1
Computational BioMedicine Laboratory (CBML), Foundation for Research & Technology, Heraklion, Crete, Greece
2
Department of Radiology, School of Medicine,, University of Crete, Greece
3
Image Analysis Research Unit, IMT School for Advanced Studies Lucca, Italy
4
Department of Pathology, , University Hospital of Crete, Greece
5
Technological Educational Institute of Crete, Department of Informatics Engineering, Greece
6
Department of Surgical Oncology, School of Medicine, University of Crete, Greece
Corresponding author details:
Katerina Nikiforaki, CBML
Foundation for Research & Technology
N.Plastira 100 Vasilika Vouton
Crete,Greece
Copyright: © 2018 Nikiforaki K, et al. This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 international License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited
Tumor necrosis factor alpha (TNF-α) is known to cause selective destruction of tumor
neovascularisation. Perfusion changes in tumor microenvironment were studied in a young
patient with a high-grade myxoid liposarcoma after isolated limb perfusion with TNF-α
and melphalan, using advanced MR imaging. Quantitative and semi-quantitative evaluation
of perfusion metrics included ktrans/f/T2* shift and automatic extraction of enhancement
patterns. Strong agreement between different methods is consistent with response to
therapy, specifically reduced perfusion (mean ktrans shifted from 0.424 to 0.099 mL/g/min,
mean f from 33.5 to 7.3%), reduced hypoxia indication (mean T2* increased from 72 to 82
ms) and prevalence of necrotic areas.
Tumor Necrosis Factor-α; MR imaging/diagnosis; Pattern Recognition; Tumor hypoxia;
Cancer imagenomics; Liposarcoma/treatment response
(TNF-α) : Tumor Necrosis Factor Alpha
ILP : Isolated limb perfusion
DCE : Dynamic Contrast Enhanced
DWI : Diffusion Weighted Imaging
IVIM : Intravoxel Incoherent Motion
T2*r : T2* Relaxometry
BOLD : Blood Oxygen Level-Dependent
PK : Pharmacokinetic
ktrans : Transendothelial permeability
dHb : Deoxyhemoglobin
IMVD : Intratumoral Microvessel Density
FISH : Fluorescence In situ Hybridization
PR : Pattern Recognition
Isolated limb perfusion (ILP) has been used for locally advanced extremity soft tissue sarcoma in order to avoid amputation when wide excision of the tumor is not feasible [1]. Subsequently (marginal) excision of the tumor is performed 6 to 8 weeks after this regional chemotherapy treatment. A drug combination of tumor necrosis factor alpha (TNF-α) and melphalan is usually used. TNF-α is multifunctional cytokine that plays a major role in innate and acquired immunity while its binding to certain receptors leads to hemodynamic and antitumor effects [2]. The administration of TNF-α during ILP inhibits devastating systemic hemodynamic effects and shows a strong synergistic antitumor action with chemotherapeutic agents in sarcoma patients [3]. In the setting of ILP TNF-α has two distinct antitumor properties that may be related to each other: increased uptake of chemotherapeutic drugs such as melphalan into the tumor and selective destruction of tumor neovascularization [2]. In experimental studies TNF-α has led to increased vessel permeability and decreased interstitial pressure as immediate effects after administration [4,5]. These early antivascular effects lead to an up to 6-fold–increased uptake of melphalan into the tumor [6]. Late antivascular effects such as tumor vessel disintegration and endothelial apoptosis which may ultimately result in tumor necrosis are best demonstrated in imaging obtained before and after ILP with TNF-α [7].
The evaluation of tumor response to therapy by non-invasively assessing blood supply and oxygen level has been a long-standing endeavor in MR imaging (MRI). To this end clinically relevant perfusion biomarkers extracted from Dynamic Contrast Enhanced (DCE) Diffusion Weighted Imaging (DWI) and T2* relaxometry (T2*r) are indicative of various pathophysiological properties of tissue vascularity [8,9].
Specifically DCE MRI evaluates the temporal pattern of enhancement from multiple dynamic acquisitions of high temporal resolution. A recent semi-quantitative analysis framework relies on a pattern recognition (PR) method namely BU-NMF [10] which extracts automatically the principal enhancement patterns that describe DCE data and assigns a certain enhancement profile to each image pixel. These patterns can be labeled as necrotic hypoxic and well-perfused and were associated with the three most common enhancement patterns found in tumors [11] which are described by steady and subtle enhancement (Type 1) by delayed enhancement and some wash-out (Type 2) and by rapid enhancement and wash-out (Type 3). Additionally quantitative analysis based on pharmacokinetic (PK) models [12] provides estimates of transfer rates of the contrast agent from plasma to tissue (transendothelial permeability-ktrans).
