Biomedical Imaging

1936–1946 · 1 of 8

X-ray Driven Biomedical Imaging

1936–1946

X-ray Driven Biomedical Imaging

Contemporary themes

X-ray driven imaging and spectroscopy have coalesced into the core biomedical toolkit, enabling radiography, diffraction, particle sizing, and spectral analysis to visualize structure and quantify radiation effects. Fluorescence-based microscopy introduces molecular-level contrast and paves the way for later high-resolution and targeted imaging. Early computational ideas and illumination methods stimulate digital image computation for cardiac and vascular measurements, while electroencephalography-based approaches advance functional brain mapping alongside anatomical observation, and chromosomal imaging under X-ray exposure begins to document cytogenetic phenomena.

  • X-ray–driven imaging and spectroscopy emerge as the core biomedical imaging toolkit, spanning radiography, diffraction, particles sizing, and spectral analysis to visualize structure and quantify radiation effects.
  • Neuroimaging advances consolidate functional and anatomical brain imaging through electro-encephalography for tumor localization, direct brain observation, and sub-cortical tumor study, signaling a shift toward imaging-based neurology.
  • Chromosome biology with X-ray exposure crystallizes a cytogenetic imaging paradigm, documenting chromosomal aberrations, mitotic chromosome effects, and chromosomal typologies across plant models and mammalian-like systems.
  • Fluorescence-based microscopy establishes a molecular-imaging foundation, with fluorescence microscopy in biology and high-power variants enabling visualization of fluorescence pathways at the cellular level.
  • Early computational imaging and instrumentation ideas surface in illumination methods, photoelasticity theory, and digital image computation for cardiac imaging and vascular measurements.

Influential works

The period establishes a unifying paradigm at the intersection of structure, function, and genetics, laying the groundwork for modern imaging modalities. The development of electroencephalography-based tumor localization and brain mapping provides early demonstrations of non-invasive functional imaging, influencing diagnostics and neurosurgical guidance. Fluorescence microscopy and chromosomal imaging seed a molecular-imaging era that would inform subsequent cellular, ophthalmic, and radiographic cytogenetics research.

  • Fluorescence Microscopy in Biology (1940) introduced fluorescence-based imaging in biology, enabling selective visualization of labeled structures with higher contrast than brightfield. It catalyzed modern cellular imaging and underpins later confocal and multi-photon techniques.
  • Atlas of Electroencephalography (1941) standardized brain electrical activity maps, enabling recognition of patterns linked to epilepsy, tumors, and lesions. It established a reference framework for functional brain imaging and influenced neurodiagnostics and research for decades.
  • Topography of the Layer of Rods and Cones in the Human Retina (1937) mapped photoreceptor distribution, linking structure to function and informing early retinal imaging concepts. This work provided anatomical benchmarks that guided subsequent visual-imaging advances in ophthalmology.
  • The location of cerebral tumours by electro-encephalography (1936) demonstrated non-invasive localization of tumors via EEG signals, inaugurating functional brain imaging concepts and shaping later imaging-guided neurosurgical approaches.

1947–1974

Computed Tomography Emergence

Contemporary themes

During 1947–1974, biomedical imaging was reshaped by the merger of computation, physics, and instrumentation to enable noninvasive visualization. The development of computable tomographic reconstruction unlocked three-dimensional visualization from radiographs and cross-sectional data, with algorithmic reconstruction approaches complementing or replacing traditional Fourier methods. At the same time, noninvasive modalities such as magnetic resonance imaging and ultrasound expanded in vivo functional assessment, while electron microscopy, autoradiography, and staining techniques supplied crucial cellular and tissue detail across imaging scales.

