageJ,version-1.48, National Institute of Health,Bethesda,Md).
Ellipse-fitting:
The dimension and volume of tumor were
assessed in each MRI slices by 2D ellipse-fitting approach. The major-axis of
the ellipse was computed as the longest-diameter of the tumor. Volume was
calculated by the summation of all tumor areas in each slice and multiplication
by slice-thickness.
Ellipsoid-fitting:
The 2D-axial slices were stacked to form
a 3D volume. Ellipsoid-fitting approach was used to best-fit the 3D anatomy of
the tumor in any direction;calculating the major-axis of the best-fit ellipsoid
as the longest-diameter. Volume of the ellipsoid was also calculated.
Based on tumor measurements, accuracy was calculated for PI-RADS v2
assessment of PCa compared to radiologist’s assessment. The workflow of our
proposed methodology is shown in fig-1. Results
Fig-2 shows the tumor-estimation results of one patient(age=58years).
The measurement of longest-dimension and volume of tumor using proposed 2D and
3D methods comparison with radiologist measurements using ImageJ software is
shown in table-1. The longest-dimension and volume of tumor(mean±SD) from 18
subjects were 1.34±0.40cm and 0.68±0.34cm3 by radiologist,1.46±0.3cm
and 0.70±0.30cm3 using ellipse-fitting,and 2.90±0.87cm and 0.99±0.40cm3 using
ellipsoid fitting. The percentage of variation in dimension and volume between radiologist-measurement
and ellipse-fitting were small(8.21±7.5%and 2.8±11.76%,respectively)
compared to ellipsoid-fitting measurements(53±54% and 31.3±15%,respectively). Ellipse-fitting tumor-estimation based PI-RADS
v2 assessment showed sensitivity,82%;specificity,71.43%;accuracy,77.80%
compared to radiologist-assessment(table-2). Ellipsoid-fitting tumor-estimation
based PI-RADS v2 assessment against radiologist-assessment showed large
variation in sensitivity and specificity. All the patients(n=11) with grade-4
using radiologist-assessment were graded as grade-5 with ellipsoid-fitting
approach. The variation of tumor-volume(from radiologist-measurement and
ellipse-fitting method) with PI-RADS v2 scores is shown in fig-3.Discussion
Accurate
non-invasive measurement of prostate tumor could significantly improve the
determination of tumor-prognosis and assist in PI-RADS v2 assessment. Few prostate
tumor-estimation methods have been presented in the past.7-9 In this study,2D ellipse-fitting and 3D ellipsoid-fitting
approach were developed for tumor-estimation. Expert radiologist marked the
tumor as considered as ground-truth. The tumor measurements between radiologist
and ellipse-fitting were found similar,because PI-RADS v2 follows the 2D
approach for tumor-marking and estimation. The ellipsoid-fitting method showed significantly
higher values for the tumor-dimension and volume as compared to radiologist-measurement.
In the literature few studies so far evaluate PI-RADS v2 for diagnosis of PCa;the
pooled sensitivity and specificity were 85-89% and 71-73%,respectively.10,11 Based on the
ellipse-fitting tumor-estimation method, proposed methodology has observed relatively
better sensitivity,82%;specificity,71.43%;and
accuracy,78% compared to radiological PI-RADS v2 assessment. As found in our
study with ellipsoid-fitting approach all the patients with grade-4 were marked
as grade-5,because of higher tumor-diameter. This led us to a bigger concern,
if longest tumor-diameter is in oblique plane as estimated by ellipsoid
approach then the grade of tumor can also change having potential impact on
treatment plan as well. In future study, we intend to evaluate and validate this
approach in a large number of patients towards improving the accuracy of
proposed framework.Conclusion
The
proposed automated framework based on 2D ellipse-fitting showed good agreement
with radiologist-assessment, which might help to make the analysis process
time-efficient. 3D tumor estimation can provide complementary information to
clinicians compared to traditional methods due to multidimensional visualization, which
needs further investigation. Acknowledgements
This
work is supported by IIT Delhi, India and AIIMS New-Delhi, India. DS was
supported with the research fellowship fund from the Ministry of Human Resource
Development, Government of India.References
1. Bray F, Ferlay J, Soerjomataram I,
et al. Global Cancer Statistics 2018: GLOBOCAN Estimates of Incidence and
Mortality Worldwide for 36 Cancers in 185 Countries. CA CANCER J CLIN. 2018; 68:
394-442.
2. Mertan FV, Berman R, Szajek K, et al. Evaluating the Role of
mpMRI in Prostate Cancer Assessment. Expert Review of Medical Devices. 2016;
13(2):129-141.
3. Park SY, Cho NH, et al. Prostate
Imaging-Reporting and Data System Version 2: Beyond Prostate Cancer Detection.
Korean J Radiol. 2018; 19(2):193–200.
4. Weinreb JC , Barentsz JO, Choyke PL,
et al. PI-RADS Prostate Imaging Reporting and Data System: 2015, Version 2.
European Urology. 2016; 69(1):16-40.
5. Malone
SC, Haridass A, Nyiri B, et al. Creation of 3-dimensional prostate cancer maps:
methodology and clinical and research implications. Arch Pathol Lab Med. 2014;138(6)
:803-808.
6. Singh D, Kumar V, et al. Gland and
Zonal Segmentation of Prostate using Diffusion-Weighted MR Imaging, Proc. Intl.
Soc. Mag. Reson. Med. 2019; 27.
7. Perera M, Lawrentschuk N, Bolton D,
et al. Comparison
of contemporary methods for estimating prostate tumour volume in pathological
specimens. BJU Int. 2014; 113(2):29-34.
8. Bratan F, Melodelima C, Souchon R,
et al. How
Accurate Is Multiparametric MR Imaging in Evaluation of Prostate Cancer Volume?
Genitourinary Imaging. 2014; 275(1):144-154.
9. Priester A, Natarajan S, Khoshnood P,
et al. Magnetic Resonance Imaging Underestimation of Prostate Cancer Geometry:
Use of Patient Specific Molds to Correlate Images with Whole Mount Pathology. J
Urol. 2017; 197(2): 320–326.
10. Woo
S, Suh CH, Kim SY, et al. Diagnostic performance of prostate imaging reporting
and data system version 2 for detection of prostate cancer: a systematic review
and diagnostic meta-analysis. Eur Urol. 2017; 72:177-188.
11. Zhang L, Tang M, Chen S, et al. A
meta-analysis of use of prostate imaging reporting and data system version 2
(PI-RADS V2) with multiparametric MR imaging for the detection of prostate
cancer. Eur Radiol. 2017; 27:5204-5214