NEW YORK (360Dx) – Researchers at Germany's Ruhr University Bochum have developed an automated pathology tool based on Fourier transform infrared (FTIR) imaging.
Described in a study published this month in the American Journal of Pathology, the technology could enable higher-throughput pathological analyses while removing some of the subjectivity inherent in current pathology practice, said Klaus Gerwert, chair of biophysics at Ruhr University Bochum and senior author on the paper.
He added that he and his colleagues are currently moving the technique to a faster and more mobile platform based on quantum cascade laser (QCL) technology that could be used in settings including operating theaters for, for instance, assessing tissue margins during cancer surgery.
Gerwert said he also envisions using the approach to isolate cancer tissue samples for genomic and proteomic analysis, allowing clinicians to combine spatial pathological information with molecular data.
The researchers aim to launch a company to commercialize the technology and are currently looking for investors, he said.
Conventional pathology typically uses H&E staining to assess cellular features along with immunohistochemical assays to look at specific molecular markers.
The issue, Gerwert said, "is that this approach depends very much on the quality of the staining and the quality of the pathologist. If you have an excellent pathologist, there is no problem."
One the other hand, diagnoses can vary when "you have average pathologists working under time pressure," he said.
Additionally, an automated tool could help address the current shortage of pathologists in Germany and other countries, he said.
In FTIR, a sample is targeted by a beam composed of multiple frequencies of light with the different absorption of those different wavelengths at different points across the sample providing information on its composition. In the case of the pathology application Gerwert and his team are working toward, the IR spectrum provides an integrated signal reflecting the biochemical status of a given cell being examined.
The researchers then take these signals and develop a machine learning model that they train on samples stained via conventional pathology techniques to build a classifier they can then use on unstained samples to distinguish between cancer and normal tissues, as well as identify different classes and subtypes of cancers.
For instance, Gerwert said, the researchers have demonstrated that the approach can distinguish between lung cancer and normal tissue and then within lung cancer between small cell and non-small cell cancer and then distinguish between different adenoma subtypes in non-small cell lung cancer.
In the American Journal of Pathology study, the authors demonstrated that in bladder cancer the approach could distinguish between cystitis, low-grade carcinoma, and invasive high-grade carcinoma.
Gerwert added that in colorectal cancer, the researchers had been able to distinguish between tissue with low and high microsatellite instability.
He said that he envisioned the technology as something like a "driver assist" for pathology labs, where pathologists could use the method to run quickly through routine samples, freeing up time for them to focus on more difficult cases.
He also said he believed the tool could prove useful in operating rooms for real-time assessment of tissue margins during cancer surgery, though he noted that for this application it would be necessary to move the technique from a traditional FTIR platform to a QCL platform, which is much faster and more portable than FTIR.
An FTIR analysis like the one presented in the American Journal of Pathology study takes around 10 hours, while the equivalent analysis done using QCL takes around 20 minutes, Gerwert said. Additionally, QCL platforms are much more portable, he said, and don't require liquid nitrogen like FTIR.
"It is a small standalone instrument that can be easily used in an operating theater," he said, noting that thus far in his presentations to clinicians about the technology, he has received the most interest from surgeons.
Gerwert and his colleagues demonstrated the technique using QCL in a study published last year in Nature Scientific Reports in which they showed it could distinguish between stage II and stage III colorectal cancer tissue and normal tissue with a sensitivity of 96 percent and specificity of 100 percent compared to standard histopathology.
Because the technique is label-free and non-destructive, analyzed samples can then be used for genomic and proteomic analysis, Gerwert noted. In the American Journal of Pathology, the researchers used laser capture microdissection to extract the cancerous portions of the tissue they inspected and then used mass spec-based proteomics to identify potential bladder cancer markers. One of the markers they identified, the protein AHNAK2, distinguished between severe cystitis and bladder carcinoma in situ with sensitivity of 97 percent and specificity of 69 percent.
"Our goal at the end of the day is to provide the proteomic and genomic information along with the morphological data so that we have all three available to help clinicians with therapeutic decisions," Gerwert said. He said that he and his colleagues were currently working with pathologists at Ruhr University and in Cologne to demonstrate the technology's utility.
A number of academic and industry researchers are exploring new technologies to supplement existing pathology methods. Bruker, for instance, is working with pathologists to evaluate potential uses for its MALDI mass spec imaging systems, while companies like Akoya Biosciences and IONPath are also eyeing the pathology market with high-throughput tissue imaging systems.
In the surgical suite, companies including Olfactomics and Waters are developing devices that could help surgeons more quickly assess margins, as is the lab of University of Texas researcher Livia Eberlin, which this month published a study in Clinical Chemistry detailing the ability of its MasSpec Pen to distinguish between ovarian cancer and normal tissue.