Development of digital pathology

Conventional histological methods achieve image contrast using stains, such as hematoxylin and eosin (H&E). This approach can deliver highly specific results; however, it also requires using chemicals during sample preparation that can prolong preparation time and induce artifacts, delaying pathological diagnosis. Traditional histopathology techniques also require thin tissue sections, usually under 5 μm in thickness. As such, the development of label-free microscopy seeks to reduce sample preparation time further, eliminate the use of harsh chemicals, and allow thicker tissue sections to be visualized.

Several label-free microscopy methods have been employed to try and circumvent these challenges¹. Nonlinear optical microscopy uses nonlinear interactions between light and matter to generate images. However, the long dwell time to collect weak nonlinear signals impedes high-speed image scanning and rapid identification of pathological regions. Optical coherence tomography has shown promise but provides limited imaging of subcellular features due to speckle noise and spatial resolution. Another candidate is quantitative phase imaging, with accelerated imaging speed, though it only provides two-dimensional information of thin tissue samples¹.

The best digital histology candidate to date is optical diffraction tomography (ODT), a 3D label-free quantitative phase imaging technique that can generate volumetric imaging information. ODT reconstructs a tissue sample’s 3D refractive index (RI) from its scattered field images obtained with various illumination angles. As the value of the RI depends on intracellular molecules, including proteins and lipids, ODT allows label-free quantitative 3D morphological mapping of biological specimens and has been widely utilized to advance our understanding of the physiology of various live cells¹. To date, ODT has struggled with imaging thick tissue slides; however, a recent publication has overcome this limitation by using digital refocusing and automated stitching. The experimental setup enabled mesoscopic imaging of various pathologic tissues over a millimeter-scale with submicrometer resolution¹.

ODT overcomes the limitations of traditional histology

Contrary to other histological imaging techniques, which use labeling agents and stains to produce contrast, optical diffraction uses the sample’s refractive index, an intrinsic optical parameter, to create imaging contrast¹. Thus, ODT uses RI to create a 3D image by combining multiple 2D quantitative phase images acquired from different illumination angles. Off-axis holography is used to acquire the scattered light transmitted through the specimen, with multiple scattered fields obtained from different angles, then combined to reconstitute the sample RI. In addition, RI values can vary based on the type of intracellular molecules being imaged and their number. This allows ODT to provide detailed structural and quantitative insight into various unstained biological samples, including live cells.

This approach works well for thin histology tissue sections, similar to traditional methods. To extend the use of optical diffraction tomography for thick tissue specimens, Hugonnet et al. employed digital refocusing and automated stitching to obtain accurate volumetric imaging information¹. Digital refocusing is a technique that generates images of different depths, exploiting the fact that an image is a 2D integral projection of a 4D light field. Imaging stitching is the process of combining multiple images with overlapping fields of view to produce a single segmented panoramic, high-resolution image. This new microscopy method allows subcellular and mesoscopic structures to be visualized using a high resolution in combination with a wide field of view, even for unstained thick tissue specimens.

ODT allowed the researchers to visualize individual cells, multicellular tissue architectures, and different morphological features with an accuracy comparable to that of conventional H&E staining. However, as the technique did not require lengthy staining procedures and can be done on thicker tissue sections, it reduced or eliminated the need for laborious traditional tissue processing and staining methods. As such, it enabled pathologists to visualize different morphological features in various tissues, allowing for quicker recognition and diagnosis of precursor lesions and pathologies, reducing overall time to diagnosis.

Future perspectives

Label-free volumetric holography holds great potential for rapid and high-resolution histopathology of thick tissue sections, bypassing the need for time-consuming tissue processing and chemical staining protocols. As this technique can be used with various types of tissue samples, it may allow pathologists to provide something of a real-time cancer diagnosis during intraoperative pathology consultations. In addition, digital pathology can enhance the analysis of multiscale volumetric images, which take considerable time using conventional methods.

Holographic histology, in its current iteration, still has limits. This technique can only be used to obtain high-quality images from tissues with a thickness of up to ∼100 μm. Even the latest digital refocusing techniques cannot overcome the low-quality tomographs obtained from thicker tissues due to multiple light scattering. Further research is required to optimize the protocols for sample preparation, improve reconstruction speed, and minimize artifacts resulting from multiple scattering. Advances in hardware and software may also further improve the performance of the method. Newly emerging machine learning methods for segmentation and AI software can also help improve the image analysis and reconstruction process.

In addition, more research is required on how RI information is interpreted histologically. As this is still a new field, data may potentially be interpreted differently by various pathologists. However, as RI tomographs generate diverse data from that obtained using traditional H&E staining, they can be utilized as complementary results during analysis. Applying both methods can significantly improve time to diagnosis, reducing the time needed to initiate treatment, which in turn can improve patient outcomes.