  • Computational imaging and three-dimensional reconstruction emerged as a core paradigm, uniting tomographic system descriptions with algorithmic reconstruction (convolutions replacing Fourier approaches) to enable volumetric visualization from radiographs and cross-sectional data.
  • Electron microscopy and autoradiography became central tools for cellular and tissue imaging, enabling cytochemistry, high-resolution autoradiography, and electron microscopy studies of collagen, neurons, and vascular beds across organs.
  • Noninvasive functional imaging modalities gained prominence for diagnostic and physiological assessment, with early tumor detection by Nuclear Magnetic Resonance, ultrasound diagnosis, and proton radiography contributing to in vivo imaging capabilities.
  • Contrast generation and tissue labeling through staining, cytochemistry, fluorescence assays, and osmium-fixed tissue preparation enabled visualization across light and electron microscopy techniques.
  • Nuclear-based imaging modalities and tracer-enabled radiography broadened tissue characterization, spanning proton radiography, autoradiography with radiolabeled substrates, and magnetic resonance as a nucleus-spin imaging approach.

Influential works

Historically, these advances established computed tomography and magnetic resonance imaging as foundational pillars of modern diagnostic imaging, inaugurating widespread volumetric imaging and standardized functional assessment. The mathematical underpinnings—Fourier-based reconstruction for tomographic imaging—became foundational tools that propelled rapid clinical adoption and continued methodological refinement. By providing standardized anatomical frameworks and high-resolution tissue visualization, these advances set the stage for future multimodal imaging and translational neuroscience.

  • Computerized transverse axial scanning (tomography): Part 1. Description of system introduced the first practical CT scanner, detailing how X-ray attenuation data are digitized and reconstructed into cross-sectional body images; this innovation launched noninvasive diagnostic imaging and reshaped radiology.
  • The Fourier reconstruction of a head section formalized CT image reconstruction from projections using Fourier methods; it provided the mathematical backbone for turning attenuation data into cross-sectional images, catalyzing widespread clinical CT adoption.
  • Tumor Detection by Nuclear Magnetic Resonance demonstrated MR imaging's potential to distinguish tumor tissue via NMR signals, spurring rapid MRI development and establishing MRI as a cornerstone of modern noninvasive imaging.
  • A stereotaxic atlas of the rat brain offered a standardized 3D coordinate framework for rat neuroanatomy, enabling precise localization in imaging studies and surgical planning; its impact underpins countless neuroimaging experiments and translational neuroscience.

1975–1981

Quantitative Tomography Emergence

Contemporary themes

Computed Tomography emerged as the central cross-sectional imaging paradigm, unifying brain, chest, and body imaging and driving early clinical trials, aging studies, and vascular applications through CAT/CT advancements. Emission tomographic imaging matured toward quantitative brain and body imaging with attenuation correction and detector optimization, as illustrated by attenuation correction methods, early scanner designs, and whole-body imaging. Myocardial perfusion imaging with Thallium-201 became a major pattern in cardiology, and neuroimaging research leveraged tomographic methods to map cerebral glucose metabolism and brain atrophy criteria, informing diagnosis.

  • Computed Tomography (CT) emerged as the central cross‑sectional imaging paradigm, unifying brain, chest, and body imaging and driving initial clinical trials, aging studies, and vascular applications across CAT/CT advancements.
  • Emissions tomographic imaging matured toward quantitative brain and body imaging with attenuation correction and detector optimization, illustrated by attenuation correction methods, ECAT-type systems, and whole‑body PET design.
  • Myocardial perfusion imaging with Thallium-201 became a major pattern in cardiology, comparing rest/exercise perfusion and advancing multi‑pin-hole and tomographic techniques for regional perfusion assessments.
  • Neuroimaging and brain metabolism research leveraged tomographic methods to map cerebral glucose metabolism and structural atrophy, with FDG PET validation and brain perfusion/atrophy criteria informing diagnosis.

Influential works

These years solidified a unified quantitative tomographic framework that underpins modern positron emission tomography, single-photon emission computed tomography, and computed tomography imaging. Foundational breakthroughs in reconstruction, attenuation correction, and standardized quantitation shaped cross-modality imaging, enabling reliable comparisons across subjects and modalities. The FDG-PET validation study and metabolic-rate measurements established a template for brain and systemic functional imaging that catalyzed decades of research and clinical translation.

