Deep Learning for Multi‐sequence MRI Lung Segmentation

Private
Public

Article Title and Summary

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

Read more

Implementable Deep Learning for Multi‐sequence Proton MRI Lung Segmentation: A Multi‐center, Multi‐vendor, and Multi‐disease Study

  • 123

Publications
Take a look

Deep Learning for Multi‐sequence MRI Lung Segmentation

Implementable Deep Learning for Multi‐sequence Proton MRI Lung Segmentation: A Multi‐center, Multi‐vendor, and Multi‐disease Study

by Joshua R. Astley BEng, Alberto M. Biancardi PhD, Paul J. C. Hughes PhD, Helen Marshall PhD, Guilhem J. Collier PhD, Ho-Fung Chan PhD, Laura C. Saunders PhD, Laurie J. Smith PhD, Martin L. Brook MSc, Roger Thompson PhD, Sarah Rowland-Jones MD, Sarah Skeoch PhD, Stephen M. Bianchi PhD, Matthew Q. Hatton MD, Najib M. Rahman DPhil, Ling-Pei Ho PhD, Chris E. Brightling PhD, Louise V. Wain PhD, Amisha Singapuri BSc, Rachael A. Evans PhD, Alastair J. Moss PhD, Gerry P. McCann MD, Stefan Neubauer MD, Betty Raman DPhil, C-MORE/PHOSP-COVID Collaborative Group, Jim M. Wild PhD, Bilal A. Tahir PhD on behalf of the TRISTAN Consortium


 Journal of Magnetic Resonance Imaging 58, Nr. 4 (Oktober 2023): 1030–44. doi: 10.1002/jmri.28643

Abstract

Background
Recently, deep learning via convolutional neural networks (CNNs) has largely superseded conventional methods for proton (1H)-MRI lung segmentation. However, previous deep learning studies have utilized single-center data and limited acquisition parameters.

Purpose
Develop a generalizable CNN for lung segmentation in 1H-MRI, robust to pathology, acquisition protocol, vendor, and center.

Study type
Retrospective.

Population
A total of 809 1H-MRI scans from 258 participants with various pulmonary pathologies (median age (range): 57 (6–85); 42% females) and 31 healthy participants (median age (range): 34 (23–76); 34% females) that were split into training (593 scans (74%); 157 participants (55%)), testing (50 scans (6%); 50 participants (17%)) and external validation (164 scans (20%); 82 participants (28%)) sets.

Field Strength/Sequence
1.5-T and 3-T/3D spoiled-gradient recalled and ultrashort echo-time 1H-MRI.

Assessment
2D and 3D CNNs, trained on single-center, multi-sequence data, and the conventional spatial fuzzy c-means (SFCM) method were compared to manually delineated expert segmentations. Each method was validated on external data originating from several centers. Dice similarity coefficient (DSC), average boundary Hausdorff distance (Average HD), and relative error (XOR) metrics to assess segmentation performance.

Statistical Tests
Kruskal–Wallis tests assessed significances of differences between acquisitions in the testing set. Friedman tests with post hoc multiple comparisons assessed differences between the 2D CNN, 3D CNN, and SFCM. Bland–Altman analyses assessed agreement with manually derived lung volumes. A P value of <0.05 was considered statistically significant.

Results
The 3D CNN significantly outperformed its 2D analog and SFCM, yielding a median (range) DSC of 0.961 (0.880–0.987), Average HD of 1.63 mm (0.65–5.45) and XOR of 0.079 (0.025–0.240) on the testing set and a DSC of 0.973 (0.866–0.987), Average HD of 1.11 mm (0.47–8.13) and XOR of 0.054 (0.026–0.255) on external validation data.

Data Conclusion
The 3D CNN generated accurate 1H-MRI lung segmentations on a heterogenous dataset, demonstrating robustness to disease pathology, sequence, vendor, and center.

