Evaluation of state-of-the-art segmentation algorithms for left ventricle infarct from late Gadolinium enhancement MR images.

Studies have demonstrated the feasibility of late Gadolinium enhancement (LGE) cardiovascular magnetic resonance (CMR) imaging for guiding the management of patients with sequelae to myocardial infarction, such as ventricular tachycardia and heart failure. Clinical implementation of these developments necessitates a reproducible and reliable segmentation of the infarcted regions. It is challenging to compare new algorithms for infarct segmentation in the left ventricle (LV) with existing algorithms. Benchmarking datasets with evaluation strategies are much needed to facilitate comparison. This manuscript presents a benchmarking evaluation framework for future algorithms that segment infarct from LGE CMR of the LV. The image database consists of 30 LGE CMR images of both humans and pigs that were acquired from two separate imaging centres. A consensus ground truth was obtained for all data using maximum likelihood estimation. Six widely-used fixed-thresholding methods and five recently developed algorithms are tested on the benchmarking framework. Results demonstrate that the algorithms have better overlap with the consensus ground truth than most of the n-SD fixed-thresholding methods, with the exception of the Full-Width-at-Half-Maximum (FWHM) fixed-thresholding method. Some of the pitfalls of fixed thresholding methods are demonstrated in this work. The benchmarking evaluation framework, which is a contribution of this work, can be used to test and benchmark future algorithms that detect and quantify infarct in LGE CMR images of the LV. The datasets, ground truth and evaluation code have been made publicly available through the website: https://www.cardiacatlas.org/web/guest/challenges.

Risk of venous congestion in live donors of extended right liver graft.

AIM: To investigate middle hepatic vein (MHV) management in adult living donor liver transplantation and safer remnant volumes (RV). METHODS: There were 59 grafts with and 12 grafts without MHV (including 4 with MHV-5/8 reconstructions). All donors underwent our five-step protocol evaluation containing a preoperative protocol liver biopsy Congestive vs non-congestive RV, remnant-volume-body-weight ratios (RVBWR) and postoperative outcomes were evaluated in 71 right graft living donors. Dominant vs non-dominant MHV anatomy in total liver volume (d-MHV/TLV vs nd-MHV/TLV) was constellated with large/small congestion volumes (CV-index). Small for size (SFS) and non-SFS remnant considerations were based on standard cut-off- RVBWR and RV/TLV. Non-congestive RVBWR was based on non-congestive RV. RESULTS: MHV and non-MHV remnants showed no significant differences in RV, RV/TLV, RVBWR, total bilirubin, or INR. SFS-remnants with RV/TLV < 30% and non-SFS-remnants with RV/TLV ≥ 30% showed no significant differences either. RV and RVBWR for non-MHV (n = 59) and MHV-containing (n = 12) remnants were 550 ± 95 mL and 0.79 ± 0.1 mL vs 568 ± 97 mL and 0.79 ± 0.13, respectively (P = 0.423 and P = 0.919. Mean left RV/TLV was 35.8% ± 3.9%. Non-MHV (n = 59) and MHV-containing (n = 12) remnants (34.1% ± 3% vs 36% ± 4% respectively, P = 0.148. Eight SFS-remnants with RVBWR < 0.65 had a significantly smaller RV/TLV than 63 non-SFS-remnants with RVBWR ≥ 0.65 [SFS: RV/TLV 32.4% (range: 28%-35.7%) vs non-SFS: RV/TLV 36.2% (range: 26.1%-45.5%), P < 0.009. Six SFS-remnants with RV/TLV < 30% had significantly smaller RVBWR than 65 non-SFS-remnants with RV/TLV ≥ 30% (0.65 (range: 0.6-0.7) vs 0.8 (range: 0.6-1.27), P < 0.01. Two (2.8%) donors developed reversible liver failure. RVBWR and RV/TLV were concordant in 25%-33% of SFS and in 92%-94% of non-SFS remnants. MHV management options including complete MHV vs MHV-4A selective retention were necessary in n = 12 vs n = 2 remnants based on particularly risky congestive and non-congestive volume constellations. CONCLUSION: MHV procurement should consider individual remnant congestive- and non-congestive volume components and anatomy characteristics, RVBWR-RV/TLV constellation enables the identification of marginally small remnants.

