Label-Free Observation of Micrometric Inhomogeneity of Human Breast Cancer Cell Density Using Terahertz Near-Field Microscopy

Research publication · Ex vivo terahertz near-field microscopy

Label-Free Observation of Micrometric Inhomogeneity of Human Breast Cancer Cell Density Using Terahertz Near-Field Microscopy

Kosuke Okada, Quentin Cassar, Hironaru Murakami, Gaëtan MacGrogan, Jean-Paul Guillet, Patrick Mounaix, Masayoshi Tonouchi and Kazunori Serita · Photonics · 2021 · Volume 8, issue 5, article 151 · DOI: 10.3390/photonics8050151

A malignant tissue region is not electromagnetically uniform. Cell-rich zones, fibrous stroma and other microscopic structures can coexist within a few hundred micrometres, yet conventional terahertz imaging averages them within a millimetre-scale spot. This study uses a scanning point terahertz source microscope to examine that local variability in an unstained, formalin-fixed and paraffin-embedded human breast cancer section. The system resolved a gold test pattern at about 10 micrometres and mapped an approximately 250 micrometre variation within invasive ductal carcinoma. The spatial changes correlated with nuclear density in histology. They constitute an ex vivo observation in one processed specimen, not a validated cancer classifier or a clinical biopsy method.

The experiment builds on a near-field geometry in which radiation is created at the sample plane. A femtosecond optical pulse is focused onto a 500 micrometre GaAs crystal, producing a localized terahertz transient through optical rectification. Since the tissue sits directly on the crystal, it interacts with the field before diffraction expands it to the free-space wavelength. A galvanometer moves the optical focus, and a synchronized low-temperature-grown GaAs photoconductive antenna detects the transmitted or reflected waveform. Spatial resolution is therefore set primarily by the pump focus and source-sample arrangement rather than by an ordinary terahertz lens.

The tissue had undergone fixation, dehydration and paraffin embedding. Those steps stabilize a thin section and reduce the strong influence of liquid water, making structural differences easier to investigate. They also alter lipids, hydration and geometry. The specimen included invasive ductal carcinoma, ductal carcinoma in situ and fibrous tissue over short distances. A stained pathological section provided the reference for tissue composition and nuclear distribution. The terahertz map and histology were compared as corresponding representations, but the physical preparation means that signal levels cannot be assumed to match fresh, living or newly excised tissue.

Featured visual: Image 1 from the Airtable record associated with this publication. Consult the original paper for the authoritative figure caption and interpretation. Source publication.

Visuals are drawn from the Airtable research archive. Figure numbering, rights and interpretation should be checked against the original publication before republication outside this site.

Mapping local amplitude in transmission and reflection

Before examining tissue, the researchers imaged a gold interdigitated structure with 9 micrometre features on GaAs. Clear separation between metal and substrate supported an effective resolution close to 10 micrometres. The breast section was then scanned over a field of roughly 280 by 280 micrometres, a small window chosen to include distinct microstructures. A bare region of the GaAs substrate supplied the reference response. Time-domain peak fields and spectra were converted into local amplitude maps, with histogram equalization used to make subtle variations more visible in some transmission images.

Transmission was weaker in the cell-dense malignant regions than in the fibrous background. Within the invasive carcinoma, a central area about 250 micrometres across differed from neighbouring zones. The authors divided the lesion into three regions, labelled IDC1 to IDC3, and compared their average pixel values and spread. DCIS produced the strongest attenuation, IDC1 and IDC2 were intermediate, IDC3 transmitted more strongly, and fibrous tissue gave the largest signal. The ordering was not inferred from colour alone; averaged measurements were compared with the distribution of nuclei visible in the pathological reference.

Reflection provided a complementary contrast. Malignant areas generally returned a larger field than fibrous tissue, and the selected subregion within the invasive carcinoma again differed from the more densely cellular zones. Because surface and paraffin effects influence transmission and reflection differently, agreement in the spatial trend gave additional support to the observation. The spectra did not reveal a unique narrow absorption line that identified a tissue type. Contrast extended broadly across the measured band and is better interpreted as a difference in effective dielectric response, attenuation and interface reflection.

Relating terahertz contrast to tissue architecture

Histology showed the highest nuclear packing in DCIS, followed by denser portions of invasive carcinoma, the less dense IDC3 region and fibrous tissue. Nuclei and tightly packed cells change the local refractive and absorptive environment, so the authors associate greater terahertz attenuation with greater cellular density. That relationship is physically plausible under the study conditions, but cell density was not manipulated independently. Section thickness, paraffin distribution, extracellular matrix and registration error remain possible contributors. The article therefore demonstrates a correlation with annotated morphology rather than a causal calibration that converts amplitude directly into a cell count.

The phrase “label-free” refers to the terahertz acquisition: the examined section did not require a stain to generate contrast. Histological staining was nevertheless essential as the reference used to identify tissue structures and interpret the map. The method did not replace pathology in this study. Nor did it test sensitivity and specificity across patients, benign mimics or multiple tumour subtypes. A field of view measured in hundreds of micrometres also captures only a small fraction of a surgical specimen, and close contact with the GaAs emitter imposes mechanical constraints.

Future validation would need many independently annotated sections, predefined analysis procedures and explicit control of preparation thickness and hydration. For fresh specimens, water absorption may dominate the signals observed here, while uneven surfaces could disturb contact and reflection. Instrument development would have to expand the field without losing resolution, raise acquisition throughput and quantify repeatability across repeated placements. Statistical or machine-learning analysis could be considered only after a sufficiently diverse data set exists; it cannot compensate for narrow sampling.

The scientific contribution is a more precise question for terahertz biophotonics. Instead of asking only whether averaged malignant and benign tissue differ, the paper shows that the response can vary within one invasive lesion at a scale of about 250 micrometres and that this variation follows microscopic cellular architecture in the reference section. The collaboration between terahertz laboratories and pathology makes that comparison possible. Under these controlled ex vivo conditions, SPoTS microscopy reveals intralesional heterogeneity that far-field systems would blur, while leaving clinical interpretation and generalization for subsequent studies.

This publication also connects with our broader work on biomedical imaging research and terahertz imaging.

Bibliographic reference