Deep Learning for Multi‐sequence MRI Lung Segmentation
Article categories

New Long-Acting [89 Zr]Zr-DFO GLP-1 PET Tracers

Private
Public

Article Title and Summary

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

Read more

New Long-Acting [89Zr]Zr-DFO GLP-1 PET Tracers with Increased Molar Activity and Reduced Kidney Accumulation

  • 123

Publications
Take a look

New Long-Acting [89 Zr]Zr-DFO GLP-1 PET Tracers

New Long-Acting [89 Zr]Zr-DFO GLP-1 PET Tracers with Increased Molar Activity and Reduced Kidney Accumulation

by Wilbs, Jonas, René Raavé, Milou Boswinkel, Tine Glendorf, David Rodríguez, Eduardo Felipe Alves Fernandes, Sandra Heskamp, Inga Bjørnsdottir, und Magnus B. F. Gustafsson on behalf of the TRISTAN Consortium


Journal of Medicinal Chemistry 66, Nr. 12 (22. Juni 2023): 7772–84. doi: 10.1021/acs.jmedchem.2c02073

Abstract

Positron emission tomography (PET) imaging is used in drug development to noninvasively measure biodistribution and receptor occupancy. Ideally, PET tracers retain target binding and biodistribution properties of the investigated drug. Previously, we developed a zirconium-89 PET tracer based on a long-circulating glucagon-like peptide 1 receptor agonist (GLP-1RA) using desferrioxamine (DFO) as a chelator. Here, we aimed to develop an improved zirconium-89-labeled GLP-1RA with increased molar activity to increase the uptake in low receptor density tissues, such as brain. Furthermore, we aimed at reducing tracer accumulation in the kidneys. Introducing up to four additional Zr-DFOs resulted in higher molar activity and stability, while retaining potency. Branched placement of DFOs was especially beneficial. Tracers with either two or four DFOs had similar biodistribution as the tracer with one DFO in vivo, albeit increased kidney and liver uptake. Reduced kidney accumulation was achieved by introducing an enzymatically cleavable Met-Val-Lys (MVK) linker motif between the chelator and the peptide.

New Long-Acting [89 Zr]Zr-DFO GLP-1 PET Tracers
Article categories

Nuclear Imaging to Quantify Tumor CD8+ T-Cell Infiltration

Private
Public

Article Title and Summary

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

Read more

Head-to-Head Comparison of Nuclear Imaging Approaches to Quantify Tumor CD8+ T-Cell Infiltration

  • 123

Publications
Take a look

Nuclear Imaging to Quantify Tumor CD8+ T-Cell Infiltration

Head-to-Head Comparison of Nuclear Imaging Approaches to Quantify Tumor CD8+ T-Cell Infiltration

by Gerwin G.W. Sandker,  René Raavé,  Inês F. Antunes, Milou Boswinkel,  Lenneke Cornelissen, Gerben M. Franssen,  VJanneke Molkenboer-Kuenen,  Peter J. Wierstra,  Iris M. Hagemans,  Erik F.J. de Vries,  Johan Bussink,  Gosse Adema,  Martijn Verdoes,  Erik H.J.G. Aarntzen,  Sandra Heskamp on behalf of the TRISTAN Consortium


Immunology, 15. Oktober 2024. doi: 10.1101/2024.10.13.618082.

Abstract

Many immunotherapies focus on (re)invigorating CD8+ T cell anti-cancer responses and different nuclear imaging techniques have been developed to measure CD8+ T cell distributions. In vivo labeling approaches using radiotracers primarily show CD8+ T cell distributions, while ex vivo labeled CD8+ T cells can show CD8+ T cell migration patterns, homing, and tumor infiltration. Currently, a comprehensive head-to-head comparison of in vivo and ex-vivo cell labeling with respect to their tumor and normal tissue targeting properties and correlation to the presence of CD8+ T cells is lacking, yet essential for correct interpretation of clinical CD8+ imaging applications. Therefore, we performed a head-to-head comparison of three different CD8+ T cell imaging approaches: 1) 89Zr-labeled DFO-conjugated Fc-silent anti-CD8 antibody ([89Zr]Zr-anti-CD8-IgG2asilent), 2) ex vivo 89Zr-oxine labeled ovalbumin-specific CD8+ T cells ([89Zr]Zr-OT-I cells), and 3) 18F-labeled IL2 ([18F]AlF-RESCA-IL2).

Methods B16F10/OVA tumor-bearing C57BL/6 mice (n=10/group) received intravenously one of the three radiopharmaceuticals. PET/CT images were acquired starting 72 h ([89Zr]Zr-anti-CD8-IgG2asilent), 24 and 48 h ([89Zr]Zr-OT-I cells), and 10 min ([18F]AlF-RESCA-IL2) post injection. Subsequently, ex vivo biodistribution analysis of the radiopharmaceuticals was performed followed by flow cytometric analysis to evaluate the number of intratumoral CD8+ T cells. Additionally, the intratumoral radiolabel distributions was assessed by autoradiography and immunohistochemistry (IHC) on tumor slices.

