Important announcement: Missinglink has shut down. Click here to learn more.

Deep Learning in Healthcare Cover

Deep Learning in Healthcare

AI in Medical Imaging

Doctors have been using medical imaging techniques to diagnose diseases like cancer for many years. However, Artificial Intelligence (AI) has the potential to take this technology further and to improve medical imaging capabilities such as higher automation and increased productivity.


AI can improve medical imaging processes like image analysis and help with patient diagnosis. With many applied AI solutions and many more AI applications showing promising scientific test results, the market for AI in medical imaging is forecast to grow exponentially over the next few years.


The FDA needs a new, flexible, regulatory approach that covers the total lifecycle of a product. This approach should factor in the frequent modifications made to AI-based medical software. The FDA will likely need to rely on tools like MissingLink to scale up their platforms.

What Is Medical Imaging?

Medical imaging uses different processes and imaging methods to represent an internal image of the human body for diagnostic and treatment purposes. Medical imaging is often used in treatment and follow-ups for diagnosed diseases.

The term, medical imaging, includes various radiological imaging techniques such as X-ray radiography, Magnetic Resonance Imaging (MRI), medical ultrasonography or ultrasound, Computed Tomography (CT), and nuclear medicine functional imaging techniques like Positron Emission Tomography (PET).

The need for artificial intelligence in medical imaging

AI can improve traditional medical imaging methods like Computed Tomography (CT), Magnetic Resonance Imaging (MRI), and X-ray by offering computational capabilities that process images with greater speed and accuracy, at scale.


AI has the potential to improve medical imaging with:

  • Higher automation—AI can automate parts of the radiology workflow.
  • Increased productivity—AI has better computational capabilities than humans, so it can analyze medical images faster than medical doctors.
  • Standardized processes—AI can supply doctors with AI tools to compute big data and help doctors and encourage them to work smarter and more efficiently.
  • More accurate diagnosis—studies show that AI can be more efficient than doctors and experts at diagnosing many diseases like cancer from medical images. For example, scientists at Google have created an AI that diagnoses breast cancer. The AI is fed with slides of medical images and uses DL algorithms to diagnose cancerous cells. The AI recorded a 99% accuracy in cancer diagnosis based on these slides compared to 38% of some doctors in the comparison group.
  • Computing quantitative data—AI has the ability to use quantitative data in ways that are beyond the limits of human cognition. For example, AI can predict if a patient will suffer from heart failure based on their medical history and rate of hospital visits.
  • Assistance for doctors—AI can compute a large amount of data, map it and represent the relevant parts in a brief and efficient format that doctors can use.

AI in Medical Imaging Applications

With the potential to improve and standardize the process AI can be applied in medical imaging for various medical tasks. However, AI is intended to be used in conjunction with human insight.


Medical applications for AI include:

  • Medical image analysis—this technology can identify anomalies and diseases based on medical images better than doctors.
  • Aids in the diagnosis of neurological conditions—AI can help doctors diagnose neurological diseases like Amyotrophic Lateral Sclerosis(ALS). A study has also shown that AI was able to predict Alzheimer’s disease years before it manifests.
  • Revealing cardiovascular abnormalities—AI can measure a patient’s heart structure and indicate their risk of cardiovascular disease or other problems that might require surgery. Automated AI can be used to detect abnormalities in common medical tests like chest X-ray and lead to quicker risk detection and less misdiagnosis.
  • Cancer screening—early cancer diagnosis often results in a better outcome for patients. Recently, scientists created an AI based on Convolutional Neural Networks (CNN), a type of Artificial Neural Network (ANN) used to identify various types of cancer (1,2,3) with a high success rate. These experiments show that AI can decrease detection times and improve the rate of diagnosis.

The AI Medical Imaging Market

AI is a driving factor behind market growth in the medical imaging field. Signify Research published a forecast that claims that AI in medical imaging will become a $2 billion industry by 2023.

