https://sinhaharsh.org/publication/ Publications Source Themes Academic (https://sourcethemes.com/academic/)en-us img/map[gravatar:%!s(bool=false) shape:circle] https://sinhaharsh.org/publication/ https://sinhaharsh.org/publication/hasib2026unifying/https://sinhaharsh.org/publication/hasib2026unifying/Large language models (LLMs) and agent systems increasingly support scientific discovery across domains. Yet the literature remains fragmented around isolated tasks, with limited attention to how components form integrated workflows. Here, we introduce exploration, hypothesis generation, and testing (EXHYTE) as an empirical workflow abstraction for reviewing AI-assisted scientific discovery. We mapped a corpus of recent studies to EXHYTE stages and substages to identify recurring strategies, workflow connections, and capability gaps. The analysis shows strong progress in exploration, especially retrieval, knowledge assembly, and representation learning as well as growing capacity for hypothesis and idea generation. The central bottleneck remains the transition from hypothesis or idea generation to executable experimental or computational testing, followed by feedback that can refine subsequent workflow steps. Because EXHYTE organizes systems by workflow function rather than agent architecture, it applies to both multi-agent scaffolds and emerging generalist tool-using agents. It also shows that integrated workflows do not imply full automation: current systems still depend on human judgment for problem formulation, contextualization, interpretation, risk assessment, and oversight. We discuss evaluation, reproducibility, hypothesis-space narrowing, external validation, and governance considerations.https://sinhaharsh.org/publication/sinha2026stile/https://sinhaharsh.org/publication/sinha2026stile/Tissue microarrays (TMAs) enable high-throughput spatial transcriptomic profiling of dozens of tissue cores on a single slide. However, existing dearraying methods operate on histological images and do not support the coordinate-based outputs of spatial transcriptomics platforms. Therefore, task of assigning cells to their respective cores (dearraying) remains a manual bottleneck. We present STiLE, a tool for automated TMA dearraying that operates solely on cell centroid coordinates. By eliminating dependence on image data, STiLE is robust to artifacts such as variable staining quality and uneven illumination. The algorithm combines connectivity-based component detection, density-based clustering (HDBSCAN), component-guided cluster merging, and optional grid-based peak detection. Validation on eleven public TMA samples (50–150 cores, three platforms) achieved ARI > 0.99, while systematic benchmarking on 396 synthetic datasets with realistic artifacts demonstrated consistently robust performance (mean ARI = 0.992). STiLE accepts standard formats (AnnData, CSV) and is platform-agnostic, supporting diverse platforms including Vizgen MERSCOPE, 10x Xenium, and NanoString CosMx. An interactive Streamlit interface enables parameter tuning, visual inspection, and region-based processing for large slides.https://sinhaharsh.org/publication/meng2025spatialatlas/https://sinhaharsh.org/publication/meng2025spatialatlas/Kaposi’s sarcoma (KS) is a highly inflammatory, angiogenic tumor driven by Kaposi’s sarcoma-associated herpesvirus (KSHV), yet the origins of tumor cells and mechanisms of progression remain unclear. Here, we present the first spatial single-cell atlas of KS, profiling 256 samples across patch, plaque, and nodular lesions and normal controls. We identify CD34⁺ progenitor lymphatic endothelial cells (LECs) as the primary targets of KSHV, whose clonal expansion drives tumor growth. KSHV infection induces widespread cellular reprogramming across the tumor microenvironment, including LECs, vascular endothelial cells, fibroblasts, and macrophages, generating hybrid phenotypes that support angiogenesis, inflammation, and immune modulation. KSHV⁺ macrophages are enriched in tumor-proximal niches, further promoting a proangiogenic, immunosuppressive environment. Spatial analysis reveals evolving tumor-associated niches, with a core-to-periphery gradient correlating with infection, immune modulation, and cellular remodeling. We identify disease progression predictive signatures, offering mechanistic insights into KS pathogenesis and potential new therapeutic strategies by reprogramming the tumor