Academic Software Demos

University of Edinburgh

Topic Overview: DeepInverse is a PyTorch-based library for solving imaging inverse problems using deep learning. The demo will showcase the library’s capabilities in handling various imaging modalities and reconstruction tasks using state-of-the-art deep learning (DL) algorithms, with hands-on examples of implementing forward operators, training and using networks, and deploying reconstruction pipelines. Website (including all Examples and User Guide): https://deepinv.github.io GitHub: https://github.com/deepinv/deepinv

University Hospital Bonn

Overview and the learning outcome: Segment Anything Models (SAM1 and SAM2) has impacted various fields, including medical imaging. However, existing integrations of SAM in medical images primarily focus on applications that require local installation and configuration. We introduce a web-based SAM2 extension integrated into the OHIF viewer, supporting all SAM2 prompts, including text-based inputs via Grounding DINO. This eliminates installation barriers, offering a user-friendly interface. A demo session will showcase the capabilities and limitations of foundation models in medical imaging.

University of Oulu

Overview: The Open Lumbar Spine Image Analysis Platform (OLSIA) is developed for automated analysis of lumbar spine MRI. Leveraging deep learning-based segmentation and grading models, OLSIA enables the precise identification of vertebral bodies (L1-S1) and intervertebral discs (L1/2-L5/S1). Our platform facilitates comprehensive reporting by implementing the Pfirrmann classification system for the objective quantification of disc degeneration and by computing the disc height index (DHI), a metric indicative of intervertebral disc integrity. Designed for efficient batch processing of MRI datasets, OLSIA enhances research and clinical workflows.

Alan Turing Institute

Overview: Fluorescent activated cell sorting (FACS) is a flow cytometry-based multiparameter high-throughput purification strategy for single-cell experiments. Of particular interest is the use of FACS for enrichment of cells that are rare or atypically shaped. Enrichment for such atypically shaped cells can benefit most from new imaging flow cytometry techniques that provide both morphological features and imaging capabilities, in addition to the spectral features obtained from standard flow cytometry assays. Flow cytometry Feature Importance (FlowFI) provides an integrated feature analysis and design platform for choosing and creating new imaging flow cytometry features that are specific to the morphology of the cell in question. These features can then be used to refine the cell purification strategy for downstream analysis, such as single cell transcriptomic or morphogenomic analysis. FlowFI is available for installation and download with a tutorial on how to use the software in combination with real data.

Center for Biomedical Image Computing & Analytics (CBICA), University of Pennsylvania

Brief Topic Overview: NiChart is a machine learning focused neuroimaging platform that streamlines MRI data analysis in clinical settings. It extracts personalized and clinically relevant biomarkers from brain MRI data and enables comparisons of the results across a reference dataset with over 75,600 scans from 25 studies. This demo provides an overview of NiChart’s components by demonstrating the structural MRI (sMRI) pipeline on our Cloud portal. We’ll also introduce packages for functional MRI (fMRI) and diffusion MRI (dMRI) analysis. Attendees will learn how to utilize different components of NiChart to derive imaging-derived phenotypes from their own data, as well as how to visualize and compare their results against NiChart-based normative ranges or distributions from specific disease subgroups. Lastly, users will be provided with a comprehensive report summarizing the insights drawn from the prior analysis.

Georgia State University

Overview: Advances in modern artificial intelligence (AI) technologies have opened unprecedented opportunities in the neuroimaging sector, particularly in accelerating semantic brain segmentation, surface generation, flagging potential tumors in Magnetic Resonance Images (MRIs) of the brain, showing promise to greatly expedite the radiology workflow. Despite the justified excitement and expected productivity boost, patient care improvement, and reduction of the overhead in research from bringing these tools to the hands of individual researcher and clinical practitioner at scale, only a select few can currently enjoy the benefits of AI. The main reasons are, roughly, requirements of large-scale hardware and unreasonably high expectations of technical skills from the users of AI.

Savvy machine learning engineers, in time, are able to utilize the most recent, newly published and accessible AI technologies on their own laptops and workstations, after some hours of configuration, library installation, and harmonization. Yet, it can’t be expected that every doctor, clinician, or radiologist also possess the same level of expertise in machine learning engineering.

University of Texas MD Anderson Cancer Center

Overview:NeuroAnalyst is an open-source software platform designed to accelerate neuroimaging research by combining reproducible analysis pipelines with contextual knowledge retrieval (KR). It addresses the critical challenges of reproducibility and interpretability in neuroimaging by providing – a) Containerized Processing Framework-Automates standardized workflows for structural, functional, and diffusion MRI, multimodal integration, and longitudinal analysis, with real-time progress tracking and quality assurance, b) Knowledge Retrieval System – Uses large language models (LLMs) to contextualize pipeline results by synthesizing insights from scientific literature, enabling researchers to ask natural language questions about their data. Attendees will learn to deploy containerized neuroimaging workflows using for reproducible analysis, query pipeline results using the KR system to obtain contextualized answers grounded in literature, implement FAIR principles through automated metadata tracking, and leverage integrated quality control metrics to troubleshoot pipeline outputs effectively. NeuroAnalyst’s relevance to the ISBI community includes its reproducibility focus through containerized pipelines, interpretability innovation via the KR system bridging computational outputs with actionable insights, and acceleration of research by streamlining workflows and integrating knowledge to reduce discovery time. NeuroAnalyst has also been accepted to be presented as a power pitch for ISMRM 2025. Target audience includes early-career to senior investigators in neuroimaging and biomedical imaging, and clinicians interested in integrating imaging biomarkers into practice. Basic familiarity with neuroimaging concepts is required. Programming knowledge is necessary only if users desire to create their own containerized processes, using whatever language they prefer.