Deep learning and medical imaging: Training machines to scan images for diseases

• Author:
• Author name

  Quantumrun Foresight
• October 28, 2022

Insight summary

Deep learning (DL), a branch of artificial intelligence (AI), is transforming healthcare by automating tasks and enhancing diagnoses through advanced pattern recognition. This technology, particularly in medical imaging, offers faster and potentially more accurate diagnostics. While DL's integration into healthcare shows promise, it still requires validation to ensure its effectiveness and reliability compared to traditional methods.

Deep learning and medical imaging context

Deep learning is increasingly being applied to automate manual tasks across the healthcare sector, from organizing electronic health records (EHR) to scanning disease databases. In particular, DL is making significant strides in improving diagnoses by identifying anomalies in medical scans and optimizing radiology assessments.

Deep learning enables more accurately descriptive, predictive, and prescriptive analyses than manual methods. For example, convolutional neural networks (CNNs) are a DL architecture initially designed to study images. These networks differ from traditional machine learning techniques because they can learn pattern recognition from raw data rather than requiring human programming to organize the information. Medical imaging may be regarded as the most promising area of all methods through which DL may be utilized in healthcare. 

Radiological examinations, such as x-rays, ultrasounds, and CT (computed tomography) scans, require human interpretation for a timely response. Alongside the increasing demand for radiologists, particularly in low-to-middle-income nations, there is a growing need for diagnosis automation. However, deep learning must achieve diagnostic accuracy comparable to that of healthcare professionals before being integrated into standard clinical practice. While specific DL-powered diagnostic software has received the US Food and Drug Administration (FDA) clearance to be used in clinical practice, many experts believe that the critical evaluation of these technologies is in its early stages.

Disruptive impact

Several companies are investing in incorporating AI in medical imaging. In 2022, AI-powered disease detection company Viz.ai raised $1.2 billion USD in Series D funding. The company focuses on improving healthcare services through intelligent software that reduces treatment time, improves care access, and increases medical innovation speed. The firm also announced its partnership with Hyperfine, which created the first FDA-approved magnetic resonance imaging (MRI) device called Swoop.

The partnership brings MRI to the patient’s bedside and delivers valuable insights to clinicians’ emails for timely decision-making. Swoop is the first MRI system that offers neuroimaging at the point of care. With this device, physicians can rapidly make clinical decisions for patients instead of making them wait for transport and images hours later. Additionally, using Swoop reduces potential adverse events linked with transporting sick patients.

Another use case of DL in medical imaging is in the organization and storage of volumes of clinical information. According to a study published in Nature, 87 percent of surveyed radiologists believe clinical data significantly impacts interpretation. Moreover, correct imaging interpretation isn’t limited to radiology but applies to pathology, ophthalmology, and dermatology.

With continuous digitization, more people require radiological imaging exams. As a result, many radiologists interpret images every 3-4 seconds over an 8-hour work day, leading to fatigue, burnout, and increased error rates. Convolutional neural networks are particularly good at recognizing and classifying images and are frequently used to organize medical imaging records.

Implications of deep learning and medical imaging

Wider implications of deep learning and medical imaging may include: 

• The US FDA approving more software designed to interpret medical images, leading to more startups focusing on this growing technology.


• Insurance providers designing insurance packages to include deep learning AI integration protection for healthcare providers. 


• Radiologists slowly transitioning to verifying diagnoses and treatment plans instead of analyzing scans.


• Other healthcare professionals, such as nurses, increasing their training to use medical imaging AI software.


• Deep learning software used to identify cancers and recommend treatment plans.


• More biotech firms investing in developing fully automated medical imaging systems.

Questions to consider

• How have you seen DL software applied in your company or industry?


• How else can automation improve medical imaging diagnoses and disease prevention?