Comparison of volumetric and linear serial CT assessments of lung metastases in renal cell carcinoma patients in a clinical phase IIB study.

RATIONALE AND OBJECTIVES: Accuracy of radiologic assessment may have a crucial impact on clinical studies and therapeutic decisions. We compared the variability of a central radiologic assessment (RECIST) and computer-aided volume-based assessment of lung lesions in patients with metastatic renal cell carcinoma (RCC). MATERIALS AND METHODS: The investigation was prospectively planned as a substudy of a clinical randomized phase IIB therapeutic trial in patients with RCC. Starting with the manual study diameter (SDM) of the central readers using RECIST in the clinical study, we performed computer-aided volume measurements. We compared SDM to an automated RECIST diameter (aRDM) and the diameter of a volume-equivalent sphere (effective diameter [EDM]), both for the individual size measurements and for the change rate (CR) between consecutive time points. One hundred thirty diameter pairs of 30 lung lesions from 14 patients were evaluable, forming 55 change pairs over two consecutive time points each. RESULTS: The SDMs of two different readers showed a correlation of 95.6%, whereas the EDMs exhibited an excellent correlation of 99.4%. Evaluation of CRs showed an SDM-CR correlation of 63.9%, which is substantially weaker than the EDM-CR correlation of 87.6%. The variability of SDM-CR is characterized by a median absolute difference of 11.4% points versus the significantly lower 1.8% points EDM-CRs variability (aRDM: 3.2% points). The limits of agreement between readers suggest that an EDM change of 10% or 1 mm can already be significant. CONCLUSIONS: Computer-aided volume-based assessments result in markedly reduced variability of parameters describing size and change, which may offer an advantage of earlier response evaluations and treatment decisions for patients.

Evaluation of current algorithms for segmentation of scar tissue from late gadolinium enhancement cardiovascular magnetic resonance of the left atrium: an open-access grand challenge.

BACKGROUND: Late Gadolinium enhancement (LGE) cardiovascular magnetic resonance (CMR) imaging can be used to visualise regions of fibrosis and scarring in the left atrium (LA) myocardium. This can be important for treatment stratification of patients with atrial fibrillation (AF) and for assessment of treatment after radio frequency catheter ablation (RFCA). In this paper we present a standardised evaluation benchmarking framework for algorithms segmenting fibrosis and scar from LGE CMR images. The algorithms reported are the response to an open challenge that was put to the medical imaging community through an ISBI (IEEE International Symposium on Biomedical Imaging) workshop. METHODS: The image database consisted of 60 multicenter, multivendor LGE CMR image datasets from patients with AF, with 30 images taken before and 30 after RFCA for the treatment of AF. A reference standard for scar and fibrosis was established by merging manual segmentations from three observers. Furthermore, scar was also quantified using 2, 3 and 4 standard deviations (SD) and full-width-at-half-maximum (FWHM) methods. Seven institutions responded to the challenge: Imperial College (IC), Mevis Fraunhofer (MV), Sunnybrook Health Sciences (SY), Harvard/Boston University (HB), Yale School of Medicine (YL), King's College London (KCL) and Utah CARMA (UTA, UTB). There were 8 different algorithms evaluated in this study. RESULTS: Some algorithms were able to perform significantly better than SD and FWHM methods in both pre- and post-ablation imaging. Segmentation in pre-ablation images was challenging and good correlation with the reference standard was found in post-ablation images. Overlap scores (out of 100) with the reference standard were as follows: Pre: IC = 37, MV = 22, SY = 17, YL = 48, KCL = 30, UTA = 42, UTB = 45; Post: IC = 76, MV = 85, SY = 73, HB = 76, YL = 84, KCL = 78, UTA = 78, UTB = 72. CONCLUSIONS: The study concludes that currently no algorithm is deemed clearly better than others. There is scope for further algorithmic developments in LA fibrosis and scar quantification from LGE CMR images. Benchmarking of future scar segmentation algorithms is thus important. The proposed benchmarking framework is made available as open-source and new participants can evaluate their algorithms via a web-based interface.

