Bone cancer diagnosis by Raman microscopy

Pathologists have an unenviable job. Armed with just their eyes, their experience and a pinky-coloured picture they must decide whether a patient requires life-saving surgery, has nothing to be concerned about or just needs to be kept an eye on.

No matter which part of your body might have started behaving badly, a biopsy is the most likely procedure to reveal a potential problem or give the reassuring all-clear. Once the biopsy sample has been taken it is preserved, thinly sliced and then typically stained with H+E (hematoxylin and eosin) stain. When an experienced pathologist looks at a microscope image of a good quality stained sample, they are extremely successful in classifying the tissue and providing an accurate diagnosis.

Even though H+E staining has been used for more than 100 years to reveal the microscopic structure of different tissues, there is no single protocol or set of chemical reagents to produce identical results time and time again. Each pathology lab technician will have their own preferred way of preparing stained biopsy samples. The introduction of digital imaging of the view down the microscope has made it easier and quicker for a pathologist to assess a particular biopsy sample but variability in staining persists and can make it challenging to classify different cell morphologies. Staining is cheap and widely used but can be time consuming and does not always give pathologists the quality of images they need.

For these reasons, researchers have been developing Raman microscopy as an alternative pathology tool. Dr Mario D’Acunto at the IBF-CNR, Istituto di Biofisica and colleagues at the University of Pisa recently reported a pilot study of patients with bone tumors. Publishing in Nature Scientific Reports, they describe how tumors can be classified by Raman microscopy without the need of stains or time-consuming sample preparation. Using statistical methods on Raman microscope images of biopsy samples, they obtained 90% sensitivity, 90% specificity and 90% accuracy in classifying tumor types. This is comparable with the performance of expert pathologists assessing the best quality stained images.

So how does Raman microscopy work and what are the statistical methods D’Acunto et al used? In Raman microscopy a laser is scanned over the surface of the biopsy sample and the light scattered by the sample gives a kind of fingerprint of the different molecules present at each point in the image. The fingerprint is known as its vibrational spectrum. Information about even subtle changes in the way molecules are joined together can be extracted from individual points in the image. It is also possible to use statistical methods to extract similarities and differences between the vibrational spectra of the whole Raman image. These similarities are known as Principle Components and by carefully comparing the relative amounts of different Principle Components of each type of biopsy, it was possible to classify different tumor types.

Accurate bone tumor classification based on biopsy is crucial. Benign EC (enchondroma) tumors are not concerning but CS (chondrosarcoma) tumors account for around 20% of malignant bone tumors and must be detected. Of increasing concern are CS1 (chondrosarcoma 1) which rarely metastasize (spread) and CS2 and CS3 which can metastasize in up to 70% of cases. Catching CS type tumors quickly is therefore important and correctly differentiating CS2 and CS3 from CS1 types is crucial so that effective treatment can be prioritised.

Raman microscopy therefore looks to have an important role to play in the future of pathology and, cancer diagnosis and treatment. Combined with statistical methods and new artificial intelligence tools, major advances in the 100 year old science of tissue imaging are knocking at the path lab door.

The whole journal article can be found at https://www.nature.com/articles/s41598-020-58848-0 . A practical introduction to H+E staining can be found on the Leica microscope website.