Clinical imaging encompasses every technique physicians use to see inside the living body without making an incision, from a plain chest X-ray to a PET scan that tracks the metabolism of individual tumors. What began with a single technology in 1895 now spans dozens of modalities, each exploiting different physical principles to reveal anatomy, physiology, or molecular activity. The field is evolving fast: artificial intelligence is reshaping how images are reconstructed, portable MRI scanners are rolling into intensive care units, and some imaging agents can now double as cancer therapies. Understanding the landscape of clinical imaging means understanding not just what each tool does, but where the real safety trade-offs lie, how new computational methods are changing the dose-versus-quality equation, and where the technology is headed.
How the Major Modalities Work
Clinical imaging modalities split roughly by the physics they exploit. X-ray and CT rely on differences in how tissues absorb ionizing radiation. Ultrasound bounces high-frequency sound waves off tissue interfaces and measures the echoes. MRI uses strong magnetic fields and radiofrequency pulses to detect how hydrogen atoms in water and fat behave when their spin is disturbed. Nuclear medicine techniques like PET and SPECT inject tiny amounts of radioactive tracers and then detect the gamma rays those tracers emit from inside the body. Each modality has a sweet spot: CT excels at rapid, high-resolution views of bone and lung; MRI is unmatched for soft-tissue contrast in the brain, joints, and abdomen; ultrasound is cheap, portable, and ideal for real-time guidance during procedures; and PET provides metabolic information no anatomical scan can replicate.
One reason MRI produces such rich soft-tissue contrast is that different tissues return their magnetic signal at different rates. These rates, called T1 and T2 relaxation times, vary between healthy and diseased tissue, between cartilage and muscle, and even between scanners of different field strengths. Tissue relaxation values measured at the increasingly common 3.0 Tesla scanners differ from those at 1.5 Tesla: T1 values tend to be higher at 3.0 T while T2 values drop slightly.1PubMed. Musculoskeletal MRI at 3.0 T: relaxation times and image contrast Knowing these values precisely matters because it lets engineers optimize the pulse sequences that generate the image, and it opens the door to quantitative MRI, where a scan does not just show a picture but assigns a reproducible number to each tissue region. Quantitative MRI techniques can now track cartilage degradation, characterize liver disease, and help differentiate benign from malignant tumors by comparing relaxation parameters against known reference values.2PubMed Central. Applications of T1 and T2 relaxation time calculation in tissue differentiation and cancer diagnostics—a systematic literature review
Ultrasound Beyond the Baby Picture
Most people associate ultrasound with pregnancy scans, but the technology reaches far beyond obstetrics. Doppler ultrasound measures the speed and direction of blood flow by detecting the frequency shift of sound waves bouncing off moving red blood cells. This makes it a frontline tool for evaluating blood vessel disease, from deep vein thrombosis to carotid artery stenosis. Validation studies comparing pulsed Doppler measurements of blood-flow velocity against invasive electromagnetic flow probes have shown strong agreement, confirming that the technique reliably captures vascular resistance and impedance without needing a catheter.3PubMed. Determination of vascular impedance in the peripheral circulation by transcutaneous pulsed Doppler ultrasound Because ultrasound involves no ionizing radiation, it can be repeated as often as needed and used safely in children and pregnant patients, which partly explains why it remains the most commonly performed imaging study worldwide.
