Comparison of human and automatic segmentations of kidneys from CT images

Rao, M. S.; Stough, Joshua V.; Yun, Yueh; Muller, Keith E.; Tracton, Gregg; Pizer, Stephen M.; Chaney, Edward L.

doi:10.1016/j.ijrobp.2004.11.014

Cited by 70 publications

(59 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…3 shows a representative example. These results are comparable to previously reported methods [11,12,13,14] without their underlying assumptions and with a significant improvement in the running time.…”

Section: Resultssupporting

confidence: 89%

Anatomical Structures Segmentation by Spherical 3D Ray Casting and Gradient Domain Editing

Kronman

Joskowicz

2012

Medical Image Computing and Computer-Assisted Intervention – MICCAI 2012

View full text Add to dashboard Cite

Abstract. Fuzzy boundaries of anatomical structures in medical images make segmentation a challenging task. We present a new segmentation method that addresses the fuzzy boundaries problem. Our method maps the lengths of 3D rays cast from a seed point to the unit sphere, estimates the fuzzy boundaries location by thresholding the gradient magnitude of the rays lengths, and derives the true boundaries by Laplacian interpolation on the sphere. Its advantages are that it does not require a global shape prior or curvature based constraints, that it has an automatic stopping criteria, and that it is robust to anatomical variability, noise, and parameters values settings. Our experimental evaluation on 23 segmentations of kidneys and on 16 segmentations of abdominal aortic aneurysms (AAA) from CT scans yielded an average volume overlap error of 12.6% with respect to the ground-truth. These results are comparable to those of other segmentation methods without their underlying assumptions.

show abstract

Section: Resultssupporting

confidence: 89%

Anatomical Structures Segmentation by Spherical 3D Ray Casting and Gradient Domain Editing

Kronman

Joskowicz

2012

Medical Image Computing and Computer-Assisted Intervention – MICCAI 2012

View full text Add to dashboard Cite

show abstract

“…Most previous studies reported on the accuracy of measuring the volume of the entire kidney by either ultrasound or computed tomography (CT) (21)(22)(23)(24)(25).…”

Section: Discussionmentioning

confidence: 99%

Performance of an automated segmentation algorithm for 3D MR renography

Rusinek

Boykov

Kaur

et al. 2007

Magnetic Resonance in Med

View full text Add to dashboard Cite

The accuracy and precision of an automated graph-cuts (GC) segmentation technique for dynamic contrast-enhanced (DCE) 3D MR renography (MRR) was analyzed using 18 simulated and 22 clinical datasets. For clinical data, the error was 7.2 ؎ 6.1 cm 3 for the cortex and 6.5 ؎ 4.6 cm 3 for the medulla. The precision of segmentation was 7.1 ؎ 4.2 cm 3 for the cortex and 7.2 ؎ 2.4 cm 3 for the medulla. Compartmental modeling of kidney function in 22 kidneys yielded a renal plasma flow (RPF) error of 7.5% ؎ 4.5% and single-kidney GFR error of 13.5% ؎ 8.8%. The precision was 9.7% ؎ 6.4% for RPF and 14.8% ؎ 11.9% for GFR. It took 21 min to segment one kidney using GC, compared to 2.5 hr for manual segmentation. One technique to determine renal function consists of the intravenous injection of radioactive tracer followed by assessment of its plasma clearance 2-4 hr later. This technique is time-consuming, requires multiple blood samples, and measures only the global glomerular filtration rate (GFR)-a disadvantage when asymmetric or unilateral renal disease is present. The gold standard technique for assessing single-kidney GFR is inulin clearance, but this method is too invasive and complex for routine clinical application. As an alternative, dynamic gamma camera imaging with 99m Tc-DTPA has been shown to provide single-kidney GFR by analysis of the renal radioactivity. By combining measures of renal physiology with depiction of anatomical detail, dynamic contrast-enhanced (DCE) 3D MR renography (MRR) has the potential to improve upon nuclear medicine techniques and also provide useful functional information to supplement anatomic renal MRI examinations (1). Good spatial, temporal, and contrast resolution is achievable with current contrast-enhanced dynamic protocols, whereby serial 3D MR images of the kidneys are generated following an injection of contrast material. Gadolinium (Gd) chelates, such as gadopentetate dimeglumine (Gd-DTPA), are suitable MR contrast agents because they are freely filtered at the glomerulus without tubular secretion or resorption.Several approaches have been proposed to analyze renography data, including the upslope method (2), deconvolution (3,4), the Rutland-Patlak method (5,6), and renal kinetic modeling (7-10). The key prerequisite is the ability to segment dynamic MR images into functional regions (i.e., the cortical and medullary compartments). Fitting concentration-time activity (CTA) curves to kinetic models yields perfusion and filtration rates per unit volume of tissue. These kinetic rates multiplied by the cortical and medullary volumes (determined from segmented images) give the renal plasma flow (RPF) and GFR for the entire kidney.Accurate segmentation of contrast-enhanced MRR data remains a difficult task. Dynamic 3D MR images of the abdomen suffer from partial-volume and respiratory-motion artifacts and have a relatively low signal-to-noise ratio (SNR). Other sources of error include signal nonuniformity and wraparound artifacts. The presence of cysts and renal masses, and reduced...

