Global effects of aging on the hemodynamic response function in the human brain.

In functional magnetic resonance imaging, the hemodynamic response function (HRF) is a transient, stereotypical response to local changes in cerebral hemodynamics and oxygen metabolism due to briefly (< 4 s) evoked neural activity. Accordingly, the HRF is often used as an impulse response with the assumption of linearity in data analysis. In cognitive aging studies, it has been very common to interpret differences in brain activation as age-related changes in neural activity. Contrary to this assumption, however, evidence has accrued that normal aging may also significantly affect the vasculature, thereby affecting cerebral hemodynamics and metabolism, confounding interpretation of fMRI aging studies. In this study, use was made of a multisensory stimulus to evoke the HRF in ~ 87% of cerebral cortex in cognitively intact adults with ages ranging from 22-75 years. The stimulus evokes both positive and negative HRFs, which were characterized using model-free parameters in native-space coordinates. Results showed significant age trends in HRF parameter distributions in terms of both amplitudes (e.g., peak amplitude and CNR) and temporal dynamics (e.g., full-width-at-half-maximum). This work sets the stage for using HRF methods as a biomarker for age-related pathology.

Effects of Thresholding on Voxel-Wise Correspondence of Breath-Hold and Resting-State Maps of Cerebrovascular Reactivity.

Functional magnetic resonance imaging for presurgical brain mapping enables neurosurgeons to identify viable tissue near a site of operable pathology which might be at risk of surgery-induced damage. However, focal brain pathology (e.g., tumors) may selectively disrupt neurovascular coupling while leaving the underlying neurons functionally intact. Such neurovascular uncoupling can result in false negatives on brain activation maps thereby compromising their use for surgical planning. One way to detect potential neurovascular uncoupling is to map cerebrovascular reactivity using either an active breath-hold challenge or a passive resting-state scan. The equivalence of these two methods has yet to be fully established, especially at a voxel level of resolution. To quantitatively compare breath-hold and resting-state maps of cerebrovascular reactivity, we first identified threshold settings that optimized coverage of gray matter while minimizing false responses in white matter. When so optimized, the resting-state metric had moderately better gray matter coverage and specificity. We then assessed the spatial correspondence between the two metrics within cortical gray matter, again, across a wide range of thresholds. Optimal spatial correspondence was strongly dependent on threshold settings which if improperly set tended to produce statistically biased maps. When optimized, the two CVR maps did have moderately good correspondence with each other (mean accuracy of 73.6%). Our results show that while the breath-hold and resting-state maps may appear qualitatively similar they are not quantitatively identical at a voxel level of resolution.

Using data-driven sublanguage pattern mining to induce knowledge models: application in medical image reports knowledge representation.

BACKGROUND: The use of knowledge models facilitates information retrieval, knowledge base development, and therefore supports new knowledge discovery that ultimately enables decision support applications. Most existing works have employed machine learning techniques to construct a knowledge base. However, they often suffer from low precision in extracting entity and relationships. In this paper, we described a data-driven sublanguage pattern mining method that can be used to create a knowledge model. We combined natural language processing (NLP) and semantic network analysis in our model generation pipeline. METHODS: As a use case of our pipeline, we utilized data from an open source imaging case repository, Radiopaedia.org , to generate a knowledge model that represents the contents of medical imaging reports. We extracted entities and relationships using the Stanford part-of-speech parser and the "Subject:Relationship:Object" syntactic data schema. The identified noun phrases were tagged with the Unified Medical Language System (UMLS) semantic types. An evaluation was done on a dataset comprised of 83 image notes from four data sources. RESULTS: A semantic type network was built based on the co-occurrence of 135 UMLS semantic types in 23,410 medical image reports. By regrouping the semantic types and generalizing the semantic network, we created a knowledge model that contains 14 semantic categories. Our knowledge model was able to cover 98% of the content in the evaluation corpus and revealed 97% of the relationships. Machine annotation achieved a precision of 87%, recall of 79%, and F-score of 82%. CONCLUSION: The results indicated that our pipeline was able to produce a comprehensive content-based knowledge model that could represent context from various sources in the same domain.

Semantic-Enhanced Query Expansion System for Retrieving Medical Image Notes.

Most current image retrieval methods require constructing semantic metadata for representing image content. To manually create semantic metadata for medical images is time-consuming, yet it is a crucial component for query expansion. We proposed a new method for searching medical image notes that uses semantic metadata to improve query expansion and leverages a knowledge model developed specifically for the medical image domain to create relevant metadata. We used a syntactic parser and the Unified Medical Language System to analyze the corpus and store text information as semantic metadata in a knowledge model. Our new method has an interactive interface that allows users to provide relevance feedback and construct new queries more efficiently. Sixteen medical professionals evaluated the query expansion module, and each evaluator had prior experience searching for medical images. When using the initial query as the baseline standard, expanded queries achieved a performance boost of 22.6% in terms of the relevance score on first ten results (P-value<0.05). When using Google as another baseline, our system performed 24.6% better in terms of relevance score on the first ten results (P-value<0.05). Overall, 75% of the evaluators said the semantic-enhanced query expansion workflow is logical, easy to follow, and comfortable to use. In addition, 62% of the evaluators preferred using our system instead of Google. Evaluators who were positive about our system found the knowledge map-based visualization of candidate medical search terms helpful in refining cases from the initial search results.