Object segmentation, the art of extracting pixel-wise masks of objects or regions of interest from images, lies at the heart of modern computer vision and machine learning. These pixel-level labels empower models to detect and precisely localize objects, making them an indispensable tool across various industries.
With Medical AI, accuracy and precision can be a matter of life and death, and medical image segmentation is a vital frontier. It includes assigning a label to each pixel in an image, thus labeling them to various classes.
Commonly, medical image segmentation entails segmenting regions/objects of medical interest such as organs, bones, tumors, etc. For most machine learning applications, segmentation needs human intervention, which implies labeling various pixels of an image and assigning them categories under human supervision. Then, we can use this data to train models to segment objects they have not encountered before.
In this blog, we will look at some applications of medical image segmentation and explore features of our radiology editor, which can speed up the annotation process significantly.
Real-world Use Cases of Medical Image Segmentation
The iMerit Radiology Annotation Product Suite is built on top of the iMerit Ango Hub platform, an end-to-end enterprise-grade technology platform designed to deliver data annotation tools for AI teams. The combination of iMerit’s technology and radiology expertise provides seamless execution from training data to regulatory benchmarking.
The Radiology Annotation Product Suite delivers automation, annotation tools, and analytics into a single platform that enables accurate data pipelines for rapidly scaling radiology AI solutions into production. It combines:
- Advanced Tooling: Designed to solve scaling challenges in digital radiology through automation
- Platform: A single tool for sourcing, staffing, annotating, and validating machine learning models
- Expertise: Combines medical experts with secure data management, workflow customization, and annotation automation
“The central vision of our suite is to overcome the scaling challenges in digital radiology by bringing together the human expertise required for accuracy with a platform for secure data management and automation into a single solution.” Radha Basu, Founder and CEO, iMerit
Faster and Efficient Annotation with AI-assistance
Covid-19 Detection and Localization
During the recent pandemic, COVID-19 detection using X-rays with deep learning surged and was highly researched. Using the segmentation of lungs with COVID-19 in X-rays, researchers and engineers could accurately predict positive and negative cases.
The following is an image from COVID-Net, an open-source deep learning initiative aiming to tackle the detection of COVID-19. The convolutional neural network returns segmented regions of interest.
Linda Wang and Alexander Wong’s paper is highly recommended for a deeper dive into COVID-Net.
Abdomen Organ Detection
Abdomen Segmentation from TransUNet
As the name suggests, this problem relies on segmenting various organs within the abdomen. It often tackles the modality of CT scans and segments organs across multiple slices.
Brain Tumor Segmentation
Brain Tumor Detection (Source)
The region of interest for this problem is the area affected by the tumor in the brain. Once this area is identified and segmented, further studies can be performed. It greatly alleviates the burden on radiologists.
Skin Lesion Masks (Source)
This is an application for dermatology where the goal is to segment the area of skin that contains a lesion. It can then be classified into benign and malignant.
Cell Nuclei Segmentation
Cell Nuclei Segmentation (Source)
This application belongs to the field of microscopy and can be used for lab-related applications. The goal is to segment out the region containing nuclei of various types of cells. This can assist in disease detection, cell counting, and various other medical use cases.
How to Perform Medical Image Segmentation
Polygons and Segmentation Masks
Labeling each pixel one by one is an incredibly tedious task, and for practical purposes, not one that humans perform to annotate (segment) objects in a medical image.
A more efficient way to segment objects is through the use of contour (edge) points. This is the so-called ‘polygonal annotation’. The annotator draws a polygon that encloses the object, and the pixels remaining inside this polygon belong to that specific class.
Another way to tackle pixel-wise annotation manually is by drawing segmentation masks similar to how one would paint using a brush. To segment the object of interest the annotator ‘paints’ over all the pixels that belong to an object.
Automated Interactors for Medical Image Segmentation
Medical objects can be incredibly complex and the process of manually clicking each edge of a polygon and painting over each pixel can be tedious. For this reason, we have launched the iMerit Radiology Editor, designed with interactors to speed up the process manifold.
The Magnetic Lasso tool, is automated way of segmenting an object. The way this tool works is by sticking to the edge of the salient object. Given a few anchor points, it forms the boundary of the object simply as the user hovers their mouse over the object. This also reduces the interaction of the annotator with the image considerably thus saving time.
Level Tracing allows you to autoselect a group of voxels based on the Hounsfield units of the selected voxel with adjustable tolerance.
This feature can be incredibly useful in medical image analysis, especially for tasks like isolating specific tissues or structures within a CT scan. For example, it might be used to automatically select and segment all the soft tissue within a certain range of Hounsfield Units while excluding bone or air. The adjustable tolerance allows radiologists or researchers to adapt the selection to their specific needs, improving the efficiency and accuracy of their analysis.
Multiplanar Translation & Crosshair
The crosshair allows you to gather your bearings in 3D space and to understand where your cursor is positioned relative to all views. It also allows you to synchronize all views in such a way that they all show the same pixel you are selecting with the crosshair. This is also known as multiplanar translation.
This isn’t just handy; it makes tasks like identifying things in medical images much easier. The crosshair and multiplanar translation help in both navigating in 3D space and simplifying image segmentation by making it more precise and straightforward.
By enabling the 3D toggle, you’ll be able to see a 3D reconstruction of the annotations completed so far in the top-right view. You can rotate the view by dragging with the mouse cursor, and zoom in and out with the scroll wheel.
Often, medical images come in the form of volumes. These volumes consist of multiple slices (i.e. multiple still frames), and often, these frames are temporally and spatially interrelated. Since labeling multiple slices is a tedious task as these slices can be in the hundreds per volume. Segmentation interpolation is another technique for speeding up annotation up to 40 times since multiple slices are automatically labeled.
Segmenting a CT Scan with Ango Hub
Medical image segmentation is at the core of AI and Computer Vision efforts in healthcare. The research and applications of segmentation are sure to revolutionize this field in the near future. However, the fuel to get to this state of the future is well-segmented (annotated) medical data.
To help teams achieve their turn-key medical AI projects we ensure that their training data needs are met in the most efficient and highest quality manner possible. For this, we utilize our array of AI assistance tools for medical data labeling, an extremely skilled and capable workforce, and state-of-the-art platform for labeling: Radiology Annotation Product Suite.
If you’d like to see how your company can get its data segmented to start or scale its cutting-edge medical AI project, get in touch with us to talk about how to solve your data labeling needs.