Determination of lung segments in computed tomography images using the Euclidean distance to the pulmonary artery.

PURPOSE: Computed tomography (CT) imaging is the modality of choice for lung cancer diagnostics. With the increasing number of lung interventions on sublobar level in recent years, determining and visualizing pulmonary segments in CT images and, in oncological cases, reliable segment-related information about the location of tumors has become increasingly desirable. Computer-assisted identification of lung segments in CT images is subject of this work. METHODS: The authors present a new interactive approach for the segmentation of lung segments that uses the Euclidean distance of each point in the lung to the segmental branches of the pulmonary artery. The aim is to analyze the potential of the method. Detailed manual pulmonary artery segmentations are used to achieve the best possible segment approximation results. A detailed description of the method and its evaluation on 11 CT scans from clinical routine are given. RESULTS: An accuracy of 2-3 mm is measured for the segment boundaries computed by the pulmonary artery-based method. On average, maximum deviations of 8 mm are observed. 135 intersegmental pulmonary veins detected in the 11 test CT scans serve as reference data. Furthermore, a comparison of the presented pulmonary artery-based approach to a similar approach that uses the Euclidean distance to the segmental branches of the bronchial tree is presented. It shows a significantly higher accuracy for the pulmonary artery-based approach in lung regions at least 30 mm distal to the lung hilum. CONCLUSIONS: A pulmonary artery-based determination of lung segments in CT images is promising. In the tests, the pulmonary artery-based determination has been shown to be superior to the bronchial tree-based determination. The suitability of the segment approximation method for application in the planning of segment resections in clinical practice has already been verified in experimental cases. However, automation of the method accompanied by an evaluation on a larger number of test cases is required before application in the daily clinical routine.

The "carving" liver partitioning technique for graft hepatectomy in live donor liver transplantation: a single-center experience.

BACKGROUND: In adult live donor liver transplantation, postoperative venous congestion of graft and remnant livers can lead to life-threatening complications. The purpose of this study was to evaluate the safety and benefits of our 3-dimensional, computed tomographic, computer-assisted donor hepatectomy using the "carving" partitioning technique. METHODS: Eighty-three consecutive adult live donor liver transplantations were performed based on data obtained from individualized preoperative 3-dimensional, computed tomographic reconstructions and virtual graft hepatectomies. RESULTS: There were 71 right and 12 left grafts. Small grafts (graft volume body weight ratio, <1.0) were used in 20 cases. We observed no clinically important differences in postoperative function between right and left grafts. Four recipients developed lethal small-for-size syndrome. Reversible small-for-size syndrome was observed in a right graft recipient and in 2 right graft donors. CONCLUSION: Preoperative 3-dimensional, computed tomographic, computer-assisted planning using virtual liver partitioning allowed for: (1) an individualized carving technique based on specific donor anatomic characteristics, (2) donor safety based on individualized patterns of venous outflow, and (3) optimized drainage of the medial area of the graft based on the preferential inclusion of the middle hepatic vein.

Computer-assisted surgical planning in adult-to-adult live donor liver transplantation: how much does it help? A single center experience.

