Gene Expressions, Hippocampal Volume Loss, and MMSE Scores in Computation of Progression and Pharmacologic Therapy Effects for Alzheimer's Disease.

We build personalized relevance parameterization method (prep-ad) based on artificial intelligence (ai) techniques to compute Alzheimer's disease (ad) progression for patients at the mild cognitive impairment (mci) stage. Expressions of ad related genes, mini mental state examination (mmse) scores, and hippocampal volume measurements of mci patients are obtained from the Alzheimer's Disease Neuroimaging Initiative (adni) database. In evaluation of cognitive changes under pharmacological therapies, patients are grouped based on available clinical measurements and the type of therapy administered, namely donepezil monotherapy and polytherapy of donepezil with memantine. Average leave one out cross validation (loocv) error rates are calculated for prep-ad results as less than 8 percent when mmse scores are used to compute disease progression for a 60 month period, and 3 percent with hippocampal volume measurements for 12 months. Statistical significance is calculated as p = 0.003 for using ad related genes in disease progression and as for the results computed by prep-ad. These relatively small average loocv errors and p-values suggest that our prep-ad methods employing gene expressions, mmse scores and hippocampal volume loss measurements can be useful in supporting pharmacologic therapy decisions during early stages of ad.

Mathematical models of tumor growth using Voronoi tessellations in pathology slides of kidney cancer.

The impact of patient-specific spatial distribution features of cell nuclei on tumor growth characteristics was analyzed. Tumor tissues from kidney cancer patients were allowed to grow in mice to apply H&E staining and to measure tumor volume during preclinical phase of our study. Imaging the H&E stained slides under a digital light microscope, the morphological characteristics of nuclei positions were determined. Using artificial intelligence based techniques, Voronoi features were derived from diagrams, where cell nuclei were considered as distinct nodes. By identifying the effect of each Voronoi feature, tumor growth was expressed mathematically. Consistency between the computed growth curves and preclinical measurements indicates that the information obtained from the H&E slides can be used as biomarkers to build personalized mathematical models for tumor growth.

Mathematical models for absorption and efficacy of ovarian cancer treatments.

The creation of personal and individualized anti-cancer treatments has been a major goal in the progression of cancer discovery as evident by the continuous research efforts in genetics and population based PK/PD studies. In this paper we use our clinical decision support tool, called ChemoDSS, to evaluate the effectiveness of three treatments recommended by the NCCN guidelines for ovarian cancer using pre-clinical data from the literature. In particular, we analyze the treatments of PC (i.e., Paclitaxel and Cispaltin), DC (i.e., Docetaxel and Carboplatin), and PBC (i.e., Paclitaxel, Bevacizumab, and Carboplatin). Our in silico analysis of the ovarian cancer treatments shows that PC was the most effective regimen for treating ovarian cancer compared to DC and PBC, which is consistent with literature findings. We demonstrate that we can successfully evaluate the effectiveness of the selected ovarian cancer treatment regimens using ChemoDSS.