Results [89Zr]Zr-anti-CD8-IgG2asilent, [89Zr]Zr-OT-I cells, and [18F]AlF-RESCA-IL2 showed uptake in CD8-rich tissues, with preferential targeting to the spleen. Biodistribution analysis showed tumor uptake above blood level for all radiopharmaceuticals, except [18F]AlF-RESCA-IL2. For all three approaches, the uptake in the tumor-draining lymph node was significantly higher compared with the contralateral axial lymph node, suggesting that all approaches allow evaluation of immune responses involving CD8+ T cells. Tumor uptake of [89Zr]Zr-anti-CD8-IgG2asilent (R2=0.65, p<0.01) and [89Zr]Zr-OT-I cells (R2=0.74, p<0.01) correlated to the number of intratumoral CD8+ T cells (flow cytometry). The intratumoral distribution pattern of the radiosignal was different for ex vivo and in vivo radiolabeling techniques. The short half-life of 18F precluded autoradiography assessment of [18F]AlF-RESCA-IL2.

Conclusion We show that [89Zr]Zr-anti-CD8-IgG2asilent and [89Zr]Zr-OT-I cells PET/CT imaging can be used to evaluate intratumoral CD8+ T cells, even though their normal tissues and intratumoral distribution patterns are significantly different. Based on their characteristics, [89Zr]Zr-anti-CD8-IgG2asilent might be most useful to immunophenotyping the TME, while the ex vivo cell labeling approach visualizes CD8+ T cell migrations patterns and the permissiveness of tumors for invasion, whereas [18F]AlF-RESCA-IL2 allows for rapid recurrent imaging and might prove useful for tracking rapid changes in CD8+ T cell distributions. In conclusion, our head-to-head comparison of the three prototype CD8+ T cell labeling approaches provides new insights which can aid in correct interpretation of clinical CD8 imaging and may guide in the selection of the optimal imaging approach for the research question of interest.

Nuclear Imaging to Quantify Tumor CD8+ T-Cell Infiltration
Article categories

Longitudinal Changes in a Bleomycin Rat Model by DCE-MRI

Private
Public

Article Title and Summary

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

Read more

Contrast Enhanced Longitudinal Changes Observed in an Experimental Bleomycin-Induced Lung Fibrosis Rat Model by Radial DCE-MRI at 9.4T

  • 123

Publications
Take a look

Longitudinal Changes in a Bleomycin Rat Model by DCE-MRI

Contrast Enhanced Longitudinal Changes Observed in an Experimental Bleomycin-Induced Lung Fibrosis Rat Model by Radial DCE-MRI at 9.4T

by In ‘T Zandt, René, Irma Mahmutovic Persson, Marta Tibiletti, Karin Von Wachenfeldt, Geoff J. M. Parker, Lars E. Olsson on behalf of the TRISTAN Consortium


PLOS ONE 19, Nr. 9 (27. September 2024): e0310643. doi: 10.1371/journal.pone.0310643

Abstract

Identifying biomarkers in fibrotic lung disease is key for early anti-fibrotic intervention. Dynamic contrast-enhanced (DCE) MRI offers valuable perfusion-related insights in fibrosis but adapting human MRI methods to rodents poses challenges. Here, we explored these translational challenges for the inflammatory and fibrotic phase of a bleomycin lung injury model in rats. Eleven male Sprague-Dawley rats received a single intratracheal dose of bleomycin (1000iU), four control rats received saline. Imaging was performed on days 7 and 28 post-induction. Ultra-short echo time imaging was used to image the lung for 7 minutes after which Clariscan was injected intravenously. Lung signal changes were measured for an additional 21 minutes. Images were reconstructed with a sliding-window approach, providing a temporal resolution of 10 seconds per image. After imaging on day 28, animals were euthanized, and lungs were collected for histology. Bleomycin-exposed rats initially exhibited reduced body weight, recovering to control levels after 20 days. Lung volume increased in bleomycin animals from 4.4±0.9 ml in controls to 5.5±0.5 ml and 6.5±1.2 ml on day 7 and 28. DCE-MRI showed no change of initial gradient of relative enhancement in the curves between controls and bleomycin animals on day 7 and 28 post-induction. On day 7, the DCE-MRI washout phase in bleomycin animals had higher signals than the saline group and than observed at a later time point. Lung pixels were binned in 7 enhancement classes. On day 28, the size of low relative enhancement bins almost doubled in volume compared to controls and animals on day 7 post-induction. Histology on day 28 suggests that findings could be explained by changes in lung tissue density due to lung volume increase. Adapting this clinical MRI method to rodents at 9.4T remains a challenge. Future studies may benefit from lower field strength MRI combined with higher temporal resolution DCE-MRI.