BACKGROUND: Preoperative imaging and donor selection are cardinal components of adult-to-adult live donor liver transplantation (ALDLT). The purpose of this study was to evaluate our three-dimensional (3D) computed tomography image-derived computer-assisted surgical planning (3D CASP) in ALDLT. METHODS: Eighty-three consecutive ALDLTs (71 right and 12 left) were planned with 3D CASP. Graft, remnant, and total liver volume compliance were calculated and compared with actual intraoperative values. Computed risk analysis encompassing territorial liver mapping, functional (safely drained) volumes, and outflow congestion volumes in grafts and remnants allowed for the individualized management of the middle hepatic vein (MHV). RESULTS: Graft volume compliance was 13.5%±4.4%. Three small-for-size (SFS) grafts with lethal SFS syndrome (SFSS) had nonsignificant volume compliance with maximal graft volume-body weight ratios of less than 0.83. Seven SFS grafts with reversible or absent SFSS showed maximal graft volume-body weight ratios of 0.9 to 1.16. Significant differences were identified for (a) virtual graft and remnant congestion volumes of risky versus nonrisky MHV types (49%±6% and 34%±7% vs. 29%±8% and 33%±12%, P<0.001 and P<0.02, respectively) and (b) virtual mean functional versus surgical volumes of grafts (527±119 vs. 963±176 mL, P<0.0001) and remnants (419±182 vs. 640±213 mL, P<0.001). CONCLUSIONS: CASP allowed for (a) prevention of SFSS in extremely small grafts by predicting donor liver plasticity and (b) individualized MHV management for both donors and recipients based on functional graft/remnant volume analysis.

Workflow-centred evaluation of an automatic lesion tracking software for chemotherapy monitoring by CT.

OBJECTIVES: In chemotherapy monitoring, an estimation of the change in tumour size is an important criterion for the assessment of treatment success. This requires a comparison between corresponding lesions in the baseline and follow-up computed tomography (CT) examinations. We evaluate the clinical benefits of an automatic lesion tracking tool that identifies the target lesions in the follow-up CT study and pre-computes the lesion volumes. METHODS: Four radiologists performed volumetric follow-up examinations for 52 patients with and without lesion tracking. In total, 139 lung nodules, liver metastases and lymph nodes were given as target lesions. We measured reading time, inter-reader variability in lesion identification and volume measurements, and the amount of manual adjustments of the segmentation results. RESULTS: With lesion tracking, target lesion assessment time decreased by 38 % or 22 s per lesion. Relative volume difference between readers was reduced from 0.171 to 0.1. Segmentation quality was comparable with and without lesion tracking. CONCLUSIONS: Our automatic lesion tracking tool can make interpretation of follow-up CT examinations quicker and provide results that are less reader-dependent. KEY POINTS: Computed tomography is widely used to follow-up lesions in oncological patients. Novel software automatically identifies and measures target lesions in oncological follow-up examinations. This enables a reduction of target lesion assessment. The automated measurements are less reader-dependent.

Precise local resection for hepatocellular carcinoma based on tumor-surrounding vascular anatomy revealed by 3D analysis.

PURPOSES: Local resection for hepatocellular carcinoma (HCC) has been traditionally performed non-anatomically. The purpose of this study is to evaluate the feasibility of precise local resection of HCC according to the anatomy of tumor-surrounding vessels revealed by three-dimensional (3D) analysis technique. METHODS: The CT datasets of the livers of the patients with HCC were analyzed three-dimensionally. The tumor-bearing vessels were identified and virtually resected, and the depending parenchymal volume was calculated for definition of an optimal liver division plane. The actual local resections were then carried out according to the simulations. RESULTS: Precise local resection based on tumor-surrounding vascular anatomy was performed in 13 HCC patients. Both resection margin and volume were significantly correlated with those predicted by preoperative simulations. After precise local resection, neither ischemia nor congestion was observed in the remnant livers. All patients obtained adequate resection margins, without recurrences in the resection sites after a median follow-up time of 18 months. CONCLUSIONS: Local resection for HCC can be carried out precisely according to the anatomy of tumor-surrounding vessels when guided by a 3D analysis. This precise procedure will enhance both the accuracy and safety of traditional local resection.

