4Quant

Big Image Analytics

Our vision is to fill the gap in the data backbone of medical AI business.

4Quant

What we do

We realize our vision by utilizing the latest Big Data and Deep Learning technology combined with medical expertise and user experience. This mix allows us to accelerate the extraction of high-quality, meaningful medical labels. We are involved in the entire cycle of building AI-powered medical tools. We provide a platform, training data, robust quality control and detailed analysis of possible issues and model weaknesses.

We are experienced in a wide range of PET, CT, and MRI modalities on a wide variety of conditions and diseases from bone fracture to neuroimaging.
We have powerful, scalable in-house developed software for processing and extracting quantitative information in a streamlined fashion for even very large data sets. The software is automatically tested and verified to ensure the highest level of reliability for the results. We have the latest infrastructure available for storing and interactively analyzing thousands of datasets with millions of images. From estimating risk to classifying disease, our tools provide the standard for scalable analysis in medical imaging.
We always deliver as a final product a detailed report distilling a series of complicated analyses into clear, legible figures and tables with thorough explanations of every step taken so the results can be reliably, automatically reproduced.

Our Products

HICCAP

The HICCAP Annotation platform is designed to enable for quickly and continuously creating curated data in the clinical setting. The platform is built on a number of core components enabling us to integrate with existing systems, provide pixel-exact annotations with single clicks, incorporate medical information into workflow and scale to thousands of annotators covering millions of patients. Look at our whitepaper to learn more.

IQAE

Big Image Data Analytics Platform (BIDAP) is our core platform for bringing the latest, scalable image analysis tools to end-users everywhere. The platform provides the ability to apply large complex operations across petabytes of data using simple commands. Beyond these commands, our tools make images search-able based on their content rather than meta-data in a real-time ad-hoc manner. Concretely this means that instead of manually going through archives of images processing them one at a time, you can search for images by visible body-part, or X-rays with tumors larger than 5 cm, or patients whose radiographs have changed by more than 20% in a week. Behind the scenes, we leverage our cloud-based technology stack to robustly process these requests with stunning speed and efficiency.

CloudImageProcessing

Leveraging the large existing communities built around ImageJ and FIJI, Cloud Image Processing (CIP) lets you scale up these tools from one machine to hundreds in a distributed, fault-tolerant manner. Whether with powerful workstations, high-performance clusters, or cloud-based machines, our framework can handle your problem and make analysis and configuration easy using any device with our web-based tools.

Consulting

With over 6 years experience in the large-scale image processing area we have experience with a large number of problems and techniques. Our experience ranges from biology to material science , from earth science and food engineering. Specifically we can help at any step of the analysis from experimental design to making sense of the results. Just contact us at info@4quant.com.

Training

Image analysis is a diverse, complex field rife with caveats and constantly changing. We offer a number of training solutions ranging from theory and basic principals to specific tools and applications. These can be further personalized to match your exact needs and competences. Furthermore much of our material is available for free online through the course we developed at the Swiss Federal Institute of Technology entitled "Quantitative Big Imaging".

Our Technology

We have developed our platform BIDAP, on top of the latest Big Data framework, Apache Spark™. Spark provides a scalable, fault-tolerant, distributed backend for robustly handling large (giga-, tera-, and petabyte) datasets. Additionally we have cloud computing integration built-in meaning you can have a cluster in the cloud up and running in minutes, not days.

4QL

Rather than requiring complicated development, compiling and testing, our tools enable image analysis on thousands of images to be conducted interactively using a simple query language 4QL. Our extensions of the SQL language which we will reference as 4QL is fully SQL-99 compliant and offers all of our imaging functionality without requiring any developer experience.

Apache Spark

Spark is currently the world record holder for the fastest public-cloud sorting of a petabyte of data. It is one of the most active Apache projects with over 50 companies and 255 contributers and has even surpassed standard data processing frameworks such as NumPy, matplotlib, and SciKit-Learn.

TensorFlow

Our platform is built from the ground up to support the most advanced Machine Learning and Deep Learning tools. TensorFlow is one of the most actively developed, flexible, and powerful machine learning platforms available. From object recognition to segmentation and feature extraction, it is unparalleled in the image analysis space and with our tools can easily scale to large image and video datasets. Our tools support the training, deployment and transfer-learning on models as well as combining the best parts of existing tools to solve complicated problems. We additionally provide higher-level access through our integration with libraries like Keras and sklearn to make building and validating models easy.

IBM Bluemix

Our software is intrinsically distributed and compatible with many cluster and cloud-based solutions to enable on-demand processing of large datasets without large infrastructure investements. For storing the data, we have out of the box support for local, Hadoop HDFS, Amazon S3™, and OpenStack Swift allowing for scalable, fault-tolerant storage of large datasets without the bottlenecks or hassle of storing everything on external hard drives. Leveraging the cloud computing infrastructure of IBM Bluemix and SoftLayer means we can focus on providing the best image analysis solutions and allow IBM to seamlessly manage the computational resources whether in the cloud or on-premise in a fully flexible and scalable manner. Finally our interactive web-based interfaces and notebooks make running and controlling complex image processing tasks on a cloud or cluster easy with any modern web browser.

Media

4Quant has given presentations, taught lectures and won competitions at a number of world renowned conferences and workshops. We have selected posters and presentations available online.

T-stage in NSCLC patients can be predicted by radiomics features in FDG-PET/CT (ECR 2018)

Infusing clinical understanding in radiomic features (ECR 2018)

Fully automatically staging of lung cancer using deep neural networks Presentation (ECR 2018)

Simulated PET images derived from CT using adversarial learning on unstructured data (ECR 2018)

Why big image analytics is essential for imaging biobanks: an osteoporosis radiomics study (ECR 2017)

Machine Learning for Improved PET/CT Fusion Images (ECR 2017)

Keeping a human in the loop, interpretable explanations for finding crushed vertebrae (ECR 2017)

Real-time MRI image segmentation using parallel image computing and machine learning algorithms

About Us

  • IBM Partner
  • ETH Spin-off

4Quant Ltd. is Spin-off of the ETH Zurich and the Paul Scherrer Institut based on the PhD Thesis of Kevin Mader done in the X-Ray Microscopy Laboratory of Marco Stampanoni. We are funded by the Pioneer Fellowship of the ETH Zurich and a member of the ETH Innovation and Entrepreneurship Lab. We are built on a strong foundation of partnerships ranging from the IBM Global Entrepreneur Program and Databricks for Cloud Support to SLS Techno Trans AG for access to the latest generation of imaging techniqes. We are committed to Open Source and have many of our projects available on Github and we push many of our latest developments back to the core project so the entire community can benefit from our developments.

Contact

We are located inside the Technopark Zurich on the first floor of the Einstein Wing. We can be reached by email at info@4quant.com or followed on Twitter