Deep learning model for automatic segmentation of lungs and pulmonary metastasis in small animal MR images

Lungs are the most frequent site of metastases growth. The amount and size of pulmonary metastases acquired from MRI imaging data are the important criteria to assess the efficacy of new drugs in preclinical models. While efficient solutions both for MR imaging and the downstream automatic segmentation have been proposed for human patients, both MRI lung imaging and segmentation in preclinical animal models remains challenging due to the physiological motion (respiratory and cardiac movements), to the low amount of protons in this organ and to the particular challenge of precise segmentation of metastases. As a consequence post-mortem analysis is currently required to obtain information on metastatic volume. In this work, we have developed a complete methodological pipeline for automated analysis of lungs and metastases in mice, consisting of an MR sequence for image acquisition and a deep learning method for automatic segmentation of both lungs and metastases. On one hand, we optimized an MR sequence for mouse lung imaging with high contrast for high detection sensitivity. On the other hand we developed DeepMeta, a multiclass U-Net 3+ deep learning model to automatically segment the images. To assess if the proposed deep learning pipeline is able to provide an accurate segmentation of both lungs and pulmonary metastases, we have longitudinally imaged mice with fast- and slow-growing metastasis. Fifty-five balb/c mice were injected with two different derivatives of renal carcinoma cells. Mice were imaged with a SG-bSSFP (self-gated balanced steady state free precession) sequence at different time points after the injection of cancer cells. Both lung and metastases segmentations were manually performed by experts. DeepMeta was trained to perform lung and metastases segmentation based on the resulting ground truth annotations. Volumes of lungs and of pulmonary metastases as well as the number of metastases per mouse were measured on a separate test dataset of MR images. Thanks to the SG method, the 3D bSSFP images of lungs were artifact-free, enabling the downstream detection and serial follow-up of metastases. Moreover, both lungs and metastases segmentation was accurately performed by DeepMeta as soon as they reached the volume of ∼ 0.02 m m 3 . Thus we were able to distinguish two groups of mice in terms of number and volume of pulmonary metastases as well as in terms of the slow versus fast patterns of growth of metastases. We have shown that our methodology combining SG-bSSFP with deep learning, enables processing of the whole animal lungs and is thus a viable alternative to histology alone.

Micron-sized iron oxide particles for both MRI cell tracking and magnetic fluid hyperthermia treatment

Iron oxide particles (IOP) are commonly used for Cellular Magnetic Resonance Imaging (MRI) and in combination with several treatments, like Magnetic Fluid Hyperthermia (MFH), due to the rise in temperature they provoke under an Alternating Magnetic Field (AMF). Micrometric IOP have a high sensitivity of detection. Nevertheless, little is known about their internalization processes or their potential heat power. Two micrometric commercial IOP (from Bangs Laboratories and Chemicell) were characterized by Transmission Electron Microscopy (TEM) and their endocytic pathways into glioma cells were analyzed. Their Specific Absorption Rate (SAR) and cytotoxicity were evaluated using a commercial AMF inductor. T2-weighted imaging was used to monitor tumor growth in vivo after MFH treatment in mice. The two micron-sized IOP had similar structures and r2 relaxivities (100 mM-1 s-1) but involved different endocytic pathways. Only ScreenMAG particles generated a significant rise in temperature following AMF (SAR = 113 W g-1 Fe). After 1 h of AMF exposure, 60% of ScreenMAG-labeled cells died. Translated to a glioma model, 89% of mice responded to the treatment with smaller tumor volume 42 days post-implantation. Micrometric particles were investigated from their characterization to their intracellular internalization pathways and applied in one in vivo cancer treatment, i.e. MFH.

Cell density quantification with TurboSPI: R2* mapping with compensation for off-resonance fat modulation

OBJECTIVE

Tracking the migration of superparamagnetic iron oxide (SPIO)-labeled immune cells in vivo is valuable for understanding the immunogenic response to cancer and therapies. Quantitative cell tracking using TurboSPI-based R₂* mapping is a promising development to improve accuracy in longitudinal studies on immune recruitment. However, off-resonance fat signal isochromats lead to modulations in the signal time-course that can be erroneously fit as R₂* signal decay, overestimating the density of labeled cells, while excluding voxels with fat-typical modulations results in underestimation of cell density in voxels with mixed content. Approaches capable of accurate R₂* estimation in the presence of fat are needed.

METHODS

We propose a dual-decay (separate R_2f* and R_2w* for fat and water) Dixon-based signal model that accounts for the presence of fat in a voxel to provide better estimates of SPIO-induced dephasing. This model was tested in silico, in phantoms with varying quantities of fat and SPIO-labeled cells, and in 5 mice injected with SPIO-labeled CD8+ T cells.

RESULTS

In silico single voxel simulations illustrate how the proposed dual-decay model provides stable R_2w* estimates that are invariant to fat content. The proposed model outperforms previous methods when applied to in vitro samples of SPIO-labeled cells and oil prepared with oil content ≥ 15%. Preliminary in vivo results show that, compared to previous methods, the dual-decay model improves the balance of R₂* mapping in fat-dense areas, which will yield more reliable analysis in future cell tracking studies.

DISCUSSION

The proposed model is a promising tool for quantitative TurboSPI R₂* cell tracking, with further refinements offering the possibility of better specificity and sensitivity.