Visualization support for the planning of hepatic needle placement.

PURPOSE: Percutaneous image-guided interventions, such as radiofrequency ablation or biopsy, are using needle-shaped instruments which have to be inserted into a target area without penetrating any vital structure. The established planning workflow is based on viewing 2D slices of a pre-interventional CT or MR scan. However, access paths not parallel to the axial plane are often necessary. For such complicated cases, the planning process is challenging and time consuming if solely based on 2D slices. To overcome these limitations while keeping the well-established workflow, we propose a visualization method that highlights less suited paths directly in the 2D visualizations with which the radiologist is familiar. METHODS: Based on a user defined target point and segmentation masks of relevant risk structures, a risk structure map is computed using GPU accelerated volume rendering and projected onto the 2D slices. This visualization supports the user in defining safe linear access paths by selecting a second point directly in the 2D image slices. RESULTS: In an evaluation for 20 liver radiofrequency ablation cases, 3 experienced radiologists stated for 55% of the cases that the visualization supported the access path choice. The visualization support was rated with an average mark of 2.2. For 2 of the 3 radiologists, a significant reduction of the planning duration by 54 and 50% was observed. CONCLUSIONS: The proposed visualization approach can both accelerate the access path planning for radiofrequency ablation in the liver and facilitate the differentiation between safer and less safe paths.

Fusion of MR coronary angiography and viability imaging: feasibility and clinical value for the assignment of myocardial infarctions.

PURPOSE: To investigate the feasibility of image fusion of MR-coronary angiography (MRCA) and delayed gadolinium enhancement imaging (LGE) and to assign areas of myocardial infarction to the corresponding supplying coronary arteries. MATERIALS AND METHODS: An interactive segmentation of the coronary arteries was performed in MRCA data sets (n=25). The LGE slices were matched onto the vessel segmentation to perform a fused analysis of coronary artery anatomy and LGE. The results were compared to the segmental model recommended by the American Heart Association (AHA). Standard of reference was the identification of the culprit lesion in the invasive coronary angiography (CA) (n=20). RESULTS: The fused analysis allowed the assignment of MI to the supplying coronary artery in 13/20 patients. The sensitivities/specificities for the assignment of MI to the three main vessels were: LAD 63%/100%, LCX 75%/100%, and RCA 56%/100%, respectively. Using the AHA segmental model the sensitivities/specificities for the correct assignment of MI to the three main vessels were: LAD 88%/58%, LCX 94%/75%, and RCA 77%/73%, respectively. CONCLUSION: Fusion images of MRCA and LGE provides added diagnostic information in the effort to determine the epicardial vessels responsible for the postischemic myocardial injury and therefore might be helpful to establish appropriate future therapeutic steps.

Probabilistic 4D blood flow tracking and uncertainty estimation.

Phase-Contrast (PC) MRI utilizes signal phase shifts resulting from moving spins to measure tissue motion and blood flow. Time-resolved 4D vector fields representing the motion or flow can be derived from the acquired PC MRI images. In cardiovascular PC MRI applications, visualization techniques such as vector glyphs, streamlines, and particle traces are commonly employed for depicting the blood flow. Whereas these techniques indeed provide useful diagnostic information, uncertainty due to noise in the PC-MRI measurements is ignored, which may lend the results a false sense of precision. In this work, the statistical properties of PC MRI flow measurements are investigated and a probabilistic flow tracking method based on sequential Monte Carlo sampling is devised to calculate flow uncertainty maps. The theoretical derivations are validated using simulated data and a number of real PC MRI data sets of the aorta and carotid arteries are used to demonstrate the flow uncertainty mapping technique.

Object-based analysis of CT images for automatic detection and segmentation of hypodense liver lesions.

