There are many reasons why some biomedical imaging modalities are a clinical mainstay. Foremost is what can actually be imaged, which is largely determined by the interactions between tissue and electromagnetic waves or sound waves (depending on the modality). On the one hand, modalities that leverage high-energy radiation (conventional X-ray radiography, computed tomography (CT), positron emission tomography (PET) and single-photon emission computed tomography (SPECT)) provide planar images (if using X-ray radiography) or cross-sectional volumetric images (if using CT) of large parts of the body, or can sensitively detect injected radiolabelled tracers (via their gamma-ray emissions, if using PET or SPECT). On the other hand, modalities leveraging mechanical waves, magnetic fields or non-ionizing radiation (ultrasound, magnetic resonance imaging (MRI) and optical imaging) can only image relatively superficial areas of tissue (this is the case for ultrasound and optical imaging, because tissue scatters, absorbs and reflects incident light and sound, attenuating and limiting their penetration), or have low molecular sensitivity (as is the case for ultrasound and MRI).

Diagnostic needs and patient-safety considerations also determine clinical uses. X-ray-based transmission imaging is advantageous for anatomical imaging of hard tissue, and emission tomography for metabolic and molecular imaging of soft tissue. Visible and near-infrared light, mechanical waves and magnetic fields are, in principle, innocuous to the body (within certain ranges of intensity and wavelength), but the limited molecular sensitivity and specificity of conventional ultrasound and MRI requires the injection of targeted agents for enhancing contrast resolution (such as nanoparticles of gadolinium or iron compounds for MRI, and microbubbles for ultrasound). Multimodal uses (such as CT/PET and PET/MRI) can provide anatomical, physiological and molecular information. And, regardless of the modality, the utility of multiplexing — detecting multiple biomarkers simultaneously — depends on whether the need is to image biopsied tissue or intact tissue, as discussed by Chrysafis Andreou and colleagues in a Review Article in this issue of Nature Biomedical Engineering.

Among the seven most frequently used imaging techniques in medical diagnostics (X-ray radiography, CT, PET, SPECT, MRI, ultrasound and optical imaging), only optical imaging can reach spatial resolutions within the imaging plane below a few tens of micrometres and hence be used to visualize cells (the typical resolutions of most clinical modalities fall within 0.1–10 mm). And the typically low temporal resolutions (upwards of seconds) of conventional clinical MRI and ultrasound preclude the observation, let alone quantification, of many microscopic and molecular phenomena. In fact, specific molecular processes (such as increased metabolism or hypoxia in cancerous tissue) in patients can more sensitively be detected (yet at macroscopic spatial resolutions) via targeted molecular probes whose gamma-ray emissions are captured by scintillation detectors for PET or SPECT.

Functionality and performance differences across biomedical imaging techniques can indeed be stark, in particular when comparing techniques for imaging human tissue being developed in research laboratories and the established clinical modalities. For instance, fast and highly sensitive cameras and optimized optical paths and instrumentation can leverage the contrast provided by endogenous fluorophores in tissue, tissue autofluorescence, or bright exogenous fluorescent or phosphorescent near-infrared probes to, respectively, image mineral deposits in excised calcified aortic valves (L. M. Baugh et al. Nat. Biomed. Eng. 1, 914–924; 2017), large surgical specimens rapidly (J. T. C. Liu et al. Nat. Biomed. Eng. 5, 203–218; 2021) or subcutaneous tumour vasculature in small animals (as shown in an Article in this issue), at resolutions of the order of a few micrometres. And newer MRI strategies (also reported in this issue) allow for the tracking of unlabelled cells injected in the brain of mice, or provide quantitative maps of pH in murine brain tumours without requiring exogenous sources of contrast by leveraging chemical exchange saturation transfer (the radiofrequency-induced magnetic re-saturation of exchangeable protons between certain chemical species and the surrounding bulk water) as a mechanism to generate better contrast.

