Medical Imaging to Benefit from Deep Learning

Share on twitter
Share on linkedin
Share on facebook
Share on reddit
Share on pinterest

Thanks to the advent of electronic medical records (EMRs), arguably one of the biggest challenges faced by clinicians and the medical establishment today is the ever-growing stream of patient data and the resulting overload of such information. Take for instance medical image data. Each patient’s image collection has an average of 250GB of data meaning that a staggering amount of data is created by medical imaging running into petabytes. In fact, after genomics, medical imaging is the fastest growing data sector in healthcare with an estimated 400 petabytes annually produced worldwide.

However, having more data is a good problem to have provided one can easily decipher and mine it. Detecting patterns of interest in patients’ medical images is a difficult and time-consuming process that is challenging for clinicians, even the best among them. While computer-aided analysis is certainly a great help, the surfeit of medical imaging data obtained from a panoply of imaging systems such as PET, CT, MRI, ultrasound, etc., each requiring significant clinical expertise, makes the task challenging. Also not to be overlooked is the fact that only a small fraction of this mountain of data is analyzed, a much smaller quantum of it using image processing algorithms.

The problem is more severely felt in those countries that have a chronic shortage of radiologists and fewer medical resources. In such geographies, the use of AI can offer remarkable benefits. A recent study done by Lakhani P. and Sundaram B. in 2017 has shown that lung tuberculosis can be detected on chest x-rays with 97% sensitivity and 100% specificity, if the images are analyzed by two different deep artificial neural networks (ANN), and only those cases in which the algorithms do not concur are re-evaluated by a radiologist.

It is clear by now that these new methods of machine learning, based particularly on “deep learning”, which is based on ANN, are a powerful tool. In totality, AI algorithms can help in handling medical imaging in several ways:

  • speed up clinical workflows to get productivity gains,
  • prevent diagnostic errors to obtain outcome-oriented clinical decision-making, and
  • bring about financial gains for medical establishments

In this scenario, the promise that artificial intelligence (AI) holds to transform the diagnostic imaging industry is starting to come to fruition. AI can bring in gains in productivity and diagnostic accuracy while also enabling more personalized treatment planning for patients. The accelerating pace of advances in deep learning technology is helping make AI-based tools more accurate and sophisticated with added functionalities. Among the reasons that are driving these advances are the access to vast quantities of patient data and images that these AI algorithms require to learn from and massive computing power.

One independent estimate forecasts that the global market for AI in medical imaging (comprising software for automated detection, quantification, decision support and diagnosis) will grow to as much as $2 billion by 2023.

In fact, many vendors such as IBM/Merge, Philips, Agfa and Siemens are beginning to integrate AI into their medical imaging software while several startups are also demonstrating the use of AI to sift through big data or offer immediate clinical decision support. has also been working on developing its proprietary deep learning algorithms that can be used on medical image data.

Related Post