However the DCE is not the only MRI method providing metrics indicative of vascularity. Intravoxel incoherent motion (IVIM) model of DWI may separate microcirculatory from thermal diffusion effects and can be used to study microcirculatory blood flow properties by estimating the microperfusion fraction f [13] i.e. the fraction of DWI signal arising from incoherent blood flow motion.
A third MRI method related to blood supply is the Blood Oxygen Level-Dependent (BOLD) which indirectly measures the total amount of deoxyhemoglobin (dHb) levels in a voxel. Conceptually the paramagnetic nature of deoxyhemoglobin as opposed to oxyhemoglobin accelerates T2* relaxation shortening thus T2* constants of tissue around the blood vessels. Importantly dHb levels depend on a number of concurrent phenomena such as blood flow blood oxygenation vasculature hemoglobin levels etc. [14] Because of the complex interplay between blood supply and oxygen extraction changes in oxygen delivery (blood flow) are not directly related to changes of oxygen consumption and dHb changes do not always match the expected action of the vascular stimulus. T2*r can provide information regarding tumor microenvironment as it reflects tissue oxygen bioavailability. Larger T2* tissue constants as a result of lower dHb concentration may indicate lower blood flow lower tissue oxygenation deficient vascular network etc. and is consistent with tumor response to therapy [9].
Although based on different grounds all three techniques have
been widely used for tumor characterization. To our knowledge MRI
analysis of TNF induced perfusion have only been studied in animals
[15] and have shown changes in endothelial permeability after
therapy as measured by perfusion biomarkers. This case study shows
the vascularity changes in a lower limb sarcoma after ILP with TNF-α
based on quantitative and semi-quantitative MRI methods.
A 40-year-old patient was referred for MRI scan of the left foot. Six months prior to imaging he had suffered from local pain during walking and had noticed a gradually growing tumor at its dorsolateral site. X-Ray showed a soft tissue mass between the 4th and the 5th metatarsal bones of the right left foot with bone remodeling and deformity (Figure 1). The mass was painless and slowly growing. Advanced MRI was performed before any biopsy. Since an ultrasoundguided core needle biopsy was inconclusive an open biopsy from the dorsal site of the soft tissue tumor was performed. At histological examination of the macroscopically grey to white color tissue with a myxoid-fleshy consistency the lesion presented a subtle multinodular conformation exhibited areas of variable cellular density with enhancement of cellularity at the periphery and was composed of bland fusiform or round cells suspended individually in a myxoid matrix with a prominent network of arborizing thin-walled capillary vessels often in a “chicken-footprint” configuration (Figure 2). Pools of extracellular mucin with cellular condensation at the rim were focally present. Mitotic figures were absent and lipoblasts were not identified. Immunohistochemical staining was positive for Vimentin only. Based on rearrangement of DDIT-3 gene which was detected using Fluorescence in Situ Hybridization (FISH)the histopathological diagnosis of high grade myxoid liposarcoma was made.
CT of the lungs revealed small nodular lesions which were considered negative for metastatic disease at the following PET-CT.
In order to avoid amputative surgery the patient underwent ILP from the popliteal vessels with 1 mg TNF-α (administered at time 0’) and 65 mg (10 mg/L limb volume administered at time 30’) melphalan for a total of 90 minutes. The postoperative course was uneventful and gradual reduction and softening of the tumor were observed. Two months after ILP and before the planned resection of the tumor MRI demonstrated reduction of the tumor size (Figure 3) whereas advanced MRI was performed for research purposes (Figure 4). A new CT of the lungs demonstrated the presence of multiple lung metastases. Since the patient was asymptomatic the planned resection of the tumor was abandoned and systemic chemotherapy was initiated.
Figure 1: Magnified oblique view of the left foot, shows a soft
tissue mass (asterisk) with displacement and remodelling of the
4th metatarsal (arrow)
Figure 2: CD34 stain highlights the thin-walled mostly arborizing vasculature (magnification X200)
Baseline perfusion study showed strong enhancement (Figure 3a) and further post processing with in-house built software [16,17] evaluated high ktrans (mean 0.424 mL/g/min) (Figure 4a) and f values (mean 33.5%) indicating functional vessel network to the neoplasm. The semi-quantitative PR results show that before therapy there is prevalence of the hypoxic enhancement pattern i.e. 59.02% (Figures. 5a, 6). T2* relaxometry showed a mean value for the 3D tumor ROI of 72 ms.
MRI findings before therapy were correlated with
immunohistochemistry results (CD34) regarding intratumoral microvessel density (iMVD). The CD34-stained neoplastic vasculature
consisted mostly in fine arborizing vessels with diameter greater than
06 mm and newly formed vessels of one to three endothelial cells.