PET Imaging and Tumor Metabolism
Positron emission tomography offers something fundamentally different from CT or MRI: it images what cells are doing, not just what they look like. The workhorse tracer, FDG (a radioactive sugar analog), is taken up avidly by cells with high glucose consumption, which includes most cancers. FDG uptake depends on several factors: the density of glucose transporters on the cell surface, the activity of the enzyme hexokinase that traps the sugar inside the cell, and the local blood supply. High FDG uptake has been linked to more aggressive tumors and poorer prognosis across many cancer types.4Clinical Cancer Research. Hypoxia and Glucose Metabolism in Malignant Tumors: Evaluation by [18F]Fluoromisonidazole and [18F]Fluorodeoxyglucose Positron Emission Tomography Imaging But the relationship between tumor biology and tracer signal is not always straightforward. Tumors can be hypoxic yet metabolically quiet, or well-oxygenated yet glucose-hungry; all four combinations exist, which is why relying on a single PET tracer to characterize a tumor’s biology has limits.5Clinical Cancer Research. Hypoxia and Glucose Metabolism in Malignant Tumors: Evaluation by [18F]Fluoromisonidazole and [18F]Fluorodeoxyglucose Positron Emission Tomography Imaging
More specialized tracers are emerging. In certain rare tumors driven by specific genetic mutations, the mechanism behind increased FDG uptake turns out to be accelerated phosphorylation by hexokinases rather than increased expression of glucose transporters, a distinction that can matter when designing targeted therapies.6Journal of Nuclear Medicine. Correlation Between In Vivo 18F-FDG PET and Immunohistochemical Markers of Glucose Uptake and Metabolism in Pheochromocytoma and Paraganglioma
Radiation Dose and the Evolving Safety Picture
Any imaging study that uses ionizing radiation carries a small potential cancer risk, and the cumulative radiation dose from medical imaging in the United States has climbed dramatically since the 1980s, driven largely by the rise of CT scanning.7PubMed Central. Cancer risks associated with external radiation from diagnostic imaging procedures Large-scale studies are now tracking the association between cumulative childhood exposure and subsequent cancer risk, though definitive results take decades to mature.8PubMed Central. Quantifying cancer risk from exposures to medical imaging in the Risk of Pediatric and Adolescent Cancer Associated with Medical Imaging (RIC) Study: research methods and cohort profile
The good news is that dose-reduction technologies have made enormous strides. Technical innovations in CT hardware and software have enabled some applications to deliver effective doses below 1 millisievert, well within the range of natural background radiation a person receives in a few months.9Physics in Medicine & Biology. Dose in x-ray computed tomography Monte Carlo simulations can now estimate the dose each individual organ receives during a scan, giving radiologists patient-specific information to guide optimization rather than relying on generic phantom-based estimates.10Physics in Medicine & Biology. Dose in x-ray computed tomography
Contrast Agents and Rethinking Kidney Risk
Many CT and MRI scans are performed with contrast agents injected into the bloodstream to make blood vessels and abnormal tissues stand out. For CT, those agents contain iodine; for MRI, they contain gadolinium. Both have safety profiles that have been reassessed in recent years.
For decades, iodinated contrast was treated as a serious threat to the kidneys. Patients with reduced kidney function were sometimes denied contrast-enhanced CT scans out of fear of contrast-induced kidney injury. Emerging evidence now suggests that this risk was substantially overstated. Many cases attributed to contrast were probably caused by other factors that happened to coincide with the scan, such as low blood pressure, dehydration, or nephrotoxic medications.11PubMed. Risk of Acute Kidney Injury Following IV Iodinated Contrast Media Exposure: 2023 Update, From the AJR Special Series on Contrast Media A joint consensus statement from the American College of Radiology and the National Kidney Foundation concluded that the risk of kidney injury caused by contrast has been overstated, largely because older studies lacked the control groups needed to separate contrast-caused injury from injury that merely happened around the same time as contrast exposure.12PubMed. Use of Intravenous Iodinated Contrast Media in Patients with Kidney Disease: Consensus Statements from the American College of Radiology and the National Kidney Foundation This shift means more patients with kidney disease can now receive contrast-enhanced scans when the diagnostic benefit is clear, rather than being steered toward less informative alternatives.
Gadolinium-based contrast agents for MRI carry a different concern. Repeated administration has been linked to gadolinium deposits in the brain, particularly in the dentate nuclei and globus pallidus.13PubMed. Gadolinium deposition within the dentate nucleus and globus pallidus after repeated administrations of gadolinium-based contrast agents-current status The deposits show up as signal changes on subsequent MRI scans. Linear gadolinium chelates appear to cause more deposition than the more chemically stable macrocyclic types, although deposition has been observed with macrocyclic agents too.14The Lancet Neurology. Recommendations for the use of gadolinium-based contrast agents No clear clinical harm from brain gadolinium deposition has been established in people with normal kidney function, but the finding has prompted regulators to restrict certain linear agents and has pushed the field toward macrocyclic formulations as the default choice.
