http://repository.iiitd.edu.in/xmlui/handle/123456789/1302 2026-04-10T22:01:12Z http://repository.iiitd.edu.in/xmlui/handle/123456789/1699 Learning from high dimensional healthcare data to improve interpretability and insights Jha, Indra Prakash; Kumar, Vibhor (Advisor) The escalating volume and intricate nature of healthcare and social care datasets necessitate the implementation of unconventional feature learning strategies to tackle current challenges. The examination of health and biological datasets enables the assessment of existing computational techniques and fosters the creation of novel algorithms and methodologies that can be applied to difficulties in other fields. By employing these concepts, we have not only developed novel algorithms but also performed meticulous analysis to tackle concerns pertaining to healthcare and social care. It should be noted that the methodologies and analysis processes can be adapted to accommodate supplementary datasets featuring diverse data types and formats. Firstly, the authors present a novel manifold learning algorithm, named "Topological Preservation and Distance Scaling" (TPDS), which aims to enhance classification and visualization of high-dimensional datasets. The proposed method addresses the challenge of the "curse of dimensionality". The methodology aims to maintain the hierarchical structure of data by preserving both local topology and distances during linear and non-linear dimension reduction. This approach is designed to prevent the collapse of data points in visualization. In the second study, the authors present a novel matrix factorization-based manifold learning algorithm, "Network Inference in Reduced Dimensions” (NIRD), for inferring very large regulatory networks with a very large number of features. The study revealed that the proposed approach exhibited superior performance compared to existing methods, namely GENIE-3 and GrnBoost2, in terms of both the computational time required to infer the network and the accuracy of estimated edges or connections. The objective was to deduce intricate dependency and regulatory networks that encompass a vast number of dimensions, with the aim of capturing non-linear dependencies among random variables. Subsequently, two causal discovery analyses were conducted on high-dimensional healthcare datasets to infer "explainable" associations and estimate public health concerns, such as the prevalence of mental health. The hypothesis posits that the utilization of generative probabilistic graphical models, specifically the Bayesian network and Markov network, in tandem with the Markov blanket concept of feature learning may yield greater interpretability. The initial investigation involved the utilization of survey data collected from a diverse group of American adults, encompassing various age groups, genders, and socioeconomic statuses. In contrast, the subsequent inquiry employed data from the Longitudinal Ageing Study in India (LASI) Wave-1 survey, which focused on elderly individuals residing in India. The methodology employed facilitated the determination of the most relevant attributes (driver factors) that could effectively represent the incidence of mental health disorders in both studies. The features chosen by our approach demonstrate relevance in facilitating actionable interventions aimed at promoting mental health well-being among adults during pandemic-induced lockdowns, as well as among elderly individuals in India. 2023-05-01T00:00:00Z http://repository.iiitd.edu.in/xmlui/handle/123456789/1372 Machine learning approaches in cancer detection and treatment Sarita; Sengupta, Debarka (Advisor); Kumar, Lalit (advisor) Cancer has become the second leading cause of mortality worldwide, and early de-tection and adequate treatment are crucial in reducing the cancer burden. Metastasis,which involves malignant cells detaching from the primary tumor and colonizing otherdistant organs, is the leading cause of cancer-related deaths. The microenvironment,immune cells, stromal cells, and drug selection pressures influence tumors’ hetero-geneity and dynamicity, making it challenging to select the most effective treatmentapproach throughout the entire course of the disease. Liquid biopsy and single-celltranscriptomics have emerged as promising techniques for cancer detection. Bodily flu-ids such as blood, urine, and saliva provide rich biomarkers. Circulating tumor cellsand other tumor-associated cell products have been identified in the bloodstream, pro-viding potential biomarkers for cancer detection. Through serial blood analysis, liquidbiopsy techniques can help track spatial and temporal heterogeneity in tumor biology.Characterizing circulating tumor cells (CTCs) provides essential biological informa-tion about the disease as they are the primary live tumor cells responsible for metas-tasis. Existing CTC detection methods rely on surface markers, which may be shedduring the epithelial-to-mesenchymal (EMT) process or due to various stressors in theblood. Therefore, marker-free detection and characterization of CTCs are necessary.To achieve the best possible outcomes, it is crucial to manage cancer and any clinicalfactors that may impact treatment response or contribute to disease relapse. By identi-fying and addressing these factors, healthcare providers can develop effective treatmentplans and improve overall cancer management. This approach can help patients achievelonger-term remission and better quality of life.Over the past two decades, machine learning (ML) has shown tremendous potentialin enhancing cancer diagnosis and treatment accuracy and efficiency. Our researchleverages the power of ML to address the pressing need for timely cancer detectionand optimal management of the disease. By employing advanced ML algorithms, weaimed to improve the accuracy and speed of cancer diagnosis, identify the most effectivetreatment options, and enable personalized cancer care. For marker-free detection and characterization of CTCs, we created a novel unsu-pervised clustering algorithm, unCTC, which can leverage single-cell transcriptomicdata to detect and characterize CTCs. In unCTC, a wide range of computational andstatistical modules are integrated, such as novel Deep Dictionary Learning with k-meansClustering Cost (DDLK) approach for scRNA-Seq clustering, expression-based infer-ence of copy number variation (CNV), and combinatorial, marker-based validation ofmalignant phenotypes. DDLK provides a robust separation of circulating tumor cells(CTCs) and white blood cells (WBCs) in the pathway space, unlike the gene expressionspace. The utility of unCTC was validated on single-cell RNA sequencing (scRNA-Seq)profiles of breast CTCs from six patients. 2023-12-01T00:00:00Z http://repository.iiitd.edu.in/xmlui/handle/123456789/1303 Modeling the impact of social determinants of health and public health interventions Awasthi, Raghav; Sethi, Tavpritesh (Advisor) Social determinants of health (SDOH) are the non-medical factors that play a vital role in public health and form the basis of health policies. Building an effective public health policy is a complex endeavor that systematically understands multi-dimensional associations. The association of SDOH with public health outcomes is poorly understood due to the complex interplay of factors. Artificial Intelligence (AI) advancements have enabled models to make robust and explainable decisions in complex environments. Causal modeling and counterfactual analysis infer direct and indirect associations from observational data and rank policy indicators. Reinforcement learning (RL) is a paradigm for sequential decision-making under uncertainty that infers the policy after simulating sequential actions and accompanying rewards from the context. However, systematic application and utilization of these models are limited to public health settings. Through our work, we’ve made an integrative model that can measure complex interdependencies, bring together different kinds of knowledge about health system indicators, and use SDOH to guide public health interventions. In our first contribution, we contribute to the discovery of potential interventions using an integrative machine learning framework incorporating structural causal models, counterfactual analysis, and predictive modeling to discover policy interventions. Using this framework, we found policy solutions for three use cases presented, i.e. (i) vi antimicrobial resistance, (ii) mitigating the spread of HIV among women who work in the sex industry, and (iii) targeted interventions to improve mental health. In our second contribution, we built a novel framework for optimizing potential interventions using reinforcement learning. Here we showcase a model to optimally allocate COVID-19 in the context of different SDOHs for states of India. This use case also aims to generalize the reinforcement learning framework for optimizing healthcare resource allocation. Our final contribution is the real-world deployment of this policy-discovering and policy-optimizing AI model in response to COVID-19. Our machine learning models powered several web applications that provided forecasts of COVID-19 trajectory, mined emerging evidence from rapidly emerging COVID-19 literature, and for optimal allocation of COVID-19 vaccines. 2023-06-03T00:00:00Z