Vessel rich “hot spots” areas were located in the tumor periphery
and interestingly correlated closely with the areas having the highest
tumor cell proliferation activity (Figure 2). Two methods were used
to assess iMVD: a) microvessel counting [18] and b) the relative
microvessel area estimate (‘Chalkley count’) [19]. The absolute
number of the vessel count obtained in the three fields x200 was
290 282 and 264 respectively resulting in a mean of 280.3 vessels.
Chalkley grid score of each hot spot was 14 18 and 16 respectively
resulting in a mean Chalkley count of 16.
Figure 3: Baseline and follow-up MR imaging
Fat suppressed contrast enhanced sagittal T1-w MR image a) before TNF therapy and b) after TNF therapy, showing the soft tissue mass
with intense and in homogenous enhancement (arrows in a) and response to treatment (arrows in b) with regard to both the size of the
lesion and the degree of enhancement.
Figure 4: Screenshot from in-house built software platform for DCE longitudinal analysis [16]
Pixel-based parametric (ktrans) maps and corresponding histograms from a single slice of the tumor a) before and b) after TNF therapy
Since follow up MRI was performed two months after therapy
this study focused on the late vascularity changes such as tumor
vessel disintegration and endothelial apoptosis. Identical protocol
showed minimal foci of enhancement and reduced number of pixels
(873 before vs 618 pixels assigned as tumor after therapy counted
in 128x128 DWI images) in the ROI volume (Figure 3b). The semi quantitative approach showed that after therapy the necrotic
enhancement pattern is dominant representing a percentage of 97%
of image pixels (Figures. 5b,6). All measured ktrans metrics [16] show
remarkably reduced perfusion in the whole tumor volume (mean
was reduced from 0.424 to 0.099 mL/g/min median from 0.349 to
0.06 mL/g/min and 95th percentile from 0.948 to 0.326 mL/g/min)
(Figures 4b,7a) strongly indicating severe impairment of the vessel
network. Reduced blood flow was also confirmed by decrease of the f
metric (mean was reduced from 33.5 to 7.3% median from 33.8 to 5%
and 95th percentile from 34.6 to 20.1%) (Figure 7b) [17]. However
perfusion changes alone do not suffice to indicate areas of deprived
oxygen supply which are important for patient management and
assessment of response to therapy. To this end T2* relaxometry was
additionally performed. In particular T2* relaxation became slower
after therapy (mean was increased from 72 to 82ms median from 61
to 62ms and 95th percentile from 164 to 214ms) implying decreased
levels of deoxyhemoglobin (Figure 7c). This finding is in accordance
with the aforementioned findings derived from both DCE-MRI and
DW-MRI.
Figure 5: Semi-quantitative results before TNF therapy
a) classification of a single slice of the tumor with the different enhancement patterns, i.e. well-perfused (green), hypoxic (blue) and necrotic (red), identified using pattern recognition techniques
b) plots of the three patterns and corresponding percentage contribution of each one to the whole tumor. The hypoxic pattern (blue) is
the principal pattern before TNF therapy
Figure 6: Semi-quantitative results after TNF therapy
a) classification of a single slice of the tumor with the different enhancement patterns, i.e. well-perfused (green), hypoxic (blue) and necrotic (red), identified using pattern recognition techniques and
b) plots of the three patterns and corresponding percentage contribution of each
one to the whole tumor. The necrotic pattern is dominant after TNF therapy.
Figure 7: Pixel based histogram analysis before (red) and after (green)
TNF therapy from 3-dimensional tumor ROI. a) ktrans (mL/g/min), b) f ( *100%), c)T2* (ms).
Quantitative MRI is a powerful tool to assess changes in tumor
microenvironment after therapeutic interventions. Although there
is no absolute threshold to characterize tumor proliferation hypoxia
or necrosis relative changes of perfusion metrics are indicative of
changes in vascular permeability oxygen consumption and vascularity
and are therefore often used to assess response after therapy. Threedimensional ROI histogram analysis permits to analyze the whole
tumor volume without the need for spatial co-registration between
different time points. Mean value shifts in all biomarkers (ktrans f T2*)
stemming from different methods (DCE DWI T2* respectively) are in
accordance with the change of the characteristic dominant perfusion
curve of the tumor volume supporting the hypothesis of response to
therapy. All metrics compose a complete multifaceted analysis which
supports satisfactorily the radiological interpretation.
To conclude this study enabled us to obtain an objective evaluation
of the non-invasive assessment of blood supply and oxygen delivery
post TNF-a therapy using a combination of DCE DWI and T2*r MRI
related biomarkers. To our knowledge no study combining more than
one perfusion monitoring techniques was found in the recent and
older literature in animal or patient population.
Copyright © 2020 Boffin Access Limited.