microenvironment.https://sinhaharsh.org/publication/sinha2024mrqa/https://sinhaharsh.org/publication/sinha2024mrqa/Pooling data across diverse sources acquired by multisite consortia requires compliance with a predefined reference protocol i.e., ensuring different sites and scanners for a given project have used identical or compatible MR physics parameter values. Traditionally, this has been an arduous and manual process due to difficulties in working with the complicated DICOM standard and lack of resources allocated towards protocol compliance. Moreover, issues of protocol compliance is often overlooked for lack of realization that parameter values are routinely improvised/modified locally at various sites. The inconsistencies in acquisition protocols can reduce SNR, statistical power, and in the worst case, may invalidate the results altogether. An open-source tool, mrQA was developed to automatically assess protocol compliance on standard dataset formats such as DICOM and BIDS, and to study the patterns of non-compliance in over 20 open neuroimaging datasets, including the large ABCD study. The results demonstrate that the lack of compliance is rather pervasive. The frequent sources of non-compliance include but are not limited to deviations in Repetition Time, Echo Time, Flip Angle, and Phase Encoding Direction. It was also observed that GE and Philips scanners exhibited higher rates of non-compliance relative to the Siemens scanners in the ABCD dataset. Continuous monitoring for protocol compliance is strongly recommended before any pre/post-processing, ideally right after the acquisition, to avoid the silent propagation of severe/subtle issues. Although, this study focuses on neuroimaging datasets, the proposed tool mrQA can work with any DICOM-based datasets.https://sinhaharsh.org/publication/sinha2024mrdataset/https://sinhaharsh.org/publication/sinha2024mrdataset/Large-scale medical imaging datasets (e.g., MR, PET, CT) play a pivotal role in healthcare research. However, the reliability of their statistical results is contingent on the quality of the acquired data. These datasets are often acquired at multiple sites using diverse scanners and formats. Divergent practices at individual sites and scanners lead to differences in acquisition and data organization. Manually reconciling these variations is not only time-consuming and labor-intensive but also error-prone. To address this need, we developed MRdataset, a Python package that serves as a seamless interface to access diverse medical imaging dataset formats. MRdataset eliminates the burden of extensive re-formatting by accounting for differences in data organization, scanner specifications and various imaging formats.https://sinhaharsh.org/publication/sinha2023shape/https://sinhaharsh.org/publication/sinha2023shape/CNNs have emerged as powerful techniques for object recognition. However, the test performance of CNNs is contingent on the similarity to training distribution. Existing methods focus on data augmentation to address out-of-domain generalization. In contrast, we enforce a shape bias by encouraging our model to learn features that correlate with those learned from the shape of the object. We show that explicit shape cues enable CNNs to learn features that are robust to unseen image manipulations i.e. novel textures with the same semantic content. Our models are validated on Toys4K dataset which consists of 4179 3D objects and image pairs. To quantify texture bias, we synthesize dataset variants called Style (style-transfer with GANs), CueConflict (conflicting texture & semantics), and Scrambled datasets (obfuscating semantics by scrambling pixel blocks). Our experiments show that the benefits of using shape is not subject to specific shape representations like point clouds, rather the same benefits can be obtained from a simpler representation such as the distance transform.https://sinhaharsh.org/publication/xing2021cycle/https://sinhaharsh.org/publication/xing2021cycle/Modern generative adversarial networks (GANs) have been enabling the realistic generation of full 3D brain images by sampling from a latent space prior Z (i.e., random vectors) and mapping it to realistic images in X (e.g., 3D MRIs). To address the ubiquitous mode collapse issue, recent works have strongly imposed certain characteristics such as Gaussianness to the prior by also explicitly mapping X to Z via encoder. These efforts, however, fail to accurately map 3D brain