Tumor segmentation from magnetic resonance MR images may aid in tumor treatment b y tracking the progress of tumor growth and or shrinkage. In this paper we present the rst automatic segmentation method which separates non-enhancing brain tumors from healthy tissues in MR images to aid in the task of tracking tumor size over time. The MR feature images used for the segmentation consist of three weighted images T1, T2 and proton density for each axial slice through the head. An initial segmentation is computed using an unsupervised fuzzy clustering algorithm. Then, integrated domain knowledge and image processing techniques contribute to the nal tumor segmentation. They are applied under the control of a knowledge-based system. The system knowledge was acquired by training on two patient v olumes 14 images. Testing has shown successful tumor segmentations on four patient volumes 31 images. Our results show that we detected all six non-enhancing brain tumors, located tumor tissue in 35 of the 36 ground truth radiologist labeled slices containing tumor and successfully separated tumor regions from physically connected CSF regions in nine of nine slices. Quantitative measurements are promising as correspondence ratios between ground truth and segmented tumor regions ranged between 0.368 and 0.871 per volume, with percent match ranging between 0.530 and 0.909 per volume.

The growing importance of the "European Protocol for the Quality Control of the Physical and Technical Aspects of Mammography Screening, Part B: Digital Mammography" dictates the need to understand the prescribed threshold contrast sensitivity test. Observers following a 4-AFC paradigm, report the location of disks varying in contrast and diameter on multiple images of a CDMAM or similar phantom. Analysis provides a contrast threshold for each disk diameter. The goals of this study were to quantify the performance of new observers, compare it to published results, compare visual scoring with software scoring of the same images, and to quantify the major sources of variability.Digital phantom images, visual scorings by four expert readers, and CDCOM software were downloaded from the EUREF website. These images were read on a 3M Barco flat-panel monitor by 13 observers and scored by CDCOM. Scores were analyzed using the published method from the CDMAM-phantom 3.4 manual and a signal detection theory-based method.The average contrast sensitivities of the 13 study observers generally exceeded the published values by ~10%. The 95% confidence limits for the mean of 6 images from the published data vary from ±20.2% to ±41.8% of their respective means, the average being 31.2%. The average confidence limit for selected study observers is ±36%.Comparisons between software and human observer results using the prescribed method of analysis-revealed marked differences, particularly for small diameter targets. These differences are mitigated by signal-detection-theory analysis of both datasets. The large inter-observer variability and the substantial time required for human scoring supports the need to qualify a readily available software solution.

No abstract

No abstract

No abstract