Longitudinal Changes in a Bleomycin Rat Model by DCE-MRI
Article categories

Assessing Hepatic Transporter Function in Rats

Private
Public

Article Title and Summary

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

Read more

Assessment of Hepatic Transporter Function in Rats Using Dynamic Gadoxetate-Enhanced MRI: A Reproducibility Study

  • 123

Publications
Take a look

Assessing of Hepatic Transporter Function in Rats

Assessment of Hepatic Transporter Function in Rats Using Dynamic Gadoxetate-Enhanced MRI: A Reproducibility Study

by Gunwhy, Ebony R., Catherine D. G. Hines, Claudia Green, Iina Laitinen, Sirisha Tadimalla, Paul D. Hockings, Gunnar Schütz, J. Gerry Kenna, Steven Sourbron, und John C. Waterton on behalf of the TRISTAN Consortium


Magnetic Resonance Materials in Physics, Biology and Medicine 37, Nr. 4 (6. August 2024): 697–708. doi: 10.1007/s10334-024-01192-5

Abstract

Objective
Previous studies have revealed a substantial between-centre variability in DCE-MRI biomarkers of hepatocellular function in rats. This study aims to identify the main sources of variability by comparing data measured at different centres and field strengths, at different days in the same subjects, and over the course of several months in the same centre.

Materials and methods
13 substudies were conducted across three facilities on two 4.7 T and two 7 T scanners using a 3D spoiled gradient echo acquisition. All substudies included 3–6 male Wistar-Han rats each, either scanned once with vehicle (n = 76) or twice with either vehicle (n = 19) or 10 mg/kg of rifampicin (n = 13) at follow-up. Absolute values, between-centre reproducibility, within-subject repeatability, detection limits, and effect sizes were derived for hepatocellular uptake rate (Ktrans) and biliary excretion rate (kbh). Sources of variability were identified using analysis of variance and stratification by centre, field strength, and time period.

Results
Data showed significant differences between substudies of 31% for Ktrans (p = 0.013) and 43% for kbh (p < 0.001). Within-subject differences were substantially smaller for kbh (8%) but less so for Ktrans (25%). Rifampicin-induced inhibition was safely above the detection limits, with an effect size of 75 ± 3% in Ktrans and 67 ± 8% in kbh. Most of the variability in individual data was accounted for by between-subject (Ktrans = 23.5%; kbh = 42.5%) and between-centre (Ktrans = 44.9%; kbh = 50.9%) variability, substantially more than the between-day variation (Ktrans = 0.1%; kbh = 5.6%). Significant differences in kbh were found between field strengths at the same centre, between centres at the same field strength, and between repeat experiments over 2 months apart in the same centre.

Discussion
Between-centre bias caused by factors such as hardware differences, subject preparations, and operator dependence is the main source of variability in DCE-MRI of liver function in rats, closely followed by biological between-subject differences. Future method development should focus on reducing these sources of error to minimise the sample sizes needed to detect more subtle levels of inhibition.

 

Assessing Hepatic Transporter Function in Rats
Article categories

Translation from rats to humans

Private
Public

Article Title and Summary

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

Read more

Translation from rat to human: issues and current status (conference abstract)

  • 123

Publications
Take a look

Translation from rat to human

Translation from rat to human: issues and current status

by Steven Sourbron


HTN Meeting 2022

Abstract

The TRISTAN project has current completed all preclinical technical and biological validation work, and has started the process of translating methods from rats to humans. While similar methods exist and have been deployed in other liver applications, the particular focus on DILI and DDI’s poses technical constraints that require new solutions. In particular, the aim to characterize uptake rates of Gadoxetate into cells as well as the slow excretion from cells into bile requires long acquisitions – especially in the presence of strongly inhibited uptake and excretion. The aim of this talk is to sketch out the trajectory of clinical translation in TRISTAN so far and present detail on challenges encountered. 