  • Image reconstruction from projections: the fundamentals of Computed Tomography (CT) introduced the core reconstruction framework for medical imaging, mapping projections to cross-sections via filtered backprojection, enabling 3D diagnostic imaging and inspiring decades of algorithmic advances
  • Principles of Computer Assisted Tomography (CAT) in radiographic and radioisotopic imaging laid the mathematical and computational groundwork for reconstructing images from projections, enabling cross-modality imaging and quantitative tissue assessment
  • A Method for Attenuation Correction in Radionuclide Computed Tomography introduced tissue attenuation correction in radionuclide imaging, reducing artefacts and bias and laying the groundwork for accurate SPECT and later PET quantitation
  • Quantitation in Positron Emission Computed Tomography (PET) established quantitative metrics and calibration for PET data, enabling cross-subject comparisons and metabolic-rate estimation, shaping standards for clinical and research imaging
  • Tomographic measurement of local cerebral glucose metabolic rate in humans with 2-[18F]fluoro-2-deoxy-D-glucose (FDG): Validation of method demonstrated a quantitative approach to brain metabolism using FDG, catalyzing widespread PET neuroimaging and metabolic studies

1982–1995

Functional Multimodal Imaging

Contemporary themes

During 1982-1995 Biomedical Imaging saw a transition toward functional and multimodal approaches. Functional magnetic resonance imaging emerged as the core paradigm for brain mapping, leveraging stimulus-evoked BOLD responses, intrinsic signal changes, and temporal analyses to delineate functional regions and networks. Multimodal image registration and atlas-based strategies enabled anatomical-functional integration across CT, PET, and MR data, linking structural maps with functional ROIs. Diffusion and perfusion imaging introduced physiologic imaging pathways to probe microstructure and blood flow, while automated image analysis and tissue characterization advanced quantitative morphometry. Optical and nanoscale modalities complemented MR technologies, broadening the imaging toolkit with optical spectroscopy, near-field methods, and nanoscale imaging.

  • Functional MRI emerged as the core paradigm for brain mapping, combining stimulus-evoked BOLD responses, intrinsic signal changes, and time-course analyses to delineate functional regions and networks in the human brain.
  • Multimodal image registration and atlas-based approaches enabled anatomical-functional integration across CT, PET, and MR data, linking structural maps with functional ROIs through automated alignment and region-based atlases.
  • Diffusion and perfusion imaging laid out physiologic imaging strategies to probe tissue microstructure and blood flow, introducing intravoxel incoherent motion concepts, diffusion anisotropy, and perfusion imaging as biomarkers.
  • Automated image analysis and tissue characterization advanced quantitative brain morphometry, with intensity normalization, segmentation by probability/connectivity, and feature extraction for robust morphologic assessment.
  • Optical and nanoscale imaging modalities complemented MR-based imaging, introducing optical stethoscopy, time-resolved tissue optics, near-field microscopy, and atomic force microscopy as novel biomedical imaging tools.

Influential works

The period produced pivotal breakthroughs that shaped the trajectory of modern imaging. Brain imaging with blood-oxygenation-level-dependent contrast (fMRI) catalyzed functional neuroimaging and cognitive neuroscience. Diffusion tensor imaging formalized mapping of white matter pathways. Optical coherence tomography provided micrometer-scale cross-sectional tissue imaging, while super-resolution approaches such as stimulated emission depletion broke diffraction limits, enabling nanoscale cellular imaging. Two-Photon laser scanning microscopy introduced deep-tissue imaging with reduced photodamage, transforming live-cell and neuroimaging. Collectively, these innovations established an integrated, multiscale imaging paradigm that underpins contemporary biomedical imaging and informed the design of multimodal pipelines and quantitative biomarker development.

  • Optical Coherence Tomography (1991) introduced OCT, a noninvasive interferometric method that achieves micrometer-scale, cross-section imaging of tissues. It rapidly transformed ophthalmology and spurred diffusion of high-resolution biomedical imaging across disciplines.
  • Breaking the diffraction resolution limit by stimulated emission: stimulated-emission-depletion fluorescence microscopy (1994) demonstrated a fluorescence technique overcoming the diffraction limit with stimulated emission, enabling nanoscale optical imaging in cells and laying the groundwork for modern super-resolution fluorescence methods.
  • Brain magnetic resonance imaging with contrast dependent on blood oxygenation (1990) introduced BOLD fMRI, linking neural activity to hemodynamic signals, catalyzing functional neuroimaging and the modern cognitive neuroscience era.
  • MR diffusion tensor spectroscopy and imaging (1994) formalized diffusion tensor imaging, mapping white-matter pathways noninvasively, enabling tractography and deep insights into brain connectivity.
  • Two-Photon Laser Scanning Fluorescence Microscopy (1990) introduced a nonlinear excitation approach enabling deep-tissue imaging with reduced photodamage, revolutionizing live-cell and neuroimaging.