Deep Learning and the Dose-Quality Trade-Off
One of the most impactful recent developments in clinical imaging is the application of deep learning to image reconstruction. Traditional CT reconstruction methods cope with low radiation doses by applying iterative algorithms that smooth out noise, but they can also blur fine detail or introduce an unfamiliar image texture that some radiologists find harder to read. Deep-learning image reconstruction (DLIR) trains neural networks on vast datasets of high-quality images and then applies the learned patterns to clean up noisy, low-dose acquisitions.
The results have been striking. In chest CT, a low-dose protocol combined with DLIR reduced the radiation dose by about 96% compared with standard-dose scans while maintaining image noise and overall quality that were statistically indistinguishable from the standard-dose images processed with conventional iterative reconstruction.15PubMed Central. Application of deep learning image reconstruction in low-dose chest CT scan Phantom studies have confirmed that DLIR reduces noise and improves spatial resolution and the ability to detect small structures without altering noise texture, suggesting even greater potential for dose optimization than older hybrid iterative methods.16PubMed. Image quality and dose reduction opportunity of deep learning image reconstruction algorithm for CT: a phantom study The noise-reduction benefit appears particularly large in the low-dose range, which has implications not just for diagnostic scanning but also for treatment planning CT in radiation oncology, where dose to the patient during planning adds to the overall radiation burden.17Scientific Reports. Validation of deep learning-based CT image reconstruction for treatment planning
Dual-Energy CT and Material Decomposition
Standard CT scans capture images at a single X-ray energy level. Dual-energy CT acquires data at two different energy levels simultaneously, which allows software to distinguish materials that look identical on a regular scan. Because different substances (calcium, iodine, uric acid, water) attenuate X-rays differently at different energies, dual-energy CT can subtract out bone to visualize blood vessels, identify uric acid crystals in a gouty joint, or map the calcium content of vertebral bone to screen for osteoporosis. In bone-density assessment, dual-energy material decomposition has shown strong correlation with dedicated quantitative CT measurements, achieving sensitivity and specificity above 85% for detecting reduced bone density.18PubMed Central. Diagnostic Accuracy of Dual-Energy CT Material Decomposition Technique for Assessing Bone Status Compared with Quantitative Computed Tomography The appeal is that bone-density information can be extracted from a CT scan performed for another clinical reason, saving the patient a separate trip and an additional radiation dose.
Radiomics and Radiogenomics
Routine clinical images contain far more information than the human eye can extract. Radiomics is the computational extraction of hundreds or thousands of quantitative features from medical images, capturing aspects of a tumor’s shape, texture, and internal heterogeneity that are invisible to visual interpretation. A landmark study of over 1,000 patients with lung or head-and-neck cancer extracted 440 image features from CT data and found that many had prognostic value, with a signature reflecting internal tumor heterogeneity linked to underlying gene-expression patterns.19Nature Communications. Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach This kind of analysis has grown rapidly, with models now predicting outcomes like overall survival and recurrence across brain, lung, head-and-neck, and other cancers.20Nuclear Medicine Communications. Radiomics: a quantitative imaging biomarker in precision oncology
A closely related field, radiogenomics, goes a step further by correlating imaging features directly with the tumor’s genetic profile. The idea is that a tumor’s appearance on a scan reflects its molecular makeup, and if that correlation can be mapped reliably, imaging could serve as a noninvasive proxy for genomic information that currently requires a biopsy.21PubMed Central. Radiogenomic imaging-linking diagnostic imaging and molecular diagnostics If validated at scale, radiogenomics could guide treatment selection, flag patients who need genomic testing, or monitor how a tumor’s biology shifts over time, all without repeated tissue sampling.22PubMed Central. Background, current role, and potential applications of radiogenomics The field is promising but still largely in the research phase; most published models need external validation before they can influence clinical decisions.
