Innovations in Radiology

Radiology has always been closely connected to advances in technology. From the discovery of X-rays to the development of CT, MRI, ultrasound, and image-guided procedures, innovation has continually changed how physicians visualize, diagnose, and treat disease.

That evolution continues today. Artificial intelligence, advances in imaging technology, molecular imaging, and increasingly precise image-guided therapies are expanding what medical imaging can do.

In this section, explore some of the innovations shaping radiology today and where the field may be headed in the future.

As technology continues to evolve, so will this section of ExploreRadiology!

What Is Artificial Intelligence?

Artificial intelligence (AI) refers broadly to computer systems designed to perform tasks that typically require aspects of human intelligence, such as recognizing patterns, processing language, making predictions, or analyzing complex information.

In radiology, AI can be applied throughout the imaging process from how images are acquired and reconstructed to how they are analyzed, interpreted, and incorporated into clinical workflow.

Image Analysis & Detection

AI algorithms can analyze medical images and identify patterns associated with particular findings. Depending on the specific application, these tools may help detect, characterize, or quantify abnormalities on radiographs, CT, MRI, ultrasound, and other imaging studies.

For example, an AI system might help identify a suspected abnormality, measure a lesion, or highlight an area of an image for the radiologist to review.

Many AI tools are designed to assist rather than independently make a final diagnosis. They can provide additional information that radiologists incorporate into their interpretation.

Workflow & Triage

AI can also help radiologists manage the large volume of imaging studies encountered in modern medical practice.

Some AI tools can analyze incoming examinations for potentially time-sensitive abnormalities and help prioritize certain studies for earlier review. Other applications can automate repetitive tasks, organize information, or help integrate relevant clinical data into the radiologist's workflow.

In this way, AI has the potential to affect not only what radiologists see, but also how efficiently imaging information moves through the healthcare system.

Segmentation & Quantification

Segmentation involves identifying and outlining a structure or region within an image.

AI can help automate tasks such as measuring organs, tumors, blood vessels, or other structures. These measurements can provide quantitative information and may help radiologists evaluate disease burden or monitor changes over time.

Automating these tasks can make quantitative measurements more efficient and reproducible.

Image Acquisition & Reconstruction

AI can contribute to the creation of medical images themselves.

After imaging data are collected, reconstruction algorithms transform those data into the images that radiologists interpret. AI-based techniques can help reduce image noise, improve image quality, accelerate image acquisition, or assist in reconstructing diagnostically useful images from limited data, depending on the modality and application.

This is an important reminder that AI in radiology extends beyond simply analyzing a finished image.

Generative AI & Radiology Reports

Generative AI and large language models are also being explored for applications involving radiology reporting, clinical information, and communication.

Potential applications include assisting with drafting or structuring radiology reports, summarizing clinical information, generating impressions, and translating complex imaging information into more understandable language.

However, generated information can be incomplete or incorrect. Clinical use therefore requires appropriate validation and human oversight.

What Are the Limitations?

AI systems are only as useful as their performance in the clinical settings in which they are used. Performance can be affected by factors such as the quality and representativeness of training data, differences between patient populations and imaging equipment, and changes in clinical practice.

AI systems can also produce errors, false positives, false negatives, and unexpected results. Questions involving bias, privacy, transparency, regulation, responsibility, and appropriate human oversight remain important as these technologies become more widely incorporated into healthcare.

Will AI Replace Radiologists?

AI is already changing how some radiology tasks are performed, but the role of a radiologist extends far beyond pattern recognition.

Radiologists integrate imaging findings with clinical information, compare examinations over time, communicate with referring physicians and patients, perform image-guided procedures, determine appropriate imaging strategies, and make complex medical judgments.

Rather than viewing AI only in terms of replacement, it can be more useful to understand it as a developing set of tools that may change how radiologists work and which tasks are performed by humans, computers, or a combination of both.

The exact role of AI in radiology will continue to evolve as the technology develops and its applications are studied in clinical practice.

