Machine Vision based Computer-Aided Detection of Pulmonary Tuberculosis using Chest X-Ray Images

Tuberculosis (TB) is a lethal disease and developing countries are struggling to overcome this health hazard especially in rural areas and faced globally. Therefore, serious measures are required to reduce this global health hazard. Millary and pulmonary are the most common types of tuberculosis occurring globally. X-ray is the preliminary method to detect tuberculosis; however, the diagnosis is quite often subject to human error. In contrast, the chances of curing Tuberculosis depend on its timely and accurate diagnosis. Therefore, an intelligent machine learning algorithm is developed in this study to assist the clinician in an accurate TB identification in x-ray images. The proposed method pre-processes the X-ray image, enhances its quality and extracts the features of each class which are further passed on to a Deep Convolutional Neural Network-based design for the X-ray image classification, followed by the identification of the tuberculosis type i.e. Millary, Cavitary, Healthy. The classification accuracy for the developed method resulted in 88% and 89% for millary and cavitary TB diseases in x ray images.

Synthesis and Characterization of Coated and Uncoated Ferrite Magnetic Nanoparticles

In this thesis, we studied the structural and magnetic properties of uncoated and coateded ferrite nanoparticles (NPs). Cobalt ferrite (CoFe2O4), maghemite (γ-Fe2O3) and manganese ferrite (MnFe2O4) NPs were preferred due to their potential use in wide range of applications including magnetic fluids, catalysis, biotechnology/biomedicine, magnetic resonance imaging, magneto optical devices, data storage, and environmental remediation. To avoid agglomeration and interparticle interactions and study surface effects in these ferrite NPs, different coating materials such as SiO2 (non-magnetic), Co3O4 (antiferromagnetic) and ZrO2 (non-magnetic) were used. The nature of surface coated material can influence the surface magnetization and modify the interparticle interactions. Initially, SiO2 coated and uncoated CoFe2O4 NPs were prepared by using solgel method. X-ray diffraction (XRD) technique revealed the spinel structure of CoFe2O4 NPs. The average crystallite size was found to be 34 and 25 nm for uncoated and 60 % SiO2 coated CoFe2O4 NPs, respectively. The SiO2 remains amorphous at annealing temperature of 900oC. TEM micrographs revealed spherical NPs with less agglomeration. Surface effects were studied by using AC and DC magnetic measurements. Dynamic scaling law fitting on frequency dependent average blocking temperature (TB) of uncoated CoFe2O4 NPs showed weak spin glass behavior with critical exponent value zv = 4. SiO2 coated CoFe2O4 NPs showed two peaks: blocking at high temperature and freezing at low temperature in imaginary part of frequency dependent AC-susceptibility. The dynamic scaling law fitting on frequency dependent freezing peak in coated NPs showed strong spin-glass state with critical exponent zv = 7 due to stiffed frozen surface spins. However, the blocking peak did not follow the thermal activation. DC field in AC-susceptibility shifted the blocking peak and broadened the freezing peak which also confirmed the existence of spin glass in SiO2 coated CoFe2O4 NPs. We have also prepared CoFe2O4 NPs with ZrO2 and TiO2 coating materials and compared their saturation magnetization (MS) values. M-H loops of SiO2, ZrO2 and TiO2 coated CoFe2O4 NPs revealed that ZrO2 coating reduced more magnetization than SiO2, and TiO2 coating materials. ere studied. XRD confirmed the spinel structure of γ-Fe2O3 NPs with phase of ZrO2. Simulated zero field cooled and field cooled (ZFC/FC) curves gave larger Keff = 1.5 ± 1 x 105 erg/cm3 than Kbulk due to enhanced surface effects by ZrO2 coating. Temperature dependent MS was examined by Bloch‘s law. Bloch‘s constant (B) = 2x10-01 K-b for ZrO2 coated γ-Fe2O3 NPs was higher as compared to SiO2 coating due to weak exchange coupling. Coercivity (HC) of SiO2 and ZrO2 coated γ-Fe2O3 NPs showed increasing trend as decrease in temperature due to enhanced surface anisotropy. Uncoated and SiO2 coated γ-Fe2O3 NPs showed slow spin relaxation due to stronger surface disorders as investigated by stretched exponential law. However, ZrO2 coated γ-Fe2O3 NPs showed weak interactions among NPs and reduced surface effects which were examined by frequency dependent AC-susceptibility data. Finally, surface effects in ZrO2 and Co3O4 coated MnFe2O4 NPs were studied. Simulated ZFC/FC measurement showed larger Keff value of Co3O4 coated MnFe2O4 NPs as compared to ZrO2 coated due to strong coupling between ferrimagnetic (FiM) core and antiferromagnetic (AFM) surface. Bloch‘s law fit showed higher value of B = 0.06 K-b for ZrO2 coating due to decreased in exchange coupling caused by ZrO2 coating. DC field and frequency dependent AC-susceptibility data were analyzed by using Arrhenius, Vogel-Fulcher and dynamic scaling laws for these NPs. Co3O4 coated NPs showed a strong spin glass behavior while ZrO2 coating reduced the interactions between NPs as obtained to ZrO2 coated γ-Fe2O3 NPs. In summary, ZrO2 coating reduced the interparticle interactions between MnFe2O4 and γ-Fe2O3 NPs and also reduced surface effects. However, SiO2 and Co3O4 coating enhanced surface spin disorder and interparticle interactions. Different types of coating can tune the magnetic properties of these ferrite NPs tremendously which make them useful for different applications such as for data storage, hyperthermia cancer treatment, microwave absorber etc.