Testing the Reference of Biological Kind Terms.

Recent experimental work on "natural" kind terms has shown evidence of both descriptive and nondescriptive reference determination. This has led some to propose ambiguity or hybrid theories, as opposed to traditional description and causal-historical theories of reference. Many of those experiments tested theories against referential intuitions. We reject this method, urging that reference should be tested against usage, preferably by elicited production. Our tests of the usage of a biological kind term confirm that there are indeed both descriptive and causal-historical elements to the reference determination of some natural kind terms. We argue that to accommodate our results and earlier ones, we should abandon the common assumption that any one theory of reference fits all natural kind terms. Rather, it is likely that some terms are descriptive, some causal-historical, some ambiguous, and some hybrid. This substantive conclusion is accompanied by a methodological one. Our experiments, like some earlier ones, found participants contradicting both each other and themselves. We argue that these contradictions indicate a lack of linguistic competence with the term. We conclude that these experiments have been faulty, because they test terms that are novel to participants and/or use fantastical vignettes. We provide some suggestions for future research.

Prostate cancer: three-dimensional sonoelastography for in vitro detection.

PURPOSE: To prospectively evaluate the accuracy of three-dimensional (3D) sonoelastographic imaging, relative to that of gray-scale ultrasonography (US), in the in vitro detection of prostate cancer. MATERIALS AND METHODS: The study was approved by the institutional review board and was HIPAA compliant. Informed consent was obtained from all patients. Nineteen prostatectomy specimens from patients aged 46-70 years with biopsy-proved prostate cancer were scanned in three dimensions by using conventional B-mode US and sonoelastography with vibrations of more than 100 Hz. Step-sectioned whole-mount histologic specimens were used to create a 3D volume of the prostate and the tumors within it. B-mode US scans and regions of low vibration on the sonoelastographic images (hard regions) were formatted in three dimensions. The lesions in the 19 cases were classified into two groups, as follows: G1 lesions were pathologically confirmed tumors with a volume of at least 1.0 cm3, and G2 lesions were pathologically confirmed tumors smaller than 1.0 cm3. G1 lesions were evaluated with B-mode US and sonoelastography and classified as true-positive, false-positive, true-negative, or false-negative; G2 lesions were evaluated only with sonoelastography. Findings at histologic examination were used as the reference standard. True-positive findings necessitated 3D lesion correlation between pathologic and imaging data. Conventional definitions of accuracy and sensitivity were used for statistical analysis. RESULTS: For G1 lesions (seven lesions with a volume of at least 1.0 cm3), sonoelastography had an accuracy of 55% and a sensitivity of 71% and B-mode US had an accuracy of 17% and a sensitivity of 29%. The mean tumor volume was 3.1 cm3 +/- 2.1 (standard deviation). For G2 lesions (22 lesions with a volume of less than 1.0 cm3), the mean tumor volume was 0.32 cm3 +/- 0.21. Sonoelastography had an accuracy of 34% and a sensitivity of 41%; there were six false-positive findings. CONCLUSION: Sonoelastography performed considerably better than did gray-scale US in the depiction of prostate cancer for tumors with volumes of more than 1 cm3.

Three-dimensional registration of prostate images from histology and ultrasound.

A whole mount histology protocol for 3-D tissue reconstruction to compare the size and spatial location of tumors (and other components) identified in histology data with that from 3-D ultrasound (US) images is presented. Prostate specimens are imaged in 3-D using B-mode (US) and sonoelastography. The prostate surface is outlined in each B-mode image and a 3-D surface reconstruction is made. The specimen is then prepared for whole mount histology and the histology slides are digitally reconstructed to make a 3-D surface. These two surfaces are then aligned using a 3-D correlation algorithm, and the tumor boundary determined by the pathologist is compared with that using sonoelastography. 3-D images showing the overlapping histology and sonoelastography of prostate surface reconstructions for one prostate are presented to illustrate the technique; results for four prostates yielded an accuracy of 92% +/- 3%.