Theranostics and the Blurring Line Between Diagnosis and Treatment
One of the more exciting shifts in nuclear medicine is theranostics, an approach that uses the same molecular target for both imaging and therapy by swapping the radionuclide attached to a targeting molecule.23PubMed Central. Recent advances in theranostics and oncology PET: emerging radionuclides and targets A diagnostic scan with a PET-emitting isotope first confirms that a patient’s tumor expresses the target; if it does, a therapeutic version of the same molecule, now carrying an isotope that delivers cell-killing radiation, is administered. The concept has already proven successful in clinical practice for neuroendocrine tumors and certain prostate cancers, and researchers are now extending it to other malignancies.
Preclinical work is targeting new antigens. One team developed a theranostic pair aimed at CD46, a protein overexpressed in multiple myeloma and prostate cancer: a zirconium-89-labeled version for PET imaging and an actinium-225-labeled version for therapy, showing feasibility in mouse models.24PubMed Central. CD46-Targeted Theranostics for PET and 225Ac-Radiopharmaceutical Therapy of Multiple Myeloma Another group is developing antibodies targeting galectin-3, a protein found in several hard-to-treat cancers including triple-negative breast cancer, with promising early imaging and therapeutic results in animal models.25PubMed Central. Development of galectin-3 targeted radioimmunoligands for cancer theranostics These are still preclinical, but they illustrate the direction: rather than imaging and treating as separate processes, the molecular probe does both jobs.
Portable MRI at the Bedside
Conventional MRI scanners are large, expensive, and require magnetically shielded rooms. Critically ill patients often cannot be safely transported to the scanner suite. Ultra-low-field portable MRI devices, operating at a fraction of the magnetic field strength of standard systems, are changing that equation. In one early study, a portable device was wheeled directly to the bedsides of 50 ICU patients, including those with strokes, brain tumors, and traumatic brain injuries. The device found brain abnormalities in 97% of patients who did not have COVID-19, and its findings were in agreement with conventional imaging in nearly every case.26JAMA Neurology. Assessment of Brain Injury Using Portable, Low-Field Magnetic Resonance Imaging at the Bedside of Critically Ill Patients No adverse events occurred during any of the bedside scans.
The technology has since been tested in even more challenging settings. A multicenter study examined bedside ultra-low-field MRI in patients on extracorporeal membrane oxygenation (ECMO), who are among the sickest in any hospital. The portable scanner detected brain injury in 44% of those patients. For ischemic strokes specifically, the portable MRI picked up all eight events while standard head CT caught only half.27PubMed Central. Clinical Use of Bedside Portable Ultra-Low-Field Brain Magnetic Resonance Imaging in Patients on Extracorporeal Membrane Oxygenation: Results From the Multicenter SAFE MRI ECMO Study Preliminary safety testing has also demonstrated that even patients with intra-aortic balloon pumps, which contain metal components, can be safely imaged at the 0.064 Tesla field used by these devices.28PubMed Central. Safety of Bedside Portable Low-Field Brain MRI in ECMO Patients Supported on Intra-Aortic Balloon Pump Portable MRI will not replace conventional scanners for detailed diagnostic work, but for triage and monitoring of critically ill patients who cannot leave the ICU, it fills a gap that previously did not have a good solution.
