Resolve Stroke’s Post

The last paper of my PhD got finally published, this was an incredible team effort involving a large number of people and many years of work (this is why you will find a VGG doing the lifting here 😉) - my thanks to all of them. In a nutshell this work enabled neural circuit understanding at brain scale, as without it the sign of a synaptic connection would be unknown for a large fraction of the fly - thus circuit function would be hard to determine. In addition, we were able to extract what exactly the network learned, revealing so far unknown ultrastructural features: This is a case where AI (NNs but this is LinkedIn) taught us something we did not know before, i.e. a form of semi automated knowledge discovery.

Part 1 Hi Fellows! One of the main ways to understand the flow and operation of a neural network is to analyze the format of the input data. Understanding the inputs well not only helps you choose the best architecture but also helps you better understand what the machine will try to learn in the process. Examples of Input Formats 1. Convolutional Neural Networks (CNN) Multi-Modality Medical Images In medical applications such as image diagnosis, it's common to deal with images from different modalities such as Magnetic Resonance Imaging (MRI), Computed Tomography (CT), and Ultrasound. The input format would be (width, height, channels), where the channels represent the different image modalities. Video For video analysis, such as in camera surveillance systems, the input format would be (number of frames, width, height, channels). Each frame is treated as an individual image. ---

Part 2... The use of #AI in cardiology can refer to computational techniques that enable machines to learn from vast amounts of data, such as imaging, clinical parameters, and patient histories, to make decisions that traditionally require human expertise. In cardiology, AI often involves: 1️⃣ #MachineLearning (ML): Algorithms that can identify patterns in data to predict clinical outcomes or suggest diagnoses. 2️⃣ #DeepLearning (DL): A more complex subset of ML that simulates the neural networks of the human brain, allowing for advanced image and signal analysis. Wouldn't it be great if we could see patients alongside an "intelligent" AI-driven #EMR that could guide us into intervening earlier and preventing disease or adverse outcomes?

🚀 Day 69: Today, I focused on Brain Tumor Classification, utilizing advanced machine learning and deep learning techniques to accurately classify brain tumor images. By leveraging convolutional neural networks (CNNs), transfer learning, and image processing algorithms, I aim to develop a robust classification system capable of assisting medical professionals in diagnosing brain tumors more efficiently and accurately. Excited to contribute to the advancement of healthcare technology and improve patient outcomes through AI-driven medical diagnostics! #BrainTumorClassification #MedicalAI #DeepLearning 🧠💻🏥

New Post: Deep learning reconstruction for coronary CT angiography in patients with origin anomaly, stent or bypass graft

🚀 Achieving 99% Accuracy in Brain Tumor Classification using Transfer Learning and CNNs! 🧠 I’m thrilled to share the successful completion of my latest project, where I combined the power of Convolutional Neural Networks (CNNs) with Transfer Learning to classify brain tumor MRI images. The model achieved an impressive 99% accuracy! 🔍 Project Overview: Pretrained Model: EfficientNetB3 for feature extraction, fine-tuning for brain tumor classification. Custom Architecture: Added custom dense layers to improve classification accuracy. Data Preprocessing: Employed data augmentation for better model generalization. Optimization: Used Adamax optimizer and EarlyStopping to prevent overfitting. Evaluation: Achieved 99% accuracy on the test set, highlighting the strength of transfer learning in medical imaging. 🔗 Next Steps: Excited to further explore AI’s potential in healthcare to bring impactful, AI-driven solutions. #MachineLearning #DeepLearning #AI #Healthcare #MedicalImaging #TransferLearning #BrainTumorClassification