show abstract

“…The discrepancies between the reference manual contours and automatic generated contours included registration errors, as well as human variations. This quantitative validation method is only valid when the ROI can be accurately defined by a human, such as has been done for the kidney (28). We have performed repeat manual delineation by two physicians on a few data sets.…”

Section: Discussionmentioning

confidence: 99%

On-line Adaptive Radiation Treatment of Prostate Cancer

Zhang¹

2008

View full text Add to dashboard Cite

Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources, gathering and maintaining the data needed, and completing and reviewing this collection of information. Send comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden to Department of Defense, Washington Headquarters Services, Directorate for Information Operations and Reports (0704-0188), 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA 22202-4302. Respondents should be aware that notwithstanding any other provision of law, no person shall be subject to any penalty for failing to comply with a collection of information if it does not display a currently valid OMB control number. PLEASE DO NOT RETURN YOUR FORM TO THE ABOVE ADDRESS. B. Developing of parallel deformable image registration algorithm. REPORT DATEWe have implemented an intensity-based deformable image registration algorithm on a 16-PE (processing element) Beowulf cluster. This algorithm has been used for deformable registration of various sites including the prostate. We have set up a deformable registration cluster in our clinic so that the research outcomes are utilized in the clinical treatment planning.The robustness of deformable image registration for head and neck cases was quantitatively evaluated using multiple CT images for 10 patients with each patient having 9 daily CT images. The automatically delineated region of interests (ROIs) were compared with manual delineation which was considered as golden standards. Results showed excellent agreement between automatically and manually delineated ROIs 4Evaluation of prostate registration is still in progress. Figure 1 shows the results of automatic delineation of the prostate. Generally the manual contours agree with automatic contours. However, bowel gas and rectum filling frequently cause problems in deformable image registration. Currently we are working in two directions to solve this problem: (1) control patient's diet and give instruction to the patient to empty bladder and rectum before treatment; (2) create organ model and perform statistical training. Rectum balloon may be another option.To accelerate registration speed, we have implemented parallel deformable image registration using Message Passing Interface (MPI). The speed up factor and efficiency of the parallel deformable image registration algorithms are show in Figure 2. Using a 16-PE cluster, we have achieved a speedup factor of 15. Registration of 3D CT images takes only about 2-3 mins. We expect the time can be reduced to within 30 sec with a 64-PE computer cluster. The results have been reported in an international conference C. Developing of parallel IMRT re-optimization algorithm.Online re-planning requires online dose calculation and online plan re-optimization. All ROIs in prostate treatment are deep inside patient's body. Dose calculation on daily ...

show abstract

Comparison of human and automatic segmentations of kidneys from CT images

Cited by 70 publications

References 26 publications

Anatomical Structures Segmentation by Spherical 3D Ray Casting and Gradient Domain Editing

Anatomical Structures Segmentation by Spherical 3D Ray Casting and Gradient Domain Editing

Performance of an automated segmentation algorithm for 3D MR renography

On-line Adaptive Radiation Treatment of Prostate Cancer

Contact Info

Product

Resources

About