Free-breathing dynamic liver MRI using GRASP

Free-breathing dynamic liver MRI using multicoil compressed sensing

Specific Aims

The incidence of hepatocellular carcinoma (HCC) has increased in the United States since 1980, with approximately 20,000 new cases diagnosed every year. This trend is expected to continue over the next decade, which will substantially increase the already enormous annual cost of HCC (over $450 million). Orthotopic liver transplants (OLT) and partial hepatectomy are the only established curative treatments available in patients with localized HCC. Given the invasiveness of the surgery, limited organ availability, and extremely poor prognosis of recurrent HCC, it is important to identify patients who are most likely to have long term curative outcome without recurrence or relapse. Tumor burden, angioinvasion, and tumor grade have been identified as important causes for tumor recurrence after curative therapy, which has been reported to occur in up to 20% of cases after OLT and in up to 50% of cases after hepatic resection. Patients with earlystage HCC (<2 cm) can be treated with potentially curative therapy with best possible long-term survival at a lower cost. Assessment of tumor burden and angioinvasion performed with imaging is the current standard clinical practice guiding treatment. Although studies have shown that contrast-enhanced MRI is superior to CT for detection of HCC, its accuracy is reduced for small tumor nodules (less than 2 cm) and tumor thrombi in small vessels (macroscopic angioinvasion). Furthermore, it is not possible to assess histologic grade reliably on conventional CT or MR imaging. Conventional contrast-enhanced liver MRI is usually performed using a T1- weighted 3D acquisition in a breath-hold, which limits temporal and spatial resolution, and produces suboptimal image quality in many patients with end stage liver disease who have limited breath-holding capabilities. Development of free-breathing contrast-enhanced MRI techniques with high spatial and temporal resolution will increase detection of small HCC (including tumor deposits in small vessels), will improve prediction of histologic grade, and thus improve selection of appropriate individualized therapy.

Compressed sensing has recently emerged as a powerful approach for fast imaging. Compressed sensing, which exploits image compressibility to reduce the quantity of data required to generate faithful images, can be combined with parallel imaging to further increase imaging speed. We have developed Golden-angle RAdial Sparse Parallel (GRASP) MRI, a highly-accelerated dynamic imaging MRI technique, which synergistically combines compressed sensing, parallel imaging and golden-angle radial sampling, a continuous acquisition approach robust to respiratory motion that is well-suited for compressed sensing. GRASP has the advantage of producing high spatial resolution images during free-breathing with access to high temporal resolution information from the same raw data. This method is therefore an ideal candidate for HCC evaluation.It also represents a promising new paradigm for clinical workflow in general, based on continuous comprehensive data acquisition with flexible spatiotemporal resolution tailored retroactively to clinical needs.

Our preliminary results demonstrate that GRASP can generate diagnostic quality multiphase images of the liver, which will improve the detection of small tumors. Furthermore, high temporal resolution dynamic imaging of any particular tumor can be performed with the same raw data without use of additional contrast or acquisition time, which will be useful for characterization of HCC grade. Our overarching goals are (a) to validate a dynamic liver MRI technique with high spatial resolution for detection of small tumors, and with high temporal resolution to accurately predict histologic grade and (b) to evaluate GRASP as a new paradigm for clinical studies using continuous data acquisition and flexible reconstruction.

Specific Aims are as follows:

  1. Evaluate the accuracy of free-breathing GRASP with respect to conventional breath-hold (BH) Cartesian MRI examination for detection of HCC tumor burden and tumor thrombi in vessels, with histopathology as a reference. Hypothesis: Free-breathing GRASP examination will demonstrate superior accuracy for detection of HCC (including small <2 cm tumors), and diagnosis of angioinvasion when compared to BH acquisitions.
  2. Evaluate high temporal resolution GRASP dynamic liver MRI for discrimination of HCC histologic grade. Hypothesis: Poorly differentiated HCC will have a higher arterial perfusion and washout rate compared to well and moderately-differentiated HCC.



Hepatocellular Carcinoma
Hepatocellular carcinoma (HCC) is the sixth most common neoplasm and the third most common cause of death worldwide (1).The incidence of HCC has increased exponentially over the last few decades in the United States with approximately 20,000 new cases diagnosed every year (1-3). As the prevalence of liver disease and cirrhosis is expected to continue to rise over the next decades (4-9), this will result in dramatic increase in incidence of HCC. In addition to the human cost, this will substantially increase the already considerable economic cost of HCC which is over $450 million annually (10).

