L S Hibbard1, D W McKeel. 1. Washington University Alzheimer's Disease Research Center, Washington University School of Medicine, St. Louis, Missouri, USA.
Abstract
OBJECTIVE: Senile plaques (SP) are one of the characteristic neuropathologic lesions of Alzheimer's Disease (AD), and studies of SP cortical distribution, density and morphology may lead to new information about the mechanism and pathogenesis of AD. We used an automated, digital image analysis program to detect and measure SP size, shape and total fractional area in digital images of silver-stained tissue sections. STUDY DESIGN: The program observed 94,000 SP in 2,800 digitized microscope fields from tissue sections from 42 postmortem cases ranging from healthy aged to severely demented subjects, studied prospectively before death. RESULTS: Automated pattern recognition can measure SP densities in excellent agreement with an expert and can generate morphometric information not obtainable by conventional microscopy. SP densities (number of SPs/mm2) strongly correlate with tissue load (fraction of tissue area occupied by lesions). SP densities strongly correlate between cortical regions within the same subjects. SP densities, while correlating with the occurrence of AD, do not display a significant trend with respect to dementia severity; likewise, mean SP area and shape properties do not vary significantly with dementia severity. Finally, all the computed SP densities would have produced the same diagnoses of AD in these subjects as the manual SP densities according to the consensus criteria. CONCLUSION: This is the first fully automated program to identify SPs and measure SP morphometry; it uses well-established digital image analysis and statistical pattern recognition methods. The computed SP densities correlate highly with expert results, and the systematic differences are smaller than the interrater differences reported in several multicenter Alzheimer's disease neuropathology studies. The program measures morphometric properties that would be impractical to measure by manual means and, with program-controlled, scanning stage microscopy, can measure lesion densities exhaustively across large cortical areas without stereologic sampling. SP densities rise from near zero to significant values at the mildest diagnosed stage of AD, but beyond this point, there is no demonstrable correlation of density, or any other SP property, with dementia severity. Computed SP densities for even the mildest dementia satisfy the consensus diagnostic criteria.
OBJECTIVE: Senile plaques (SP) are one of the characteristic neuropathologic lesions of Alzheimer's Disease (AD), and studies of SP cortical distribution, density and morphology may lead to new information about the mechanism and pathogenesis of AD. We used an automated, digital image analysis program to detect and measure SP size, shape and total fractional area in digital images of silver-stained tissue sections. STUDY DESIGN: The program observed 94,000 SP in 2,800 digitized microscope fields from tissue sections from 42 postmortem cases ranging from healthy aged to severely demented subjects, studied prospectively before death. RESULTS: Automated pattern recognition can measure SP densities in excellent agreement with an expert and can generate morphometric information not obtainable by conventional microscopy. SP densities (number of SPs/mm2) strongly correlate with tissue load (fraction of tissue area occupied by lesions). SP densities strongly correlate between cortical regions within the same subjects. SP densities, while correlating with the occurrence of AD, do not display a significant trend with respect to dementia severity; likewise, mean SP area and shape properties do not vary significantly with dementia severity. Finally, all the computed SP densities would have produced the same diagnoses of AD in these subjects as the manual SP densities according to the consensus criteria. CONCLUSION: This is the first fully automated program to identify SPs and measure SP morphometry; it uses well-established digital image analysis and statistical pattern recognition methods. The computed SP densities correlate highly with expert results, and the systematic differences are smaller than the interrater differences reported in several multicenter Alzheimer's disease neuropathology studies. The program measures morphometric properties that would be impractical to measure by manual means and, with program-controlled, scanning stage microscopy, can measure lesion densities exhaustively across large cortical areas without stereologic sampling. SP densities rise from near zero to significant values at the mildest diagnosed stage of AD, but beyond this point, there is no demonstrable correlation of density, or any other SP property, with dementia severity. Computed SP densities for even the mildest dementia satisfy the consensus diagnostic criteria.