Such improvements in performance can make imaging technology more translationally suitable. Five research Articles included in this issue exemplify this well. Jan Grimm and colleagues show, with prospective clinical testing of 96 patients with existing or suspected tumours who had undergone routine molecular imaging, that Cerenkov luminescence imaging (CLI) for the detection of tumour location is clinically feasible for a range of radioisotopes. CLI scans for the distribution of the transmitted portion of the really weak and predominantly blue light generated by superluminal charged particles emitted by radiotracers in the tissue. To enhance the signal for detecting tumour location, the researchers designed a lightproof enclosure, to remove ambient light, and a fibrescope, to enhance the manoeuvrability of the camera. Because of the portability of the camera and its lower cost compared with PET and SPECT, CLI may provide supplementary radiology services in low-income countries.

Only one system for photoacoustic imaging (to image breast tissue) has been approved by the United States Food and Drug Administration. Lihong Wang and co-authors show that photoacoustic CT can be used to image the human brain. By using parallel ultrasonic transducers arranged hemispherically around the human head to enable a larger field of view, and by detecting acoustic waves stemming from both haemoglobin and deoxyhaemoglobin, the authors performed functional tomographic brain imaging (through the measurement of changes in oxygen saturation) in patients who had undergone a hemicraniectomy (in these patients, the removed bone flap provides an acoustic window that allows for better image reconstruction).

Rapid volumetric histological imaging of intact living tissue in real time could substantially reduce the duration of surgical procedures, as highlighted in a Perspective by Gooitzen van Dam and colleagues. Elizabeth Hillman and colleagues report two optimized microscope systems, one of which is compact, for light-sheet fluorescence microscopy. The researchers used a fast digital camera and a galvanometric mirror to scan the tissue along the direction perpendicular to the light sheet, allowing for roving image acquisition (each tissue section is illuminated by an oblique light sheet, and autofluorescence is detected through the same illumination lens). The compact microscope could be used intraoperatively to perform histology in real time, for both biopsied and intact tissue, as the researchers show for fresh biopsies of the kidney of a patient with chronic kidney disease (Fig. 1) and for oral mucosa in a healthy volunteer.

Fig. 1: Autofluorescence from a fresh biopsy of the kidney of a patient with chronic kidney disease.
figure 1

The image was stitched from 12 images of the biopsied tissue acquired via a benchtop system for light-sheet microscopy within approximately 3 minutes. Yellow and blue correspond to autofluorescence emitted at the approximate wavelengths of 525 nm and 618 nm, respectively. Scale bar, 1,000 µm. Reproduced with permission from the Article by Hillman and colleagues, Springer Nature Ltd.

In tissue microscopy, lensless systems would allow for large fields of view and for image refocusing. However, the low contrast provided by optically dense tissue has precluded lensless microscopy in vivo. Jacob Robinson and colleagues now show that this is possible. They report a prototype lensless microscope and the use of a bespoke optical phase mask (which generated diffraction patterns with high-contrast contours) to image calcium dynamics in the cortex of live mice and the microvasculature in the oral mucosa of human volunteers. The small form factor and the large field of view of systems for lensless microscopy may be beneficial for imaging vasculature in difficult-to-reach regions of the body.

By repetitively localizing a small number of microbubbles with micrometrical precision (the microbubbles act as subwavelength sources), ultrasound localization microscopy bypasses the diffraction limit of sound waves (which caps the resolution of conventional ultrasound) and can thus provide high-quality haemodynamic maps of the vasculature. To facilitate faster imaging with this super-resolution ultrasound technique, Baptiste Heiles and colleagues have benchmarked algorithms for the localization of the microbubbles and the rendering of their trajectories.

Of course, performance comparisons across imaging modalities or across preclinical and clinical uses are rarely absolute. Yet the standardization of the technical validation of instrumentation, as discussed for optical imaging modalities by Sarah Bohndiek in a Perspective article, can foster clinical adoption. In this regard, practical considerations also matter. For example, ultrasound systems can be portable, robust and relatively inexpensive, and hence widely available; yet the magnets and their enclosures required for MRI are bulky and expensive to run and maintain, and therefore are mostly available in large centres in developed countries. Overall, intrinsic physicochemical limitations, engineering trade-offs, clinical needs, regulatory constraints, intended uses and other factors can determine whether an imaging technology is advantageous for a particular application. Nevertheless, preclinical systems that provide performance advantages when imaging the human body or biopsied tissue to detect disease have a high chance to eventually make a clinical impact.