1996–2002

Tensor-Based Diffusion MRI

Contemporary themes

Diffusion Magnetic Resonance Imaging emerged as the central paradigm for revealing white-matter microstructure and connectivity, enabling tensor-based tractography and quantitative diffusion metrics. Concurrently, automated image analysis ecosystems and registration techniques strengthened multi-modal workflows, underpinning functional Magnetic Resonance Imaging designs and cross-modality alignment. Image preprocessing advances, including bias field correction and intensity normalization, improved reliability of structural and functional analyses in routine workflows.

  • Diffusion MRI emerged as a central paradigm for revealing microstructure and white-matter connectivity, enabling tensor-based tractography, white-matter morphometry, and noninvasive fiber mapping..
  • Computational MRI analysis ecosystems and automated registration techniques underpinned functional and multi-modal imaging, exemplified by AFNI, fMRI designs, and MR-PET/MR-CT alignment..
  • Image preprocessing advances targeted MRI bias field correction and intensity nonuniformity, improving reliability of structural and functional analyses..
  • Stroke-imaging paradigms combined diffusion and perfusion MRI for early detection, evolution tracking, and outcome prediction in ischemic events..
  • Multi-modal data fusion through cross-modality registration (MR-PET, MR-CT) supported integrated anatomical and functional interpretation..

Influential works

The period laid the groundwork for diffusion tensor imaging concepts and tractography that became foundational to subsequent connectomics research. It fostered a culture of open, tool-driven pipelines and cross-modality fusion, shaping reproducibility and multi-modal interpretation in later decades.

  • Visual molecular dynamics (VMD) introduced an integrated, extensible platform for visualizing and analyzing biomolecular structures and molecular dynamics (MD) trajectories, enabling scripting and a plugin ecosystem.
  • Analysis and Visualization of Functional Magnetic Resonance Neuroimages (AFNI) provided an extensible, open‑source suite for functional magnetic resonance imaging (fMRI) data analysis, visualization, and statistical modeling, promoting reproducibility in neuroimaging workflows.
  • In vivo fiber tractography using diffusion tensor magnetic resonance imaging (DT-MRI) data enabled reconstruction of white‑matter pathways from MRI, catalyzing connectivity studies and methodological advances in brain networking.
  • Diffusion tensor imaging (DTI): Concepts and applications offered a foundational framework summarizing diffusion metrics, modeling, and early neuroimaging uses that shaped subsequent literature in white matter mapping.
  • Voxel-based morphometry (VBM) — The Methods presented automated, whole‑brain statistical analysis of brain structure from MRI, launching a wave of morphological investigations and cross‑study comparability.

2003–2009

Multimodal Targeted Molecular Imaging

Contemporary themes

During 2003-2009, biomedical imaging coalesced around a multimodal, targeted paradigm, using ligand- or antibody-guided contrast to visualize biological processes in vivo across magnetic resonance imaging (MRI), optical, and nanoparticle-enabled readouts. Fluorescent nanoscale probes enabled high-resolution visualization of tumors and vascular dynamics, while engineered nanoparticles supported cross-modality platforms that link anatomical and functional contrast for surgical guidance. Quantitative imaging advanced through diffusion-weighted and diffusion spectrum imaging with uncertainty analysis, and novel contrast agents such as manganese oxide nanoparticles and gadolinium oxide nanoparticles pushed sensitivity and specificity in molecular imaging.