Radial MP2RAGE sequence for rapid 3D T1 mapping of mouse abdomen: application to hepatic metastases

OBJECTIVES

The T₁ longitudinal recovery time is regarded as a biomarker of cancer treatment efficiency. In this scope, the Magnetization Prepared 2 RApid Gradient Echo (MP2RAGE) sequence relevantly complies with fast 3D T₁ mapping. Nevertheless, with its Cartesian encoding scheme, it is very sensitive to respiratory motion. Consequently, a radial encoding scheme was implemented for the detection and T₁ measurement of hepatic metastases in mice at 7T.

METHODS

A 3D radial encoding scheme was developed using a golden angle distribution for the k-space trajectories. As in that case, each projection contributes to the image contrast, the signal equations had to be modified. Phantoms containing increasing gadoteridol concentrations were used to determine the accuracy of the sequence in vitro. Healthy mice were repetitively scanned to assess the reproducibility of the T₁ values. The growth of hepatic metastases was monitored. Undersampling robustness was also evaluated.

RESULTS

The accuracy of the T₁ values obtained with the radial MP2RAGE sequence was > 90% compared to the Inversion-Recovery sequence. The motion robustness of this new sequence also enabled repeatable T₁ measurements on abdominal organs. Hepatic metastases of less than 1-mm diameter were easily detected and T₁ heterogeneities within the metastasis and between the metastases within the same animal were measured. With a twofold acceleration factor using undersampling, high-quality 3D T₁ abdominal maps were achieved in 9 min.

CONCLUSIONS

The radial MP2RAGE sequence could be used for fast 3D T₁ mapping, to detect and characterize metastases in regions subjected to respiratory motion.

KEY POINTS

• The Cartesian encoding of the MP2RAGE sequence was modified to a radial encoding. The modified sequence enabled accurate T ₁ measurements on phantoms and on abdominal organs of mice. • Hepatic metastases were easily detected due to high contrast. Heterogeneity in T ₁ was measured within the metastases and between each metastasis within the same animal. • As implementation of this sequence does not require specific hardware, we expect that it could be readily available for clinical practice in humans.

Regularly incremented phase encoding - MR fingerprinting (RIPE-MRF) for enhanced motion artifact suppression in preclinical cartesian MR fingerprinting

PURPOSE

The regularly incremented phase encoding-magnetic resonance fingerprinting (RIPE-MRF) method is introduced to limit the sensitivity of preclinical MRF assessments to pulsatile and respiratory motion artifacts.

METHODS

As compared to previously reported standard Cartesian-MRF methods (SC-MRF), the proposed RIPE-MRF method uses a modified Cartesian trajectory that varies the acquired phase-encoding line within each dynamic MRF dataset. Phantoms and mice were scanned without gating or triggering on a 7T preclinical MRI scanner using the RIPE-MRF and SC-MRF methods. In vitro phantom longitudinal relaxation time (T₁ ) and transverse relaxation time (T₂ ) measurements, as well as in vivo liver assessments of artifact-to-noise ratio (ANR) and MRF-based T₁ and T₂ mean and standard deviation, were compared between the two methods (n = 5).

RESULTS

RIPE-MRF showed significant ANR reductions in regions of pulsatility (P < 0.005) and respiratory motion (P < 0.0005). RIPE-MRF also exhibited improved precision in T₁ and T₂ measurements in comparison to the SC-MRF method (P <  0.05). The RIPE-MRF and SC-MRF methods displayed similar mean T₁ and T₂ estimates (difference in mean values < 10%).

CONCLUSION

These results show that the RIPE-MRF method can provide effective motion artifact suppression with minimal impact on T₁ and T₂ accuracy for in vivo small animal MRI studies. Magn Reson Med 79:2176-2182, 2018. © 2017 International Society for Magnetic Resonance in Medicine.

Fast 3D ultrashort echo-time spiral projection imaging using golden-angle: A flexible protocol for in vivo mouse imaging at high magnetic field

PURPOSE

To develop a fast three-dimensional (3D) k-space encoding method based on spiral projection imaging (SPI) with an interleaved golden-angle approach and to validate this novel sequence on small animal models.

METHODS

A disk-like trajectory, in which each disk contained spirals, was developed. The 3D encoding was performed by tilting the disks with a golden angle. The sharpness was first calculated at different T2* values. Then, the sharpness was measured on phantom using variable undersampling ratios. Finally, the sampling method was validated by whole brain time-of-flight angiography and ultrasmall superparamagnetic iron oxide (USPIO) enhanced free-breathing liver angiography on mouse.

RESULTS

The in vitro results demonstrated the robustness of the method for short T2* and high undersampling ratios. In vivo experiments showed the ability to properly detect small vessels in the brain with an acquisition time shorter than 1 min. Free-breathing mice liver angiography showed the insensitivity of this protocol toward motions and flow artifacts, and enabled the visualization of liver motion during breathing.

CONCLUSIONS

The method implemented here allowed fast 3D k-space sampling with a high undersampling ratio. Combining the advantages of center-out spirals with the flexibility of the golden angle approach could have major implications for real-time imaging. Magn Reson Med 77:1831-1840, 2017. © 2016 International Society for Magnetic Resonance in Medicine.