PURPOSE: Hypodense liver lesions are commonly detected in CT, so their segmentation and characterization are essential for diagnosis and treatment. Methods for automatic detection and segmentation of liver lesions were developed to support this task. METHODS: The detection algorithm uses an object-based image analysis approach, allowing for effectively integrating domain knowledge and reasoning processes into the detection logic. The method is intended to succeed in cases typically difficult for computer-aided detection systems, especially low contrast of hypodense lesions relative to healthy tissue. The detection stage is followed by a dedicated segmentation algorithm needed to synthesize 3D segmentations for all true-positive findings. RESULTS: The automated method provides an overall detection rate of 77.8% with a precision of 0.53 and performs better than other related methods. The final lesion segmentation delivers appropriate quality in 89% of the detected cases, as evaluated by two radiologists. CONCLUSIONS: A new automated liver lesion detection algorithm employs the strengths of an object-based image analysis approach. The combination of automated detection and segmentation provides promising results with potential to improve diagnostic liver lesion evaluation.

A robust and extendible framework for medical image registration focused on rapid clinical application deployment.

Development and integration of image registration methods become increasingly important for clinical workstations. Due to the complexity of such methods, prototyping, evaluation and workflow integration require in-depth knowledge foremostly available to registration developers. Rapid development and deployment is therefore often difficult, particularly for comprehensive software frameworks. In this article, we introduce a novel rapid prototyping framework for voxel-based registration. It is specifically designed to allow evaluation and adaption by developers with less knowledge on registration internals. Based on a unique "one-iteration" paradigm, it enables accelerated algorithm development. Furthermore, methods are interchangeable at runtime using an intuitive graphical plugin interface, allowing a genuine comparison of methods. We discuss the concepts of this framework, compare it to other software and demonstrate its effectiveness for several demanding registration applications, highlighting its versatility and reliability.

In vivo validation of a therapy planning system for laser-induced thermotherapy (LITT) of liver malignancies.

PURPOSE: In situ ablation is increasingly being used for the treatment of liver malignancies. The application of these techniques is limited by the lack of a precise prediction of the destruction volume. This holds especially true in anatomically difficult situations, such as metastases in the vicinity of larger liver vessels. We developed a three-dimensional (3D) planning system for laser-induced thermotherapy (LITT) of liver tumors. The aim of the study was to validate the system for calculation of the destruction volume. METHODS: LITT (28 W, 20 min) was performed in close contact to major hepatic vessels in six pigs. After explantation of the liver, the coagulation area was documented. The liver and its vascular structures were segmented from a pre-interventional CT scan. Therapy planning was carried out including the cooling effect of adjacent liver vessels. The lesions in vivo and the simulated lesions were compared with a morphometric analysis. RESULTS: The volume of lesions in vivo was 6,568.3 ± 3,245.9 mm(3), which was not different to the simulation result of 6,935.2 ± 2,538.5 mm(3) (P = 0.937). The morphometric analysis showed a sensitivity of the system of 0.896 ± 0.093 (correct prediction of destructed tissue). The specificity was 0.858 ± 0.090 (correct prediction of vital tissue). CONCLUSIONS: A 3D computer planning system for the prediction of thermal lesions in LITT was developed. The calculation of the directional cooling effect of intrahepatic vessels is possible for the first time. The morphometric analysis showed a good correlation under clinical conditions. The pre-therapeutic calculation of the ablation zone might be a valuable tool for procedure planning.

Image-guided therapies.

This paper presents image-guided therapies development and advantages using real time imaging modalities such as computed tomography, ultrasound imaging, and magnetic resonance imaging. The following are also highlighted: image-guided cancer interventions and image-guided cardiovascular interventions.

Probabilistic 4D blood flow mapping.

Blood flow and tissue velocity can be measured using phase-contrast MRI. In this work, the statistical properties of 4D phase-contrast images are derived, and a novel probabilistic blood flow mapping method based on sequential Monte Carlo sampling is presented. The resulting flow maps visualize and quantify the uncertainty in conventional flow visualization techniques such as streamlines and particle traces.