  • Targeted molecular imaging integrates ligand- or antibody-guided contrast to visualize biological processes in vivo, spanning magnetic resonance, optical, and nanoparticle-enabled readouts.
  • Fluorescent nanoscale probes such as near-infrared quantum dots and related dyes enable high-resolution, targeted visualization of tumors and vascular dynamics in living subjects.
  • Multimodal imaging platforms exploit engineered nanoparticles to couple anatomical and functional contrast (MRI/optical/photoacoustic), aiding surgical guidance and cross-modality interpretation.
  • Quantitative imaging advances rely on diffusion-weighted and diffusion spectrum MRI together with uncertainty analysis to improve tissue microstructure inference and diagnostic confidence.
  • Development of novel contrast agent designs (MnO nanoparticles, gadolinium oxide nanoparticles, artificial magnetic nanoparticles) pushes sensitivity and specificity for molecular imaging in vivo.

Influential works

These developments built on cross-period innovations, establishing scalable pipelines for fast, quantitative imaging and cross-modal integration. Pioneering approaches in sparse MRI with compressed sensing enabled rapid acquisitions from undersampled data and spurred broader adoption across modalities. Diffusion MRI progress, including tractography and automated brain parcellation, standardised white-matter mapping and cortical localization, shaping future brain-wide connectivity studies.

  • Sparse MRI: The application of compressed sensing for rapid MR imaging introduced compressed sensing to MRI, enabling accurate reconstruction from undersampled data and dramatically faster scans; catalyzed widespread adoption of CS in MRI and spurred cross-modal imaging research.
  • A diffusion tensor imaging tractography atlas for virtual in vivo dissections presents a tractography atlas to standardize white-matter pathways across subjects, enabling reproducible in vivo dissections and guiding neurosurgical planning.
  • An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest automates gyral-based cortical parcellation, providing scalable ROIs and improving cross-study comparability in functional and structural imaging.
  • A robust and high-throughput Cre reporting and characterization system for the whole mouse brain enables systematic mapping of neural populations across transgenic lines, accelerating circuit-level genetics and brain-wide connectivity studies.
  • Characterization and propagation of uncertainty in diffusion‐weighted MR imaging analyzes and quantifies diffusion-weighted MRI uncertainty, guiding interpretation and enhancing robustness of diffusion metrics used in tractography.

2010–2016

Radiomics-Driven Deep Imaging

Contemporary themes

During 2010–2016, biomedical imaging merged deep learning, radiomics, and multimodal approaches to enable quantitative analysis and image-guided interventions. Molecular imaging integrated targeted probes with cross-modal performance across fluorescence, photoacoustic, and intraoperative imaging to visualize biological processes and steer therapy. Simultaneously, photoacoustic and optical imaging advanced depth and resolution using engineered contrast agents, while diffusion-based neuroimaging emphasized robust quantitative metrics and artifact mitigation to map brain structure and networks; theranostics and image-guided therapy linked imaging with real-time treatment monitoring using activatable nanomaterials.

  • Molecular imaging integrates targeted probes with multiple modalities to enable simultaneous visualization and therapy in vivo, spanning design of contrast agents, pharmacokinetics, and cross-modal performance across fluorescence, photoacoustic, and intraoperative imaging.
  • Photoacoustic and optical imaging advances push depth and resolution by leveraging engineered contrast agents and tissue chromophores for organ-to-cellular scale visualization, with high sensitivity and molecular specificity.
  • Neuroimaging and diffusion methods emphasize robust quantitative analysis and artifact mitigation, coupling diffusion theory with functional connectivity metrics to map brain structure and network dynamics.
  • Theranostics and image-guided therapy unify imaging and treatment, using PDT mechanisms and photoactivatable nanomaterials to monitor response and optimize interventions in real time.

Influential works

U-Net: Convolutional Networks for Biomedical Image Segmentation (2015) introduced a symmetric encoder–decoder with skip connections, enabling precise segmentation from limited data and spawning widespread 2D and 3D variants. ImageNet Classification with Deep Convolutional Neural Networks (2012) demonstrated deep learning's transformative power for large-scale vision, catalyzing rapid transfer learning adoption in biomedical imaging. V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation (2016) extended end-to-end three-dimensional segmentation to volumetric data; Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach (2014) introduced radiomics as a high-throughput feature toolkit for predicting biology and outcomes.