Hello connections in this post we will talk about the "Impact of Convolutional Neural Networks in Medicine" Convolutional Neural Networks (CNNs) have emerged as a groundbreaking technology revolutionizing the field of medicine. CNNs, a specialized form of deep learning, excel in processing and analyzing visual data, making them invaluable in medical imaging and diagnostics. In healthcare, CNNs are being used to interpret complex medical images such as MRIs, CT scans, and X-rays with remarkable accuracy. By detecting patterns and anomalies that may elude the human eye, CNNs aid in early disease diagnosis, treatment planning, and monitoring patient progress. One of the significant applications of CNNs in medicine is in the field of pathology. By training on vast datasets of histopathology slides, CNNs can identify cancerous tissues with high precision, leading to quicker and more accurate diagnoses. This not only saves time but also enhances patient outcomes by enabling timely intervention. Moreover, CNNs have also been instrumental in drug discovery by predicting the efficacy and potential side effects of new compounds, thus accelerating the development of novel treatments. As CNNs continue to evolve and adapt, their role in revolutionizing healthcare is set to expand further, offering new possibilities for personalized medicine, improved patient care, and enhanced medical research. The integration of CNNs in medicine signifies a transformative shift towards more efficient, accurate, and innovative healthcare practices. #snsinstitutions #snsdesignthinkers #designthinking

📃Scientific paper: U-Mamba: Enhancing Long-range Dependency for Biomedical Image Segmentation Abstract: Convolutional Neural Networks (CNNs) and Transformers have been the most popular architectures for biomedical image segmentation, but both of them have limited ability to handle long-range dependencies because of inherent locality or computational complexity. To address this challenge, we introduce U-Mamba, a general-purpose network for biomedical image segmentation. Inspired by the State Space Sequence Models (SSMs), a new family of deep sequence models known for their strong capability in handling long sequences, we design a hybrid CNN-SSM block that integrates the local feature extraction power of convolutional layers with the abilities of SSMs for capturing the long-range dependency. Moreover, U-Mamba enjoys a self-configuring mechanism, allowing it to automatically adapt to various datasets without manual intervention. We conduct extensive experiments on four diverse tasks, including the 3D abdominal organ segmentation in CT and MR images, instrument segmentation in endoscopy images, and cell segmentation in microscopy images. The results reveal that U-Mamba outperforms state-of-the-art CNN-based and Transformer-based segmentation networks across all tasks. This opens new avenues for efficient long-range dependency modeling in biomedical image analysis. The code, models, and data are publicly available at https://lnkd.in/eVewBv-6. Continued on ES/IODE ➡️ https://etcse.fr/eU8 ------- If you find this interesting, feel free to follow, comment and share. We need your help to enhance our visibility, so that our platform continues to serve you.

researchers from Google and Harvard University have recently released a new set of detailed images of the human brain and its neural networks using AI and 3D mapping. they pieced the image data together and generated it into a 3D structure detailing each cell. by combining brain imaging with AI-based image processing and analysis, google and harvard university researchers managed to reconstruct nearly every cell and all of its connections within a small volume of the donated human brain tissue, which was about half the size of a grain of rice, and produce images that could help researchers understand neurological disorders and learn more about how the brain works. read more here: https://buff.ly/4but6jP

🔹️ Brain Tumor Detection using Deep Learning In this project, I developed a convolutional neural network (CNN) to classify MRI images into four categories: Glioma, Meningioma, No Tumor, and Pituitary. The project involved organizing the dataset and preprocessing the images by cropping the region of interest based on contour detection. The images were then resized and normalized to improve model performance. The dataset was divided into training, validation, and testing sets. Data augmentation techniques were applied to the training set to enhance the model's generalization capabilities. The CNN architecture was carefully designed, incorporating dropout layers to prevent overfitting. The model was trained on the training set and validated on the validation set, using learning rate scheduling for optimization. After training, the model's performance was evaluated using a confusion matrix and learning curves to ensure accuracy and to detect any signs of overfitting. The trained model was saved for future use, and an inference pipeline was developed to process new MRI images and predict their categories. This project demonstrates my proficiency in deep learning workflows, including data preprocessing, neural network design, performance optimization, and real-world deployment, resulting in a robust MRI image classification model. ✳️ GitHub Repository: https://lnkd.in/d5iX8Phk #MachineLearning #DeepLearning #CNN