Data Science Center in Health (DASH)’s Post

Exploring Counterfactual Image Generation in Biomedical Research Continuing our series on innovative research, we turn to a new benchmarking framework for counterfactual image generation, aiming to push the boundaries in medical imaging. This approach involves evaluating how well different models can generate realistic and causally consistent images, enabling new possibilities in understanding medical scenarios. Key Takeaways: • Hierarchical VAEs Shine: HVAE models demonstrate superior performance across datasets, including natural and medical images. • SCM-Based Evaluation: Utilizes Structural Causal Models (SCM) for realistic edits, maintaining the integrity of original data. • Comprehensive Metrics: The study compares image quality and causal consistency, offering a new standard for the field. https://lnkd.in/etfdpu2h Thomas Melistas Nefeli Gkouti Pedro P. Sanchez Dr Athanasios Vlontzos Yannis Panagakis Giorgos Papanastasiou, PhD Sotirios (Sotos) Tsaftaris #GenerativeAI #BiomedicalResearch #AIinHealthcare #InnovationInMedicine

Our work with King's College London (KCL) School of Biomedical Engineering & Imaging Sciences to design machine learning approaches to evaluate imaging of patients with rheumatoid arthritis using maraciclatide is continuing to make good progress. Congratulations to our KCL colleagues Andrew Reader, Gary Cook and Robert Cobb who have had a paper on their ongoing research titled: “Improved Classification Learning from Highly Imbalanced Multi-Label Datasets of Inflamed Joints in ⁹⁹ᵐTc-Maraciclatide Imaging of Arthritic Patients by Natural Image and Diffusion Model Augmentation” accepted for presentation at the International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI Society) later this year. Maraciclatide is in development to diagnose and detect inflammation in patients with inflammatory arthritis and endometriosis. The development of AI tools could enhance the potential of maraciclatide as a new imaging marker. See more details here: https://lnkd.in/eUpdj33T Maraciclatide is for investigational use only and is not approved by the FDA or UK and European regulatory authorities. #molecularimaging #medicalimaging #AI #machinelearning #rheumatoidarthritis #endometriosis #nuclearmedicine #precisionmedicine #MICCAI2024

ISBI 2024 coming up soon (International Symposium on Biomedical Imaging 2024). #medicalimaging #imageprocessing #imageanalysis #machinelearning

Further research on the development of AI tools to enhance the potential of ⁹⁹ᵐTc-maraciclatide as a new imaging marker has been presented today by Robert Cobb from the School of Biomedical Engineering & Imaging Sciences, King’s College London (KCL). A poster at the annual conference of the Medical Image Computing and Computer Assisted Intervention Society (MICCAI Society) illustrates the latest methodology used in the ongoing research collaboration led by Professor Andrew Reader and Professor Gary Cook at KCL to develop AI tools to help clinicians read and interpret scans using ⁹⁹ᵐTc-maraciclatide. For more details, please see the press release here: https://lnkd.in/eQMJyu4s And high res poster here: https://lnkd.in/evi8UQux ⁹⁹ᵐTc-maraciclatide is for investigational use only and is not approved by the FDA or UK and European regulatory authorities. #molecularimaging #medicalimaging #AI #machinelearning #rheumatoidarthritis #endometriosis #nuclearmedicine #precisionmedicine #MICCAI2024 #inflammatoryarthritis

Last year, over 1 million people in the U.S. suffered from acute coronary events like heart attacks. Due to its demonstrated effectiveness, IV-OCT (intravascular optical coherence tomography) is a type of medical imaging that is increasingly used to guide interventions for acute coronary events. Even so, adoption of IV-OCT is lagging due to the time-consuming process of interpreting the images. Mohammad Haft-Javaherian, Martin Villiger, Kenichiro Otsuka, Joost Daemen, Peter Libby, Polina Golland, and Brett Bouma propose a deep learning model that can automate the process of simplified image interpretations, which was recently published in Optica's Biomedical Optics Express. Paper: https://lnkd.in/ekhCcFam #AmericanHeartMonth #aiforgood #aiforhealth #machinelearningsolutions #computervision

VASCage imparts STEM knowledge in schools together with klasse!forschung. Here Nadja Gruber presents to students of Reithmann Gymnasium in Innsbruck how artificial intelligence helps radiologists to better detect very small structures on medical images and diagnose diseases such as #stroke. AI for automated segmentation of 3D tissue images is an important focus of VASCage research. #mintbildung #imageanalysis #imageprocessing #imagesegmentation