Five year relative survival rate for localized HCC is 21%, as compared to 6% or lower for HCC with regional or distant spread (from National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) database). Orthotopic liver transplantation (OLT), and surgical resection are considered curative therapeutic options in patients with localized HCC, with the possibility of achieving a high rate of complete response in appropriately selected patients (11-14). Given the invasiveness of the surgery, limited organ availability, and extremely poor prognosis of recurrent HCC, it is important to identify patients who are most likely to have long term curative outcome without recurrence or relapse

Success of surgical resection as curative treatment decreases drastically with increase in tumor number, increase in tumor size or presence of vascular invasion (15-17). Small single tumors are amenable to resection or even ablative therapy, such that asymptomatic patients with early-stage HCC (<2 cm) have the best possible outcome (18) at a lower treatment cost (19).

Limitations of Conventional MRI
burden and macroscopic angioinvasion is performed with imaging (in clinical practice), guiding management. Several studies have consistently shown contrast-enhanced MRI to be superior to other imaging modalities in detection of HCC and tumor burden (20-23). A meta-analysis comparing different modalities demonstrated higher pooled sensitivity of MRI compared to CT and sonography for detection of HCC (81% versus 68% and 60% respectively) (24). However, contrast-enhanced liver MRI is routinely performed during a breath-hold, which limits the achievable spatial resolution and volumetric coverage, and introduces motion-related artifacts in cirrhotic patients who cannot adequately suspend respiration during the acquisition (25). In addition, accurate timing is required for the arterial phase acquisition to optimize contrast between HCC and background liver parenchyma (26). Studies have proposed that mid arterial phase acquisition has higher accuracy for detection of HCC compared to early or late arterial phase acquisition (27). However, on conventional MRI it is not possible to acquire multiple arterial phases with high spatial resolution, or to time the mid arterial phase accurately in patients with cirrhosis who may have volume overload and/or cardiac dysfunction.

These technical limitations result in suboptimal image quality, which reduces sensitivity for detection of small (<2 cm) tumors (28, 29). In a study performed at our institution, when we compared conventional MRI to pathology at explants, lesions ‘missed’ at MRI were all smaller than 2.5 cm in size (30). Furthermore, MRI accuracy is also limited in predicting Edmonson and Steiner histologic grade of HCC (31, 32), although some studies suggest that washout on contrast-enhanced MRI may be associated with high-grade HCC (33, 34).


This project will take advantage of the novel acquisition and reconstruction approaches developed as part of TR&D project #1 to address the urgent clinical need for a free-breathing dynamic liver MRI technique with high spatial and temporal resolution. This will allow us to (a) improve detection of small HCC and tumor thrombi in small peripheral vessels, and (b) assess dynamic enhancement pattern and perfusion of HCC to improve predication of tumor histologic grade


Golden-angle RAdial Sparse Parallel (GRASP) MRI In collaboration with TR&D project #1 co-PI Dr. Otazo and the compressed sensing research group at NYUSOM, we have developed a highly-accelerated free-breathing volumetric dynamic MRI technique named Golden-angle RAdial Sparse Parallel MRI (GRASP) (35), using continuous acquisition of radial MRI data in a golden-angle scheme together with multicoil compressed sensing reconstruction. The reconstruction algorithm uses an extension of k-t SPARSE-SENSE (36), our multicoil compressed sensing method for Cartesian sampling of kspace, to radial sampling. The synergistic combination of compressed sensing, parallel imaging and golden-angle radial sampling is expected to deliver previously impossible combinations of temporal resolution, spatial resolution and volumetric coverage for dynamic MRI studies without the need of breath-holding. Furthermore, GRASP introduces a new paradigm for clinical workflow, involving continuous data acquisition and reconstruction with flexible temporal information. In other words, reconstructions with different temporal information, such as temporal resolution, position, and number of temporal frames, are feasible using the same raw data. As shown in Figures 1 and 2, initial application to liver MRI studies demonstrated diagnostic quality images similar to breath-hold examinations, even in subjects with adequate breath-holding capacity, with substantial expected improvements for subjects incapable of prolonged breath-holds (37). We were also able to generate reconstructions with different temporal resolutions at arbitrary time points using the same raw data sets (35, 38).

Principal Investigator: 
Hersh Chandarana


Latest Updates

07/16/2019 - 15:15
07/10/2019 - 10:06

Philanthropic Support

We gratefully acknowledge generous support for radiology research at NYU Langone Health from:
• The Big George Foundation
• Raymond and Beverly Sackler
• Bernard and Irene Schwartz

Go to top