Putting It All Together

Artificial intelligence has potential applications throughout the radiology workflow:

Image Acquisition → Reconstruction → Analysis → Detection → Quantification → Workflow → Reporting → Communication

AI is therefore not one single tool or algorithm. It represents a broad and rapidly developing group of technologies that can interact with many different parts of medical imaging.

As AI continues to develop, understanding both what these technologies can do and where their limitations lie will be increasingly important for the next generation of radiologists.

What Is Photon-Counting CT?

Photon-counting CT is an advanced form of computed tomography that uses a different type of detector to measure the X-rays that pass through the patient.

Conventional CT scanners typically use energy-integrating detectors, which measure the combined energy deposited by many incoming X-ray photons. Photon-counting detectors instead detect individual X-ray photons and measure information about their energy.

This difference in how X-rays are detected allows photon-counting CT to extract additional information from the X-ray beam.

Why Is It Different?

Photon-counting detectors directly convert incoming X-ray photons into electrical signals. Because individual photons can be detected and separated according to their energy, the scanner can preserve spectral information that conventional energy-integrating detectors do not measure in the same way.

The detector design can also support smaller detector elements, allowing images to be acquired with very high spatial resolution.

What Are the Potential Advantages?

Photon-counting CT can provide several advantages compared with conventional CT, including:

Higher spatial resolution: Smaller structures and fine anatomical details can be visualized more clearly.

Spectral information: Measuring information about photon energies can help distinguish and characterize different materials.

Improved dose efficiency: The technology can make more efficient use of detected X-rays, creating opportunities to maintain diagnostic image quality at lower radiation doses for some applications.

Improved tissue and material characterization: Spectral information can help differentiate materials such as iodine, calcium, and other substances based on their X-ray attenuation characteristics at different energies.

The exact benefit depends on the clinical application and imaging protocol.

Connecting It to the Physics

Remember that CT creates images by measuring how much X-rays are attenuated as they pass through the body.

Conventional CT with energy-integrating detectors primarily measures the combined energy deposited by many detected X-ray photons. Photon-counting CT goes a step further by detecting individual photons and obtaining information about their energies.

Because different materials interact with X-rays differently across the energy spectrum, this additional information can help the scanner distinguish and characterize different materials.

Why Is It Exciting?

Photon-counting CT demonstrates how changing just one fundamental part of an imaging system (the detector) can expand what the entire modality is capable of doing.

Higher spatial resolution and spectral imaging are already being explored and used across areas such as cardiac, vascular, thoracic, and musculoskeletal imaging, while additional clinical applications continue to be studied.

As photon-counting CT becomes more widely available, continued research will help determine where its technological advantages provide the greatest improvements in patient care.

Why Does MRI Continue to Evolve?

Magnetic resonance imaging is a highly versatile imaging modality. By changing how MRI signals are acquired and processed, different tissue properties can be emphasized and many different types of images can be created.

Advances in scanner hardware, imaging techniques, and computing continue to improve how quickly MRI examinations can be performed, the detail that can be visualized, and the information that can be obtained from tissues.

Faster MRI

One challenge of MRI is that examinations can take considerably longer than many other imaging studies. Long acquisition times can increase the chance of patient motion and make MRI more difficult for patients who have trouble remaining still.

Techniques such as parallel imaging, compressed sensing, and AI-assisted reconstruction can help accelerate MRI acquisition and image reconstruction.

Faster imaging has the potential to shorten examinations, reduce motion artifacts, and make MRI more accessible and comfortable for patients.

Improving Image Quality With AI

Artificial intelligence can also be incorporated into MRI reconstruction and image processing.

AI-based techniques can help reduce image noise, improve image quality, correct certain artifacts, and reconstruct useful images from more limited acquired data.

This creates an important connection between advances in MRI and the artificial intelligence tools discussed earlier: innovation in radiology increasingly occurs at the intersection of imaging physics, hardware, and computing.

Quantitative MRI

Traditional MRI interpretation often relies on the relative appearance of tissues—for example, whether a structure appears bright or dark on a particular sequence.

Quantitative MRI aims to measure specific tissue properties and produce numerical information that can potentially be compared across examinations or over time.