One issue in human quantitative MRI in general is the large heterogeneity of clinical scanners by multiple vendors, the lack of standardisation in hardware and software and the fact that these are devices optimized for qualitative rather than quantitative imaging. This is in contrast to preclinical MRI scanners which are scientific instruments by definition and at higher fields are currently supplied by one single vendor. In order to estimate and help understand the impact of this scanner heterogeneity under real-world conditions of today, the first step towards clinical translation was a multi-vendor multi-site repeatability and reproducibility study of quantitative MRI in the absence of Gadoxetate [Tadimalla et al JMRI 2022]. The methodology, results and conclusions of this study will be reviewed briefly, as well as the practical experience of running them. 

The second stage towards clinical translation of the assay is an experimental medicine study in healthy volunteers, aiming to (1) derive benchmark values for Gadoxetate uptake and extraction under normal conditions and in the presence of a strong inhibitor (rifampicin), and (2) demonstrate that the effect of a strong inhibitor can be characterized reliably and consistently across subjects. Because these are the first Gadoxetate-enhanced dynamic data in humans, the study has an adaptive design allowing in the initial stages for modifications in the methodology should the data indicate the need to do so. Initially the dose of Gadoxetate for these studies is chosen at the lowest feasible value, and the study design allows this too to be stepped up if noise levels are deemed too high for reliable quantification. In this second part of the talk, I will present our first experience with the adaptive stage of the study, showing methodology and results in 3 volunteers with and without rifampicin.  

We will conclude the talk with an outlook on clinical studies that are planned in the next stage of the development.

Translation from rat to human
Article categories

Assess liver transporters in rats

Private
Public

Article Title and Summary

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

Read more

Assessment of hepatic transporter function in rats using dynamic gadoxetate-enhanced MRI: A reproducibility study (conference abstract)

  • 123

Publications
Take a look

Assess reproducibility of liver transporter function

Assessment of hepatic transporter function in rats using dynamic gadoxetate-enhanced MRI: A reproducibility study 

by Catherine D. G. Hines, Sirisha Tadimalla, Claudia Green, Iina Laitinen, Ebony R. Gunwhy, Steven Sourbron, Issam Ben Khedhiri, Paul D. Hockings, Gunnar Schütz, John C. Waterton


HTN Meeting 2022.

Abstract

Background: Drug-induced perturbations of liver transporter fluxes contribute to both drug-induced liver injury and drug-drug interactions, which are significant problems in healthcare and in drug development. Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) using gadoxetate has been proposed for assessing liver transporter-mediated drug injury, with compartmental modelling yielding gadoxetate hepatic plasma clearance (Ktrans) and biliary efflux (kbh) rate constants as biomarkers.  

Purpose/Hypothesis: Quantify the reproducibility and repeatability of gadoxetate Ktrans and kbh in the absence of drugs and investigate robustness by comparing with the effect size of a potent inhibitor, as measured by the TRISTAN rat assay. 

Study Type: data collected from five retroprospective and eight prospective longitudinal substudies. 

Population/Subjects/Phantom/Specimen/Animal Model: 76 male Wistar-Han rats. 

Field Strength/Sequence: Two 4.7T and two 7T Bruker (Rheinstetten, Germany) scanners at three facilities using a T2-weighted (T2W) spin echo sequence for anatomy identification and a retrospectively triggered 3D Fast Low Angle Shot (FLASH) RF-spoiled gradient echo T1W acquisition. 

Assessment: 13 substudies covering three centres, two MRI field strengths, three time periods, and two substances were assessed (ndatasets=108). All 13 substudies included between three to eight rats either scanned once (baseline: Day 1) with saline or study-specific vehicle (nrats=76) or twice (follow-up, 2-7 days apart: Day 2) with saline (nrats=19) or 10 mg/kg of the strong inhibitor rifampicin (nrats=13). 

Methods: Images were analysed using a tracer kinetic (TK) two-compartment exchange model that characterises Ktrans and kbh kinetics of gadoxetate using liver ROIs with a standardised arterial input function derived from a simplified model of the rat circulation. Average Ktrans and kbh values from each study along with 95% confidence intervals were reported as the TRISTAN rat assay. From the assay, reproducibility (between-substudies) and repeatability (between-day) errors were quantified for saline data, only. Reproducibility errors were then deconstructed across centres, field strengths, and time periods to examine the relative impact of different variables. Effect sizes were calculated from data where a follow-up scan of rifampicin was acquired. One-way ANOVAs and paired T-tests were also performed, where p<0.05 was considered to be statistically significant. 