World Market for AI-Based Medical Image Analysis Software by Algorithm Type

Source: Signify Research


Some possible applications for AI in medical imaging are already applied in general healthcare:


  • GE medical imaging—in a collaboration with NVIDIA, GE healthcare has 500,000 imaging devices in use worldwide. These devices use AI to speed up the process of analyzing CT scans with improved accuracy.
  • Siemens medical imaging—AI Rad Companion Chest CT is a software assistant that uses AI for CT. The software measures and identifies orangs and lesions in thorax CT scans and automatically generates a quantitative report to help increase efficiency and correct diagnosis in radiology.
  • Aidoc—an AI for analyzing radiology scans. Aidoc is installed in over 100 hospitals throughout the world. It has helped doctors diagnose over a million patients. After Aidoc initially struggled with resource management, which caused inflated GPU cloud service costs, they turned to MissingLink and with MissingLink’s automated platform, they managed to scale up. The platform let Aidoc’s team automate and control their deep learning lifecycle, their core cloud infrastructure, and their experiment results. Furthermore, Aidoc’s AI team can use MissingLink to view and control their experiments, aggregated from multiple cloud servers, via a single web page. The team can also queue and prioritize experiments while MissingLink automatically manages deployment. Read the full story on how MissingLink helped Aidoc.

How Is The FDA Preparing For The Future of Medical Imaging?

The U.S Food and Drug Administration (FDA) is taking steps to prepare for the regulation of AI/ML-based medical imaging tools. In an article published by the FDA, they explain that the traditional model they used for medical device regulation was not designed to deal with the dynamic nature of AI and ML technologies. Thus, the FDA proposed a new regulatory framework.


The FDA lists 4 key points it will use to regulate AI/ML-based Software as a Medical Device (SaMD):


  1. Establish what are quality systems and good ML practices.
  2. Review SaMD premarket submission to demonstrate assurance of effectiveness and safety measures. Establish clear expectations for manufacturers of how AI/ML-based SaMD will continually manage patient risks throughout the lifecycle.
  3. Expect manufacturers to incorporate a risk management approach and monitor their AI/ML devices.
  4. Increase user and FDA transparency to use in real-world performance reporting and maintain continued assurance of safety and effectiveness.


A total product lifecycle is a new and different regulation approach for the FDA, it requires different procedures and, most importantly, different resources. To accommodate this new approach, the FDA will rely on tools like MissingLink to audit AI-based medical models and algorithms and to establish the regulatory framework for modifications to AI/ML-based SaMDs.

Read the full article on how the FDA is regulating AI in healthcare.

Manage Your Experiments With MissingLink

Using AI in medical imaging involves intensive tasks like training it to detect lesions and other signs of cancer in thousands of scans. Running AI experiments in medical imaging demand powerful hardware and might prove challenging if you need to manage multiple experiments simultaneously.


MissingLink organizes and tracks all of your code, data, and experiments within a single visual dashboard that all team members can access to improve collaboration. Team members can use this easy to use “single source of truth” to evaluate how data changes impact model performance between experiments.


MissingLink provides a platform to easily manage multiple experiments and trace and safeguard your data. MissingLink standardizes your servers and provides you with support for tools that facilitate shared computational environment. With MissingLink, you and your teams can manage all your machines and environments in a standardized computational environment.


With MissingLink’s version control feature, you can track all the specific dataset changes that individual contributors make. MissingLink locks each data volume version to protect the integrity of your data━you can commit a new version if you want to make changes.


With MissingLink, you can schedule, automate and record your experiments. This allows you to track all your experiments, code, machines and results.

missinglink screenshot

Learn more about how easy it is to use deep learning in healthcare with MissingLink.

Train Deep Learning Models 20X Faster

Let us show you how you can:

  • Run experiments across hundreds of machines
  • Easily collaborate with your team on experiments
  • Reproduce experiments with one click
  • Save time and immediately understand what works and what doesn’t

MissingLink is the most comprehensive deep learning platform to manage experiments, data, and resources more frequently, at scale and with greater confidence.

Request your personal demo to start training models faster

    Thank you!
    We will be in touch with more information in one business day.
    In the meantime, why not check out how Nanit is using MissingLink to streamline deep learning training and accelerate time to Market.