Depending on the technique, MRI can provide quantitative information related to properties such as relaxation, diffusion, fat content, iron concentration, and blood flow.

These measurements have the potential to provide imaging biomarkers that help characterize tissue and monitor disease.

Higher-Field MRI

The strength of an MRI scanner's magnetic field is measured in tesla (T). Many clinical MRI examinations are currently performed using 1.5T or 3T scanners, while higher-field systems such as 7T MRI can provide additional capabilities in certain applications.

Increasing magnetic field strength can provide greater signal, which can potentially be used to achieve higher spatial resolution or visualize structures in greater detail.

However, higher magnetic field strength also introduces technical challenges and does not automatically make every MRI examination better.

Connecting It to the Physics

Remember that MRI creates images by manipulating and measuring signals related to hydrogen protons within a magnetic field.

MRI innovation often involves finding new ways to generate, detect, encode, reconstruct, and analyze those signals.

Rather than representing one completely new imaging modality, many advances in MRI build upon the same fundamental physics you learned earlier while improving how efficiently we acquire the signal or how much information we can extract from it.

Why Is It Exciting?

MRI already allows radiologists to examine anatomy and tissue characteristics without using ionizing radiation. Continued advances are expanding what information can be obtained while also working to address some of MRI's traditional limitations, including long examination times and sensitivity to motion.

Future developments in faster acquisition, quantitative imaging, artificial intelligence, scanner hardware, and other emerging techniques may continue to expand how MRI is used to detect, characterize, and monitor disease.

From Images to 3D Models

CT and MRI examinations often produce a series of cross-sectional images through the body. Although radiologists commonly interpret these individual images, the underlying imaging data can also be processed to create three-dimensional representations of anatomy and disease.

Using specialized software, structures such as bones, blood vessels, organs, and tumors can be isolated from surrounding tissues and reconstructed into 3D models.

These models allow complex anatomy to be viewed from different angles and can make spatial relationships easier to understand.

3D Reconstruction

3D reconstruction uses data from an imaging examination to create a three-dimensional representation of structures within the body.

Different visualization techniques can emphasize different anatomical features. For example, a CT angiogram can be processed to display a three-dimensional representation of blood vessels, while a CT of a complex fracture can be reconstructed to better demonstrate the relationship between bone fragments.

Importantly, these images are not created from a new scan. They are additional representations generated from imaging data that have already been acquired.

Surgical & Procedural Planning

Three-dimensional visualization can help physicians understand complex anatomy before certain surgeries or procedures.

A 3D model may help demonstrate the relationship between a tumor and nearby blood vessels, show the anatomy of a complex fracture, or help physicians plan an approach to a cardiovascular or other anatomically complex procedure.

By transforming imaging data into a more intuitive representation of anatomy, 3D visualization can complement traditional cross-sectional images during surgical and procedural planning.

3D Printing

Imaging data can also be used to create physical three-dimensional models.

After structures are identified and converted into a digital 3D model, that model can be used to produce a patient-specific object with a 3D printer.

These models can be used in selected settings for surgical planning, medical education, device development, and communication with patients.

This is a unique example of medical imaging moving beyond the screen and becoming a physical representation of a patient's anatomy.

Virtual & Augmented Reality

Advances in computing are also creating new ways to interact with medical imaging.

Virtual reality (VR) can place users inside an immersive digital environment where three-dimensional anatomical models can be viewed and manipulated.

Augmented reality (AR) instead overlays digital information onto the user's view of the real world.

These technologies are being explored for applications such as medical education, visualization of complex anatomy, surgical and procedural planning, and image-guided interventions.

Connecting It to Radiology

Radiology generates enormous amounts of spatial information about the human body. Traditionally, much of that information is viewed as a series of two-dimensional images on a screen.

3D reconstruction, printing, and immersive visualization provide different ways to organize and interact with the same anatomical information.

These tools do not replace careful interpretation of the original images. Instead, they can provide additional perspectives that may make complex anatomy and spatial relationships easier to understand.

Why Is It Exciting?

