📢 Announcement! A new paper in the latest MLHS issue by Molchanov & Pal demonstrates a DenseNet201 framework achieving 96.69% accuracy in classifying retinal diseases (CNV, DME, Drusen) from OCT scans.
This work underscores the power of transfer learning and specialized loss functions in medical imaging. Congratulations to the authors on their contribution to the field!
Not every LLM is best for every task.
In ☸️SAIMSARA, one scientific review is built through a multi-model workflow:
One review. Four LLMs. One evidence-driven pipeline.
📢 Announcement! SAIMSARA receives ISSN 3054-3991 and becames an AI-native scientific journal!
Now operating in two layers:
Build. Read. Ask the evidence with ☸️SAIMSARA
How it was built — and how it screens research papers for retraction risk.
For authors: pre-submission check. For editors: early screening layer. For reviewers: triage signal.
📢 Announcement! ML in Health Science launches a new screening tool that estimates how closely an abstract resembles cardiovascular-medicine papers retracted within the last ~5 years.
Retract is powered by a fine-tuned BERT text-classification model trained on 520 retracted research articles (2021–2026).
Upload your abstract prior to submission to assess its retraction probability and plus an AI explanation with clarity improvements.
Meet the latest SAIMSARA update:
🎥 Watch the story behind ☸️SAIMSARA 4.0
🌍 We invite clinicians, researchers, and everyone interested in digital health to explore MLHS apps. This is a growing hub of real-world, ready-to-use medical, research, and education utilities — running entirely in your web browser, all in one place.
💡 For developers:
📢Announcement! ML in Health Science’s new app for medical image augmentation + dataset cleanup.
Synthetica generates realistic acquisition-level variants (findings preserved) and reduces common overlays (text/UI/borders).
Supports:
☢︎ Radiology (XR, DSA/Angio, MRI, Ultrasound, CT*)
🔬 Microscopy (histology / pathology / microbiology)
📷 Macroscopy (endoscopy / dermatology / clinical images)
New paper in our journal.
Architecture and Workload as Primary Sources of Error in RAG and Agentic AI Systems
presents insights gained from over two years of ☸️SAIMSARA development.
The article summarizes practical lessons learned while building a multilayer, agent-based evidence extraction system capable of processing millions of scientific papers and thousands of references at scale.
To our knowledge, no stronger system is currently in operation. Readers aware of comparable or more powerful systems are invited to share references or insights in the comments.
📄 Not just URL/TXT anymore: export a clean, paper-style PDF you can read offline, print, and share with colleagues in one click.
The AI Act still cannot function effectively, because there are no real, practical comparators that can reliably distinguish black-box, low-quality “AI tools” from genuinely validated, transparent, clinically aligned systems. At ML in Health Science, we have been working on this challenge since 2023. Today, we introduce our latest solution — a fast, structured tool for individual assessment of healthcare AI quality and alignment with human-centered and regulatory expectations. The evaluation takes only a few minutes.
⚡ Result: 1000× faster than a human analyst.
⏱️ Stop wasting time searching for biomedical evidence.
☸️ Subscribe to SAIMSARA and take your research to the next level — faster, smarter, and backed by reliable data.
🌍 We invite clinicians, researchers, and digital health experts to contribute as editors and reviewers of innovative digital health project
💡 What we offer:
Medical images are machine outputs, not holiday photos. In MLHS, every image goes through the same 224×224 bilinear resize (no cropping) at both training and prediction. Augmentations are train-only and used thoughtfully to preserve anatomical truth. Join the discussion on X to learn more!
💡Digital Health Developers: Limited Clinical Background?
🎯 Until 01.01.26 — Apply for an Editorial Peer Review (FREE, limited to 3 projects!)
Your Bionic Hand is a Web AI application that allows users to train and calibrate our custom ESP32- and proximity sensor–driven bionic prosthesis without relying on external services. Its low-cost, low-power design makes it accessible to over 95 % (≈3 million) of upper-limb amputees worldwide. For more information, watch the attached video.
This Web AI application is trained on grayscale 2D ultrasound scans of the forearm, each corresponding to one of four distinct hand positions: Thumb, Forefinger, Open Palm, and Fist. Clicking a photo loads the related ultrasound scan, and the AI model instantly predicts the virtual bionic hand’s movement in real time.
Build custom Web AI models using live video or camera input for real-time computer vision computer vision utilizing transfer learning with the MobileNetV2 layer model and the TensorFlow.js library. The model can be stored after training for deployment, sharing, and retraining with new data. All calculations are performed on the client side and are not transferred elsewhere, ensuring data privacy and safety.
Systematic, AI-powered, Medical Scientific Article Review Agent
🔹 Take your first step toward crafting evidence-based meta-analyses, systematic reviews, and new research directions in health sciences.
🔹 Let AI filter and extract relevant research based on your input, providing new directions to build upon.
DApp designed for easy participation in MLHS ecosystem.
🔹 BNB Network ensures fast transactions
🔹 Reown AppKit enables smooth wallet connection
🔹 1inch Aggregator provides fair and optimized fees
Hybrid AI built on a Tensorflow.js and empowered by Gemini and Deepseek. 🔍 What does it do?