images to the desirable prior, which the generator assumes to be sampling the random vectors from. On the other hand, Variational Auto-Encoding GAN (VAE-GAN) solves mode collapse by enforcing Gaussianness by two learned parameter, yet causes blurriness in images. In this work, we show how our cycle consistent embedding GAN (CCE-GAN) both accurately encodes 3D MRIs to the standard normal prior, and maintains the quality of the generated images. We achieve this without a network-based code discriminator via the Wasserstein measure. We quantitatively and qualitatively assess the embeddings and the generated 3D MRIs using healthy T1-weighted MRIs from ADNI.https://sinhaharsh.org/publication/muraliunderreviewischemia/https://sinhaharsh.org/publication/muraliunderreviewischemia/Carotid endarterectomy carries the risk of intraoperative cerebral ischemia, which is monitored by expert neurophysiologists through continuous electroencephalography (cEEG). Because expert availability is limited, we developed a hybrid novice–artificial intelligence (AI) system that detects ischemia using novice monitors with limited cEEG training. The hybrid system dynamically weights novice and AI inputs to arrive at a final output. Using four novices, we compared hybrid systems against experts alone, novices alone, and AI alone. Hybrid systems were statistically non-inferior to experts in sensitivity and false-positive rate (FPR), whereas novices alone were not. At 80% sensitivity, hybrid systems reduced FPR by half compared with the AI-only system, with similar benefits at 90% sensitivity. Further, the area under the precision-recall curve improved from 0.546 to 0.610–0.726, the area under the receiver operating curve improved from 0.957 to 0.967–0.971, and calibration improved compared with AI alone. These results highlight the potential of a hybrid system to monitor intraoperative cerebral ischemia.https://sinhaharsh.org/publication/dasunderreviewspatialytica/https://sinhaharsh.org/publication/dasunderreviewspatialytica/Spatial transcriptomics and proteomics map tissue architecture and cellular interactions, but analysis remains limited by programming demands and text-centered AI agents that lack viewer grounding and cross-turn context. We present spatiAlytica, a viewer-centric multimodal interactive agentic system embedded in the Napari viewer that enables non-programmer biologists to perform iterative, hypothesis-driven spatial omics analysis via natural language. spatiAlytica couples viewer-state serialization, agentic memory, biological concept-to-data-field mapping, code generation and debugging, Spatial VQA, and grounded interpretation to support an exploratory analysis and interpretive reasoning workflow. We introduce spatiAlyticaBench, a comprehensive benchmark spanning 222 single-turn spatial analytical coding questions, 178 multi-turn sequential workflow questions, and 7,350 image-grounded reasoning questions. spatiAlytica outperformed strong agentic baselines, while using less time and tokens. Case studies across Kaposi’s sarcoma, colorectal cancer, and ovarian cancer recapitulated known spatial patterns and uncovered progressive CD8 T-cell dysfunction during KS progression. https://sinhaharsh.org/publication/sinha2021learning/ Fri, 06 Nov 2020 15:44:46 +0530 https://sinhaharsh.org/publication/sinha2021learning/ https://sinhaharsh.org/publication/mehta2021domain/ Mon, 02 Nov 2020 16:21:27 +0530 https://sinhaharsh.org/publication/mehta2021domain/ https://sinhaharsh.org/publication/mehta2020hidegan/ Mon, 20 Apr 2020 09:55:30 +0530 https://sinhaharsh.org/publication/mehta2020hidegan/ https://sinhaharsh.org/publication/sinha2020text/ Sun, 05 Apr 2020 16:18:20 +0530 https://sinhaharsh.org/publication/sinha2020text/ https://sinhaharsh.org/publication/sinha2020audio/ Sun, 05 Apr 2020 16:16:17 +0530 https://sinhaharsh.org/publication/sinha2020audio/ https://sinhaharsh.org/publication/sinha2019upgrading/ Tue, 29 Jan 2019 16:18:48 +0530 https://sinhaharsh.org/publication/sinha2019upgrading/ https://sinhaharsh.org/publication/sinha2019convolutional/ Wed, 31 Oct 2018 16:18:37 +0530 https://sinhaharsh.org/publication/sinha2019convolutional/ https://sinhaharsh.org/publication/sinha2018studying/ Mon, 01 Oct 2018 16:19:02 +0530 https://sinhaharsh.org/publication/sinha2018studying/ https://sinhaharsh.org/publication/sinha2018interweaving/ Thu, 23 Aug 2018 16:19:14 +0530 https://sinhaharsh.org/publication/sinha2018interweaving/