New approaches and algorithms are developed for the identification and estimation of low order models that represent multipath channel effects in Code Division Multiple Access (CDMA) communication systems. Based on these parsimonious channel models, low complexity receivers such as RAKE receivers are considered to exploit these propagation effects and enhance the system performance. We consider the scenario where multipath is frequency selective slowly fading and where the channel components including delays and attenuation coefficients are assumed to be constant over one or few signalling intervals. We model the channel as a long FIR-like filter (or a tapped delay line filter) with the number of taps related to the ratio between the channel delay-spread and the chip duration. Due to the high data rate of new CDMA systems, the channel length in terms of the chip duration will be very large. With classical channel estimation techniques this will result in poor estimates of many of the channel parameters where most of them are zero leading to a reduction in the system performance. Unlike classical techniques which estimate directly the channel response given the number of taps or given an estimate of the channel length, the proposed techniques in this work will firstly identify the significant multipath parameters using model selection techniques, then estimate these identified parameters. Statistical tests are proposed to determine whether or not each individual parameter is significant. A low complexity RAKE receiver is then considered based on estimates of these identified parameters only. The level of significance with which we will make this assertion will be controlled based on statistical tests such as multiple hypothesis tests. Frequency and time domain based approaches and model selection techniques are proposed to achieve the above proposed objectives. / The frequency domain approach for parsimonious channel estimation results in an efficient implementation of RAKE receivers in DS-CDMA systems. In this approach, we consider a training based strategy and estimate the channel delays and attenuation using the averaged periodogram and modified time delay estimation techniques. We then use model selection techniques such as the sphericity test and multiple hypotheses tests based on F-Statistics to identify the model order and select the significant channel paths. Simulations show that for a pre-defined level of significance, the proposed technique correctly identifies the significant channel parameters and the parsimonious RAKE receiver shows improved statistical as well as computational performance over classical methods. The time domain approach is based on the Bootstrap which is appropriate for the case when the distribution of the test statistics required by the multiple hypothesis tests is unknown. In this approach we also use short training data and model the channel response as an FIR filter with unknown length. Model parameters are then estimated using low complexity algorithms in the time domain. Based on these estimates, bootstrap based multiple hypotheses tests are applied to identify the non-zero coefficients of the FIR filter. Simulation results demonstrate the power of this technique for RAKE receivers in unknown noise environments. Finally we propose adaptive blind channel estimation algorithms for CDMA systems. Using only the spreading code of the user of interest and the received data sequence, four different adaptive blind estimation algorithms are proposed to estimate the impulse response of frequency selective and frequency non-selective fading channels. Also the idea is based on minimum variance receiver techniques. Tracking of a frequency selective varying fading channel is also considered. / A blind based hierarchical MDL model selection method is also proposed to select non-zero parameters of the channel response. Simulation results show that the proposed algorithms perform better than previously proposed algorithms. They have lower complexity and have a faster convergence rate. The proposed algorithms can also be applied to the design of adaptive blind channel estimation based RAKE receivers.

Immunotherapy of cancer has been shown to be promising in prolonging patient survival. However, complete elimination of cancer and life-long relapse-free survival remain to be major challenge for anti-cancer therapeutics. We have previously reported that ex vivo reprogramming of tumor-sensitized immune cells by bryostatin 1/ionomycin (B/I) and the gamma-chain (γ-c) cytokines IL-2, IL-7, and IL-15 resulted in the generation of memory T cells as well as CD25+ NKT cells and CD25+ NK cells. Adoptive cellular therapy (ACT) utilizing these reprogrammed immune cells protected FVBN202 mice from tumor challenge, and overcame the suppressive functions of myeloid-derived suppressor cells (MDSCs). We then demonstrated that the presence of CD25+ NKT cells was required for anti-tumor efficacy of T cells as well as their resistance to MDSCs. Similar results were obtained by reprogramming of peripheral blood mononuclear cells (PBMC) from patients with early stage breast cancer, demonstrating that an increased frequency of CD25+ NKT cells in reprogrammed immune cells was associated with modulation of MDSCs to CD11b-HLA-DR+ immune stimulatory cells. Here, we tested the efficacy of immunotherapy in a therapeutic setting against established primary breast cancer (Chapter One), experimental metastatic breast cancer (Chapter Three) as well as against minimal residual disease (MRD) in patients with multiple myeloma (Chapter Two). We evaluated the ability of reprogrammed immune cells, including CD25+ NKT cells, to convert MDSCs to myeloid immune stimulatory cells, in vivo; this resulted in the identification and characterization of a novel antigen presenting cell (APC). These novel immune stimulatory cells differed from conventional APCs, including dendritic cells (DCs) and macrophages. We have also demonstrated that enhancing immunogenicity of mammary tumors by treatment with Decitabine (Dec) along with overcoming MDSCs by utilizing reprogrammed T cells and NKT cells in ACT prolongs survival of animals, but fails to eliminate the tumor. However, targeting cancer during a setting of MDR, when tumor cells are dormant, results in objective responses as evidenced in our multiple myeloma studies. This suggests that targeting breast cancer with immunotherapy following conventional therapies, in a setting of residual disease when tumor cells are dormant, may be effective in eliminating such residual cells or maintaining dormancy and extending time-to-relapse for breast cancer patients.