Split-and-merge segmentation of magnetic resonance medical images: performance evaluation and extension to three dimensions.

Intensity- or edge-based methods of segmentation are often insufficiently robust to be applied to images containing complex anatomical objects, such as those seen in high-resolution magnetic resonance imaging systems. Split-and-merge techniques attempt to overcome these difficulties by using homogeneity measures. Simple modifications to the basic 2D split-and-merge method, based on the principles of simulated annealing and controlled boundary elimination, are developed and discussed. Simulated annealing reduced the number of regions by 22% with a further reduction of 21% achieved through boundary elimination. Smoother regional boundaries are also produced. These methods are extended to true 3D and quantitatively compared with their 2D counterparts. The main advantage of 3D methods is that they produce segmented volumes by directly preserving the connectivity between slices, whereas in 2D, segments have to be grouped together in a post-split-and-merge process. Finally, the properties of the 3D approach are demonstrated by the automatic quantitation of brain ventricle volume, producing estimates to within 7% of validated manual methods.

Validation of the in vivo measurement of adipose tissue by magnetic resonance imaging of lean and obese pigs.

In vivo quantification of adipose tissue with magnetic resonance imaging (MRI) was validated with pigs. Thirteen transaxial MRI sections were collected, at intervals proportional to body length, from each pig, which was then killed, frozen, and sliced at the locations of the MRI sections. Adipose-tissue quantities were determined by dissecting each slice, and lipid contents of the dissected slices and of the tissue segments between slices were measured. Compared with dissection, MRI underestimated abdominal percent adipose tissue and overestimated cervical percent adipose tissue by less than 6%. When all 13 sections were used, MRI closely predicted percent lipid and dissected percent adipose tissue with small residual SDs (RSD = 1.9 and 2.1, respectively), which increased only slightly if two sections (4, upper thorax and 8, upper abdomen) were used (RSD = 2.3 and 2.6, respectively). In conclusion MRI accurately quantifies adipose tissue in vivo, matching values produced by dissection and chemical analysis.

Total and subcutaneous adipose tissue in women: the measurement of distribution and accurate prediction of quantity by using magnetic resonance imaging.

Total and subcutaneous adipose tissue in seven lean and seven obese women were quantified using magnetic resonance imaging (MRI). The distributions of adipose tissues along the body were closely correlated: subcutaneous with total, both within and between lean and obese groups. Lean women had proportionally less adipose tissue in the lower thorax and upper abdomen than did obese women. Reducing the number of MRI scans from 17 to 4 did not increase the residual SD of predicted body adipose tissue (2.9 percent) when body density was used as the reference measure. MRI gave an estimate of total-body adipose tissue significantly closer to the value for fat percent produced when the results from five other techniques (skinfold thickness, underwater weighing, 40K whole-body counting, isotopic water dilution, and tetrapolar bioelectrical impedance) were averaged than when any other technique was used alone. MRI-determined percent body adipose tissue in women is close to, and proportional to, estimates derived by underwater weighing.

Cyclic changes in composition and volume of the breast during the menstrual cycle, measured by magnetic resonance imaging.

The volumes and spin-lattice (T1) relaxation times of breast tissues and parenchymal water content were measured non-invasively by magnetic resonance imaging (MRI) in eight healthy women during four to eight consecutive menstrual cycles. Total breast volume, and parenchymal volume, T1 relaxation time and water content were lowest between days 6 and 15. Between days 16 and 28, parenchymal volume, T1 relaxation time and water content rose sharply by 38.9%, 15.1% and 24.5%, respectively, and peaked after day 25. Within 5 days of the onset of menses, parenchymal volume fell sharply by 30.3%, while water content declined by 17.5%. Rising parenchymal volume in the second half of the menstrual cycle is not solely due to increased tissue water content and provides in vivo evidence for both growth and increased tissue fluid at this time.

Use of magnetic resonance imaging in the study of goat mammary glands in vivo.

A technique is described for the in-vivo determination of mammary gland size and gross composition in goats by using nuclear magnetic resonance imaging (MRI). The volume of test objects determined with MRI had an error of +0.4 +/- 1.6% of the actual volume. In lactating goats the in-vivo MRI estimate of mammary parenchymal volume was significantly greater than, but highly significantly correlated with, the weight of parenchyma determined post mortem (for the whole udder, r = 0.88, P less than 0.001; for individual glands, r = 0.85, P less than 0.001). MRI-determined estimates of the volume of fluid within the mammary gland were within 1.2% of the volume of milk removed from the udders after imaging. The spin-lattice (T1) relaxation time of the whole udder correlated closely with the volume of fluid within the udder. The T1 relaxation time of parenchymal tissue measured in vivo did not differ significantly from that determined immediately after post-mortem excision.

In-vivo studies of mammary development in the goat using magnetic resonance imaging (MRI).

Mammary development and regression were measured in goats in vivo using magnetic resonance imaging (MRI). Measurements were made during the first and second cycles of pregnancy, lactation and involution. In primiparous goats, and exponential pattern of growth was evident during gestation and for the first 2 weeks of lactation. Parenchyma volume correlated significantly with milk yield across goats during early lactation, and across stage of lactation within goats. Milking was discontinued in Week 26 of the first lactation. Involution was characterized by an initial accumulation of fluid (over 2 days) followed by reabsorption; parenchyma volume did not decrease significantly until the 3rd week of involution, which was also the time at which these goats were mated to start their second gestation. Their udders still contained significant quantities of fluid (40-60% of the gross volume), but parenchyma volume was also greater (by 4.7-fold) than in goats beginning their first gestation. By Week 15 of gestation there was no longer a parity difference in parenchyma; the udders of first-gestation goats had grown significantly, but those of second-gestation goats had not. Conversely, between gestation Week 15 and lactation Week 2 mammary growth was significantly more rapid in the second cycle, such that the udder was larger at the start of the second lactation.

The case of the missing vault: a "cold" bone lesion following electrical burn.

Osteonecrosis of the skull produced by an electrical burn presented as a dramatic "cold" bone lesion on an MDP bone scan, despite normal skull radiographs. Four months later the skull radiograph showed marked bone resorption, and the three-phase bone scan confirmed healing and new osteoblastic activity. The pathophysiology of high-voltage injuries is outlined.

Comparative efficacy of ampicillin and trimethoprim-sulfamethoxazole in otitis media.

Seventy-nine children with acute otitis media were the subjects in a study designed to compare the effectiveness of ampicillin and trimethoprim-sulfamethoxazole in this infection. They received either of these two agents according to a double-blind randomized procedure that also took the child's weight into account. No significant difference was found in the clinical outcome between the two treatment regimens. Undesirable side effects from TMP-SMX were notably few.