×
×

Deep learning has been the buzz word in the last few years and it actually presents a lot of probabilities for future innovation. Every day we see fresh innovations in this field of computing, leading to its popularity and applicability to more and more realms of day to day applications. It seems that the field of artificial intelligence has terminated its fallow period of hibernation, and is undergoing a revolutionary uprising, thanks to this new technology. And the world is now a witness to a plethora of applications ranging from simple software which finds application in our daily life to large scale niche scientific applications catering the needs of multiple domains including medical, consumer electronics and automotive. While this technology is imminent in the medical imaging field, implementation of deep learning algorithms have already been done in imaging technologies of other industries. However, the research advancements are happening very quickly in healthcare as well.

Deep learning applications, which have evolved over the last few years, appear poised to take the world ahead of the present reality and bring in new possibilities in the medical field.

A Revolutionary Technology in Radiology

A few years down the lane, deep learning will be achieving several amazing things in different fields, which are beyond our present imagination. It could be anything from virtually trying glasses online to translation of spoken language in real time. We can rightly say that data learning is unleashing such futuristic applications that they seem to be taken straight out of a science fiction movie.

With such unlimited potential, it isn’t difficult to imagine a deep learning technology that will allow computers to evaluate medical images. With this kind of tasks, deep learning technology in medical imaging may turn out to be the most revolutionizing technology seen by radiology after the introduction of digital imaging. It has the potential to augment the current radiology practice by acting as a second reader to the radiologist and by complementing the expert skills of humans.

Vendors and researchers like QuEST GLOBAL who are progressing in this field today boldly recommend radiologists to embrace deep learning if they want to become stronger and also state that failure to follow this trend could make them obsolete down the lane.

What Results Can Deep Learning Actually Drive?

Deep learning algorithms, which are poised to be launched in the market, could choose and pull out features from the medical images and at the same time they can identify a disease, classify it, and measure the disease patterns with negligible input from the radiology experts. These algorithms are powerful enough to assist in discovering the disease features in medical images and provide assistance using decision support tools. These applications include image segmentation, registration, computer aided detection and diagnosis, automatic labelling and captioning, reading assistant and automatic dictation as well as integration to healthcare big data to achieve precision medicine.

Where do deep learning algorithms find applications in medical imaging? What are the results that this technology can actually drive? Read on to find out.

  • Deep learning algorithms, especially Convolutional Neural Networks (CNN), have quickly become the preferred methodology for the analysis of medical images.
  • With respect to image processing, this technology will not only help in choosing features and extracting them from the medical images, but also build new ones. This will result in depictions of imaging researches never seen earlier.
  • On the image interpretation side, the applications of this technology will assist in not only identifying, classifying, and measuring the patterns of the disease from the medical images, but also let measuring prognostic targets and creating actionable forecast models of the treatment schedule.
  • The role of radiologists in future will be improved and further integrated with the patient care team. Many repetitive and distracting tasks which the radiologists currently do, can be reduced. They will become consultants with clinical focus, such as cardiology, oncology etc. With this new role, radiologists can concentrate more on giving recommendations for post- or pre-imaging care decisions. They will also have a greater influence at enhancing a patient’s treatment results.
  • Mammographic screening is a highly promising application of deep learning in medical imaging. This technology could greatly enhance the effectiveness of breast screening to improve results and cut costs. Deep learning could do extremely well at the same type of pattern recognition and analysis that a radiology expert does.
  • Apart from breast screening, brain tumor segmentation and lung cancer screening are also some of the promising applications of this technology. Early researches in these fields using images from MRI, CT, Ultrasound and other modalities have given good results, with better accuracy and sensitivity than traditional machine learning approaches and plenty of efforts are being put in towards advancements. The resulting developments have the potential to benefit many patients. Digital Pathology is another area where deep learning can have a significant impact.

With all the advancements that are happening, the healthcare industry will benefit the most from this technology when compared to any other industry. Though debates have been on about the disruption to the role of radiologists with the advent of deep learning applications in radiology, the potential benefits seem to outweigh every hurdle that comes in the way. At the same time, careful consideration should be given to ethical, regulatory, and legal issues before using patient clinical image data for the development of deep learning networks. This can be achieved through collaborative discussions among radiologists, scientists, engineers together with law and ethics experts.

There is no doubt that deep learning has plenty of untapped potential for creating breakthroughs in clinical radiology, which is indeed promising. This in turn can help the radiologists to make accurate diagnosis and provide quality healthcare, thereby ushering in a new era in patient care.

Written by Sindhu Ramachandran S

on 30 Nov 2017

Ms. Sindhu Ramachandran S is a Principal Architect at QuEST Global. She is currently involved in algorithm implementation and optimization in the Medical Devices and Hi-Tech Domains. Her key area of interest is image processing and machine learning. She leads the deep learning initiative in QuEST Global and provides her expertise in project activities related to deep learning. She can be contacted at sindhu.ramachandran@quest-global.com