Results: Reproducibility errors were 31% and 43% for Ktrans and kbh. Differences between substudies were significant. When isolating variables, reproducibility errors were as follows for choice of (i) centre: Ktrans<26% (p=0.13), kbh<93% (p=0.03); (ii) field strength: Ktrans<16% (p=0.51), kbh<84% (p=0.34); (iii) time period: Ktrans<29% (p=0.35), kbh<54% (p=0.008). Differences between baseline and follow-up saline data were not significant, with repeatability errors (Ktrans=14+/-2%; kbh=7+/-12%) much smaller than reproducibility errors. Rifampicin significantly reduced Ktrans (-170+/-8%) and kbh (-130+/-23%) across all centres. 

Conclusion: The TRISTAN rat assay is sufficiently robust for quantifying inhibition levels over >50% in absolute or relative terms. This safely includes potent inhibitors like rifampicin (>130% inhibition). For weaker inhibitors (20%-50% inhibition) only relative changes can be measured reliably. Inhibition levels below 20% cannot be quantified. This would require further technical development to reduce the uncertainty caused by the choice of centre, field strength, and drifts over time. 

Assess liver transporters in rats: reproducibility
Article categories

DL for lung cavity estimation from Xe and H MRI

Private
Public

Article Title and Summary

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

Read more

A Dual-Channel Deep Learning Approach for Lung Cavity Estimation From Hyperpolarized Gas and Proton MRI

  • 123

Publications
Take a look

DL for lung cavity estimation from Xe- and H-MRI

A Dual-Channel Deep Learning Approach for Lung Cavity Estimation From Hyperpolarized Gas and Proton MRI

by Joshua R. Astley BEng, Alberto M. Biancardi PhD, Helen Marshall PhD, Paul J. C. Hughes PhD, Guilhem J. Collier PhD, Laurie J. Smith PhD, James A. Eaden PhD, Rod Hughes MD, Jim M. Wild PhD, Bilal A. Tahir PhD


JMRI (2022). doi: 10.1002/jmri.28519

Abstract

Background
Hyperpolarized gas MRI can quantify regional lung ventilation via biomarkers, including the ventilation defect percentage (VDP). VDP is computed from segmentations derived from spatially co-registered functional hyperpolarized gas and structural proton (1H)-MRI. Although acquired at similar lung inflation levels, they are frequently misaligned, requiring a lung cavity estimation (LCE). Recently, single-channel, mono-modal deep learning (DL)-based methods have shown promise for pulmonary image segmentation problems. Multichannel, multimodal approaches may outperform single-channel alternatives.

Purpose
We hypothesized that a DL-based dual-channel approach, leveraging both 1H-MRI and Xenon-129-MRI (129Xe-MRI), can generate LCEs more accurately than single-channel alternatives.

Study Type
Retrospective.

Population
A total of 480 corresponding 1H-MRI and 129Xe-MRI scans from 26 healthy participants (median age [range]: 11 [8–71]; 50% females) and 289 patients with pulmonary pathologies (median age [range]: 47 [6–83]; 51% females) were split into training (422 scans [88%]; 257 participants [82%]) and testing (58 scans [12%]; 58 participants [18%]) sets.

Field Strength/Sequence
1.5-T, three-dimensional (3D) spoiled gradient-recalled 1H-MRI and 3D steady-state free-precession 129Xe-MRI.

Assessment
We developed a multimodal DL approach, integrating 129Xe-MRI and 1H-MRI, in a dual-channel convolutional neural network. We compared this approach to single-channel alternatives using manually edited LCEs as a benchmark. We further assessed a fully automatic DL-based framework to calculate VDPs and compared it to manually generated VDPs.

Statistical Tests
Friedman tests with post hoc Bonferroni correction for multiple comparisons compared single-channel and dual-channel DL approaches using Dice similarity coefficient (DSC), average boundary Hausdorff distance (average HD), and relative error (XOR) metrics. Bland–Altman analysis and paired t-tests compared manual and DL-generated VDPs. A P value < 0.05 was considered statistically significant.

Results
The dual-channel approach significantly outperformed single-channel approaches, achieving a median (range) DSC, average HD, and XOR of 0.967 (0.867–0.978), 1.68 mm (37.0–0.778), and 0.066 (0.246–0.045), respectively. DL-generated VDPs were statistically indistinguishable from manually generated VDPs (P = 0.710).