🔹 Individual risk estimation for developing varicose veins
🔹 Supports patients and healthcare professionals with the latest scientific knowledge on diagnosis, treatment, and prevention of the disease
Science 3.0 is a DApp and the utility token of the Web3 Journal: ML in Health Science, designed to preserve published articles and valuable health science research data on the opBNB blockchain using the SCI3 Token . 🔑 Use your author or reviewer login credentials for the 📔Web3 Journal: ML in Health Science to access the system for free.
This web application is a poll associated with the corresponding editorial, "Artificial Intelligence in Academic Writing: Time for Science 3.0," published in Web3: Journal of ML in Health Science. Participate in the poll and share your thoughts in the article's comments section.
Computer vision utilizes numerical color values in the RGB scale for image preprocessing, resulting in 255 x 255 x 255 possible color variants for each pixel. This complexity can make the analysis of high-resolution images challenging. Dividing images into smaller patches helps balance computational resources. This tool demonstrates how visual content is transformed into numerical values for each patch and the entire image.
This experimental tool uses Markov Chains and TensorFlow.js Universal Sentence Encoder to generate new text on the client side based on input information, without relying on large language models.
The computer vision model provides the ability to predict the probability of lower limb salvage or loss in peripheral artery disease patients using pedal angiograms (DSA) in sagittal view.
Open-source AI chat powered by the Gemma2b. It runs on your device, ensuring full privacy - crucial in healthcare settings!
🔹 Tasks: Text generation, context extraction, and conversation, including chat with PDF documents.
⚠️CAVE: Mobile devices are currently not supported! During the first start, Gemma2b will be loaded, please be patient. Subsequent starts will be much faster.
Learn the principles of machine learning through gameplay. The Snake is guided by a reinforcement learning using neural network predictions, with training data generated from auto gameplay. Experiment with parameters and observe the performance. If you would like to see how the ML works, right-click, select "Inspect," open the Console of your desktop browser, and enjoy the heartbeat of the machine.
🗃️️️ In the list of deployed WebAI applications you can find model trained and deployed in the Web AI Playground. Each model has an individual link to use and share. This list is moderated. If you find an invalid model, please contact us.
Create your own web applications using the AI model trained in the Playground. Simply upload the model, add a cover image, and press the "Submit" button. After submission, you can add a description, your custom data, and your Web3 wallet if you wish to receive donations and monetize your work.
Beta-Test. Control the position of the circle indicator with your voice.
Build custom machine learning regression and classification web applications from numerical data using feedforward neural networks powered by TensorFlow.js. After training, the model can be saved for deployment, sharing, or retraining with new data. All computations are performed locally on the user’s device, ensuring full data privacy and security.
Build custom computer vision web applications from images utilizing transfer learning with the MobileNetV2 layer model and the TensorFlow.js library. The model can be stored after training for deployment, sharing, and retraining with new data. All calculations are performed on the user side and are not transferred elsewhere, ensuring data privacy and safety.
This web app overlays GLOBOCAN cancer incidence data with NASA's Earth emissivity maps to explore the potential association between chronic infrared exposure from Earth's emissivity and overall cancer incidence rates. For more information, please refer to the accompanying article.
This Web AI app is a computer vision tool for classifying femoropopliteal angiograms based on the FP GLASS classification, as detailed in the accompanying article. For comprehensive guidance, please refer to the research article.
Opinion Piece published in our Journal on the human-centered “Xavatar” project, which aims to offer unique and engaging experiences by creating a deeper sense of presence and interaction for patients with mobility disorders. Read the paper, try the editorial application "Immersive Content", and share your comments on the use of extended reality (XR) in medical education.
Opinion Piece published in our Journal on the human-centered “AI_4_Healthcare” project, which explores proven approaches to utilizing AI in healthcare. Read the paper, try the editorial application "Keyword Generator", and share your thoughts on using proper search strategies and systematic literature reviews to achieve accurate and reliable information extraction in healthcare.
This web application is a poll associated with the corresponding article, "Confounders in Predictive Medical Models: The Role of Religion," published in Web3 Journal: of ML in Health Science. Participate in the poll and share your thoughts in the article's comments section.
This web application is a poll associated with the corresponding article, "Confounders in Predictive Medical Models: Roles of Nationality and Immigrant Status," published in Web3 Journal: of ML in Health Science. Participate in the poll and share your thoughts in the article's comments section.
This web app calculates the complexity score of CME online courses to evaluate their resistance to AI agents. For more detailed information, please refer to the accompanying article.
This web application is a poll associated with the corresponding article, "Do You Consent to the Use of Your Biological Data for Training ML and AI Models? Online Survey Targeting Clinicians and Researchers" published in Web3 Journal: of ML in Health Science. Participate in the poll and share your thoughts in the article's comments section.
This web application is a poll associated with the corresponding article, "Do You Agree with AI Making Decisions About Your Treatment? A Comparative Survey of IT and Healthcare Practitioners," published in Web3 Journal: of ML in Health Science. Participate in the poll and share your thoughts in the article's comments section.
💡 On December 30, 2023, we launched the ML in Health Science: 🔗 Blockchain verification.
Our mission is to bring clinicians and digital health developers into one community and to support human-centered, patient-oriented science in the age of AI and machine learning.
📖 Why ML in Health Science — the philosophical introduction and first publication in our journal:
https://doi.org/10.62487/e4ccm968
✍️ Write the new chapter of health science with us!