Fluorescence-Guided Surgery and Photoacoustic Imaging
Not all clinical imaging happens before or after treatment. Some of the most impactful recent advances involve imaging during procedures. Near-infrared fluorescence imaging uses a dye, typically indocyanine green (ICG), injected into the bloodstream or near a tumor. When excited by near-infrared light, the dye glows, revealing structures that would be invisible under normal operating room lighting. In gastric cancer surgery, ICG fluorescence can guide lymph node dissection, help localize the tumor, and define surgical margins in real time.29PubMed Central. Indocyanine green and near-infrared fluorescence-guided surgery for gastric cancer: a narrative review In minimally invasive surgery for rare abdominal tumors, fluorescence guidance has identified lesions that were completely invisible under standard white-light visualization, making the difference between finding the tumor and missing it.30Scientific Reports. Fluorescence-guided minimally-invasive resection of abdominal paragangliomas using indocyanine green
Photoacoustic imaging represents a different hybrid approach: short pulses of laser light are absorbed by tissue, which then emits ultrasound waves that can be captured by conventional transducers. Because hemoglobin absorbs light strongly, the technique is particularly good at mapping tiny blood vessels without any injected contrast.31PubMed Central. Photoacoustic imaging for microcirculation It can achieve resolution down to the micron scale and has pushed past the depth limits of purely optical methods, reaching into tissue at depths where scattered light alone would produce no usable image.32PubMed Central. Photoacoustic imaging and characterization of the microvasculature Clinical applications are still emerging, but research is active in areas like breast cancer detection, skin lesion characterization, and monitoring tumor response to anti-angiogenic therapies.
Fetal MRI and Prenatal Diagnosis
Ultrasound remains the primary imaging tool in pregnancy, but when ultrasound findings are ambiguous or incomplete, fetal MRI provides a valuable second look. A systematic review and meta-analysis found that fetal MRI makes a significant contribution to the accurate diagnosis of congenital abnormalities, with particular strength in detecting genitourinary anomalies.33PubMed Central. The value of fetal magnetic resonance imaging in diagnosis of congenital anomalies of the fetal body: a systematic review and meta-analysis Because MRI uses no ionizing radiation, the main safety consideration is the acoustic noise and the potential for tissue heating from the radiofrequency pulses, both of which are managed through standard protocols. Fetal MRI is typically performed in the second or third trimester, when the fetus is large enough for detailed anatomical assessment and when findings might alter pregnancy management or delivery planning.
Ultra-Portable X-Ray and Global Access
Two-thirds of the world’s population has limited or no access to diagnostic imaging. Ultra-portable X-ray systems, light enough to be carried by one person and powered by a battery, are starting to address this gap. Field evaluations have shown that these devices can be deployed at the community and even household level, extending X-ray access to populations who would otherwise need to travel long distances to a hospital.34PubMed Central. Early Evaluation of an Ultra-Portable X-ray System for Tuberculosis Active Case Finding The primary use case so far is tuberculosis screening, where a chest X-ray combined with computer-aided detection software can flag suspected cases in the field, allowing earlier referral for confirmatory testing.
Qualitative studies of healthcare providers using these systems found that the portability was generally well received and that the pairing with AI-based detection software helped decentralize radiological assessment for TB. However, improved portability came with trade-offs: smaller detectors limit the field of view, battery life constrains throughput, and some providers found that image quality, while diagnostically adequate, did not match hospital-grade equipment.35PLOS ONE. Early user experience and lessons learned using ultra-portable digital X-ray with computer-aided detection (DXR-CAD) products: A qualitative study from the perspective of healthcare providers Despite these compromises, the systems represent a meaningful step toward bringing basic imaging to the estimated billions of people who currently lack it, and they are likely to improve rapidly as detector technology, battery capacity, and AI algorithms continue to advance.
Resting-State Brain Connectivity
Functional MRI, or fMRI, is often associated with studies that map which brain regions “light up” during a task. But a different application has quietly become one of the most productive areas of neuroimaging research: resting-state functional connectivity. Instead of asking a person to perform a task inside the scanner, researchers simply acquire a continuous run of images while the brain idles. Spontaneous low-frequency fluctuations in the MRI signal turn out to be correlated between brain regions that belong to the same functional network, and these correlations can be detected reliably across subjects.36PubMed Central. Assessing functional connectivity in the human brain by fMRI Resting-state scans have revealed connectivity patterns in the visual, motor, language, and working-memory systems without the patient needing to do anything, which makes the technique especially useful for populations who cannot follow instructions, including infants, sedated patients, and people with severe neurological impairment. Clinical applications are still being refined, but resting-state connectivity maps are increasingly used in presurgical planning for brain tumor patients and in research on disorders like Alzheimer’s disease, autism, and depression, where the connections between brain regions may be more informative than the structure of any single region.