Advances in 3D visualization are changing the ways imaging information can be viewed, communicated, and applied.

A CT or MRI examination can begin as a collection of cross-sectional images and ultimately become an interactive digital model, a physical 3D-printed object, or an immersive visualization.

As these technologies continue to develop, medical imaging may become increasingly interactive—allowing physicians, trainees, and patients to engage with anatomy in ways that extend beyond the traditional radiology workstation.

What Is Molecular Imaging?

Many medical imaging examinations primarily provide information about anatomy and structure. Molecular imaging adds another dimension by allowing physicians to visualize and measure biological processes occurring within the body.

Molecular imaging commonly uses small amounts of radioactive substances called radiopharmaceuticals. These agents are designed to participate in or localize to particular physiological, biochemical, or molecular processes.

By detecting where a radiopharmaceutical travels and accumulates, physicians can gain information about how tissues are functioning, sometimes before major structural changes become apparent on conventional imaging.

How Does It Work?

A radiopharmaceutical contains a radioactive component that allows its distribution within the body to be detected, often combined with or incorporated into a molecule that determines where it travels or accumulates.

After the radiopharmaceutical is administered, its distribution can be detected using imaging technologies such as positron emission tomography (PET) or single-photon emission computed tomography (SPECT).

The resulting images provide information about the distribution of the radiopharmaceutical and therefore about the biological process it is designed to evaluate.

PET Imaging

Positron emission tomography (PET) is one of the most widely used forms of molecular imaging.

Different PET radiopharmaceuticals can be designed or selected to evaluate different biological processes. One commonly used radiopharmaceutical is 18F-fluorodeoxyglucose (FDG), a glucose analog that can provide information about glucose metabolism.

Because many cancers demonstrate increased glucose metabolism, FDG PET can help detect and evaluate certain malignancies. However, increased FDG uptake is not specific to cancer and can also occur in normal tissues, inflammation, infection, and other processes.

This illustrates an important principle of molecular imaging: understanding an image requires understanding why a particular radiopharmaceutical accumulates where it does.

Combining Molecular & Anatomic Imaging

Molecular imaging is frequently combined with anatomical imaging.

PET/CT combines information about radiopharmaceutical distribution from PET with detailed anatomical information from CT. PET/MRI similarly combines molecular information from PET with the soft-tissue contrast and other capabilities of MRI.

Combining these types of information can help physicians determine both where an abnormal biological process is occurring and what anatomical structures are involved.

What Are Theranostics?

Theranostics combines diagnosis and therapy by using radiopharmaceuticals that target the same or closely related biological targets.

A diagnostic radiopharmaceutical can first be used to determine whether a particular molecular target is present and where it is located within the body.

If appropriate, a related therapeutic radiopharmaceutical can then deliver ionizing radiation to cells or tissues expressing that target.

In this way, molecular imaging can help identify patients who may benefit from a targeted radiopharmaceutical therapy and can help visualize the biological target that the treatment is designed to reach.

An Example: Prostate Cancer

Prostate-specific membrane antigen (PSMA) is a protein that is highly expressed in many prostate cancers.

Radiopharmaceuticals targeting PSMA can be used with PET imaging to identify PSMA-expressing disease. For selected patients, a therapeutic radiopharmaceutical targeting PSMA can then be used to deliver radiation to sites of disease.

This is an example of the theranostic concept:

Identify the target → Image the target → Treat the target

Other theranostic approaches use different molecular targets and radiopharmaceuticals for different diseases.

Connecting It to Radiology

Molecular imaging demonstrates that medical imaging can provide information about more than what a structure looks like.

By selecting radiopharmaceuticals that interact with particular biological processes or molecular targets, imaging can provide information about what tissues are doing at a molecular or physiological level.

Theranostics takes this concept one step further by connecting that information with targeted treatment.

Why Is It Exciting?

Molecular imaging and theranostics are helping move medical imaging toward increasingly personalized diagnosis and treatment.

Instead of relying only on the location and appearance of disease, physicians may be able to identify specific biological targets, visualize those targets throughout the body, and select therapies designed to act on them.