I'm thrilled to share my latest project, "GANs in Medical Imaging: Synthesizing Realistic Images for Analysis." This innovative endeavor harnesses the power of Generative Adversarial Networks (GANs) to revolutionize medical imaging. Author's:- Mrunmay Mete Krutika Tathe Krunal Jane Parul bhusari Under the guidance of Dr. Smita Nirkhi Singh 🔍 Project Highlights: > Advanced GAN Models: Exploring cutting-edge GAN architectures to generate high-quality, realistic medical images. > Data-Driven Insights: Utilizing diverse datasets from MRI, CT scans, and X-rays to train and optimize our models. > Quality Evaluation: Implementing robust metrics to ensure our synthesized images meet the highest standards of realism and utility. > Real-World Applications: Enhancing diagnostic tools, augmenting training data, and supporting rare disease research through synthetic imagery. Check it out here: https://lnkd.in/gj-X3vwb #GenerativeAdversarialNetworks #GenerativeAI #MachineLearning #DeepLearning #SyntheticData #TechInHealthcare #ArtificialIntelligence #HealthTech #Radiology #ImagingScience #MedicalImaging

Carnival is over (at least in Germany), spring is around the corner - and so is the BVM, the national conference for Medical Image Computing. Among many other scientists, our PhD student Jonas Bohn is participating. He presents his #MICCAI 2023 paper via poster: „RPTK: The Role of Feature Computation on Prediction Performance.“ The abstract highlights the growing importance of #radiomics in #medical #imaging for disease analysis and treatment prediction. It acknowledges challenges in standardization and manual pipeline construction. To address these issues, the Radiomics Processing Toolkit (RPTK) is proposed, integrating advanced feature extraction and selection components. Comparative analysis with existing frameworks demonstrates significantly improved performance, offering valuable insights for optimized radiomics analyses in clinical settings. The paper is available here: https://lnkd.in/eS4xCUjN.

📢 Groundbreaking Advances in Microscopy: Introducing APIC 🧬🔬 For centuries, the capabilities of microscopes were constrained by the limitations of optical lenses, forcing scientists to choose between high resolution and a small field of view or low resolution and a large field of view. In 2013, the introduction of Fourier Ptychographic Microscopy (FPM) revolutionized this field by combining conventional microscopy with advanced computer algorithms, enabling high-resolution images over large areas with affordable equipment. Fast forward to now, Caltech engineers have pushed the boundaries even further with the development of Angular Ptychographic Imaging with Closed-form method (APIC). APIC surpasses FPM by eliminating the iterative trial-and-error approach, instead solving a linear equation to correct optical aberrations. This breakthrough not only provides sharper, more accurate images but also speeds up the process by reducing the need for repeated refocusing. Under the guidance of Professor Changhuei Yang, the team has created a method that ensures true representation of sample details, enhancing applications in biomedical imaging, digital pathology, and drug screening. APIC’s ability to deliver uniformly in-focus, high-quality images makes it a pivotal tool for optimizing AI applications in medical diagnostics. This innovation is set to transform microscopy, ensuring researchers can achieve greater precision and efficiency in their work. Kudos to the Caltech team for this remarkable achievement! #Microscopy #Innovation #BiomedicalImaging #DigitalPathology #DrugScreening #AI #OpticalEngineering #ResearchAndDevelopment

📢 Groundbreaking Advances in Microscopy: Introducing APIC 🧬🔬 For centuries, the capabilities of microscopes were constrained by the limitations of optical lenses, forcing scientists to choose between high resolution and a small field of view or low resolution and a large field of view. In 2013, the introduction of Fourier Ptychographic Microscopy (FPM) revolutionized this field by combining conventional microscopy with advanced computer algorithms, enabling high-resolution images over large areas with affordable equipment. Fast forward to now, Caltech engineers have pushed the boundaries even further with the development of Angular Ptychographic Imaging with Closed-form method (APIC). APIC surpasses FPM by eliminating the iterative trial-and-error approach, instead solving a linear equation to correct optical aberrations. This breakthrough not only provides sharper, more accurate images but also speeds up the process by reducing the need for repeated refocusing. Under the guidance of Professor Changhuei Yang, the team has created a method that ensures true representation of sample details, enhancing applications in biomedical imaging, digital pathology, and drug screening. APIC’s ability to deliver uniformly in-focus, high-quality images makes it a pivotal tool for optimizing AI applications in medical diagnostics. This innovation is set to transform microscopy, ensuring researchers can achieve greater precision and efficiency in their work. Kudos to the Caltech team for this remarkable achievement! #Microscopy #Innovation #BiomedicalImaging #DigitalPathology #DrugScreening #AI #OpticalEngineering #ResearchAndDevelopment