  • U-Net: Convolutional Networks for Biomedical Image Segmentation (2015) introduced a symmetric encoder–decoder architecture with skip connections that enable precise segmentation from limited data, catalyzing widespread adoption and enabling 3D variants and numerous clinical applications.
  • ImageNet Classification with Deep Convolutional Neural Networks (2012) presented AlexNet, a deep CNN that achieved breakthrough performance on ImageNet, proving deep learning's viability for large-scale vision and triggering rapid transfer learning adoption in biomedical imaging.
  • V-Net: Fully Convolutional Neural Networks for Volumetric Medical Image Segmentation (2016) adapted fully 3D CNNs to medical volumes, enabling end-to-end volumetric segmentation of MRI/CT data with efficient training and dice-based loss, shaping modern 3D medical image analysis.
  • Decoding tumour phenotype by noninvasive imaging using a quantitative radiomics approach (2014) framed radiomics as a high-throughput extraction of quantitative image features to predict tumor biology and outcomes, launching a new era of imaging-based precision oncology.

2017–2024

End-to-End Deep Medical Imaging

Contemporary themes

The period is defined by a shift toward end-to-end deep learning systems that automate detection and diagnosis across imaging modalities such as chest radiography, dermoscopic skin images, mammography, and brain imaging. These methods leverage large-scale public datasets and transfer learning to mitigate data scarcity, enabling rapid generalization across institutions and populations. Advanced segmentation and localization are dominated by multi-scale and three-dimensional architectures, including nested and densely connected U-Net variants, often augmented with probabilistic post-processing to refine organ and lesion delineation. The community also emphasizes transfer learning and domain adaptation to address emerging diseases and cross-modality translation, along with standardized benchmark datasets and evaluation tasks that drive reproducible progress. Contextual reasoning and post-processing stages begin to complement pure end-to-end models, integrating imaging patterns with clinical interpretability and radiomic phenotypes that support precision medicine.

  • Theme 1: Deep learning enables automated detection and diagnosis across imaging modalities (X-ray, skin, mammography, brain) via end-to-end learning on large datasets, as shown by ChestX-Ray8, Dermatologist-level skin cancer, Large-scale mammography CAD, Brain tumor imaging, and surveys.
  • Theme 2: Segmentation and localization with multi-scale and 3D architectures and CRF refinement—UNet-like and DenseUNet variants achieve accurate organ/tissue delineation across MRI/CT.
  • Theme 3: Transfer learning and domain adaptation for data-scarce medical imaging and emerging diseases, including COVID-19 X-ray detectors and translation-based domain adaptation.
  • Theme 4: Benchmark datasets and evaluation frameworks drive progress through standardized tasks and large-scale public data (ChestX-Ray8, LUNA16, CAD datasets).
  • Theme 5: Contextual reasoning and probabilistic post-processing improve segmentation accuracy via CRFs and context modeling in CNN-based pipelines.

Influential works

  • ChestX-Ray8 introduced a hospital-scale chest X-ray image database and benchmarks on weakly-supervised classification and localization of common thorax diseases, catalyzing the shift to large-scale radiology AI research and enabling standardized evaluation across hospitals.
  • Dermatologist-level classification of skin cancer with deep neural networks demonstrated dermatologist-level accuracy on dermoscopic cancer classification, accelerating clinical trust in AI and spurring vast dermatology AI research on image-based diagnosis.
  • UNet++: A Nested U-Net Architecture for Medical Image Segmentation proposed a nested skip-connection design that boosted segmentation accuracy with limited labeled data, becoming a go-to blueprint for medical image segmentation in academia and industry.
  • Mask R-CNN extended Faster R-CNN with a parallel mask prediction branch, enabling robust instance segmentation and precise delineation of anatomical structures and lesions, widely adopted as a modular framework in biomedical imaging challenges.
  • Computational Radiomics System to Decode the Radiographic Phenotype proposed automatic radiomic feature extraction and phenotype mapping from radiographs, catalyzing radiomics research and linking imaging patterns to clinical endpoints for precision medicine.