Data Conclusion
Our dual-channel approach generated LCEs, which could be integrated with ventilated lung segmentations to produce biomarkers such as the VDP without manual intervention.

DL for lung cavity estimation from Xe and H MRI
Article categories

DIILD mouse model

Private
Public

Article Title and Summary

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

Read more

Translational chronic drug-induced ILD mouse model, characterized by low-grade inflammation, fibrosis and dilated large airways (conference abstract)

  • 123

Publications
Take a look

DIILD mouse model

Translational chronic drug-induced ILD mouse model, characterized by low-grade inflammation, fibrosis and dilated large airways

by I Mahmutovic Persson, J Liu, R In 'T Zandt, N Fransén Petterson, A Örbom, H Falk-Håkansson, C Carvalho, K Von Wachenfeldt, L E Olsson


European Respiratory Journal 2022 60: 581 (conference abstract). doi: 10.1183/13993003.congress-2022.581

Abstract

Many systemically administrated drugs have reportedly shown to cause drug-induced interstitial lung disease (DIILD). Early disease detection is important in order to gain the best treatment outcomes. Here, we aimed to develop non-invasive MRI biomarkers, to allow for assessment of disease progression in a chronic model of bleomycin (BL)-induced ILD.

Methods: C57BL/6 mice received i.p. injections of BL (or Saline as control) 2 d/wk, for 4 wks. MRI (RARE and UTE sequences) was performed in wks 3 and 4, as well as 1-2 wks after final dosing. Lung sections from each group were stained with Masson’s-Trichrome followed by modified Ashcroft scoring.

Results: BL-challenged mice showed increased lung/body weight-ratio (p<0.05) while significant signs of low-grade inflammation and fibrosis (p<0.05) were found by histological analysis, indicating lesions emanating from the vascular side. Fibrosis progression was most apparent during the resting period (4+2wk) (p<0.001). These changes were also visualized by MRI (RARE), with increasing lesion size over time (p<0.05). MRI (UTE) analysis also showed increasing airway diameter during disease progression in the BL group.

Conclusion: With non-invasive MRI we could map the lesions and follow the progression of dilated airways over time. This model is clinically relevant and therefore suitable to use for studying DIILD as well as progressive fibrosis.

DIILD mouse model
Article categories

Methotrexate DIILD in rodents

Private
Public

Article Title and Summary

Lorem ipsum dolor sit amet, consectetur adipiscing elit, sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.

Read more

Methotrexate induces DIILD sporadically and at low incidence in rodents, similar to clinical scenario in humans (conference abstract)
 

  • 123

Publications
Take a look

Methotrexate DIILD in rodents

Methotrexate induces DIILD sporadically and at low incidence in rodents, similar to clinical scenario in humans

by N Fransén Pettersson, C Carvalho, I Mahmutovic Persson, H Falk-Håkansson, J Liu, L E Olsson, K Von Wachenfeldt


European Respiratory Journal 2022 60: 2542 (conference abstract). doi: 10.1183/13993003.congress-2022.2542

Abstract

Drug-induced interstitial lung disease (DIILD) is underdiagnosed with increasing incidence. To better diagnose and treat DIILD, translational animal models are warranted. Methotrexate (MTX) is commonly used in patients who present with DIILD. Studies indicate that MTX alone does not induce DIILD, rather underlying genetic/environmental factors or combination therapies seem to have an impact on DIILD development. Here, we attempted to develop a translational chronic model of MTX-induced ILD.

MTX was given to mice or rats, by different routes/concentrations/time points (Table). Histological assessment was the main readout in all studies, with additional MRI and lung function measurements, in selected animal groups. Histology showed limited fibrosis and/or inflammation in MTX treated animals, which developed in some of the animals, across various treatment groups. Appeared lung lesions resided from the vasculature. In cases of detected ILD, also systemic effects in other organs were present. However, there was no correlation between disease severity/dose/frequency of MTX administration.

Conclusion: Neither disease incidence nor severity correlated with MTX concentration or exposure route/time. These observations correspond to clinical observations, where MTX treatment alone does not seem to induce DIILD. Therefore, using MTX only as a DIILD-inducing agent in preclinical research poses great challenges.

Methotrexate DIILD in rodents
Article categories
Subscribe to Publications