Imaging: gearing up for the next big reveal

Diagnostic imaging is ready to take on the next challenge
Diagnostic imaging is ready to take on the next challenge: to deliver precision health through earlier detection of disease

The sepia-coloured photo from an experiment conducted on a long-ago December in 1895 by Wilhelm Röntgen may not be much to look at now. But its shadowy metacarpals and indistinct phalanges presented a momentous discovery. Using a new kind of energy, the photo provided the first glimpse into the living human anatomy—in this instance, the left hand of Röntgen’s wife, Anna-Bertha. This earned Röntgen a Nobel Prize six years later and gave rise to a new discipline in medicine: diagnostic imaging.

Imaging has come a long way since Röntgen’s first crude two-dimensional x-ray. In the mid-1950s, x-rays were joined by ultrasound technology. Nuclear medicine soon emerged in the 1960s to pinpoint cancer hotspots in the lung. Over the next two decades, the field exploded with the debut of computed tomography (CT), positron emission tomography (PET), magnetic resonance imaging (MRI) and combinations thereof. Diagnostic imaging is now three-dimensional and more detailed. Its hardware has shrunk and its software is more powerful. Most importantly, it is more affordable, accessible, and safer than ever before.

“We have reached the point where we are able to image very, very fine structures very very, well, and in my honest opinion, I think we are reaching the limits of how small we are really going to want to be able to image,” said Dr Charles Goh, a consultant with the Department of Nuclear Medicine and Molecular Imaging at the Singapore General Hospital (SGH).

But far from spelling the end for innovation in imaging, the field is ready to take on the next set of challenges: to first deliver precision medicine and then to support precision health through the earlier detection of diseases.

“Exciting new therapies hold promise as being curative and perhaps even preventative. However, these treatments can be prohibitively expensive and drive-up health care costs. What we need are better biomarker-driven tests to help us make more informed medical decisions,” said Ann-Marie Chacko, an assistant professor from the Cancer and Stem Cell Biology Programme.

“The holy grail are disease prevention strategies, where we can identify preconditions and intervene to maximise health outcomes and cost-effectiveness,” she explained.

Take a polyp for example. Long before a physical polyp manifests itself in the colon, a molecular or genetic change would have triggered a cascade of events too small to observe on a structural level.

“To capitalise on the power of non-invasive imaging beyond anatomical visualisation, we need to overlay molecular information from sensitive diagnostic probes that capture early disease or pre-disease states that are identified through early and regular screening,” said Chacko.


Decrypting the body’s enigma code

To make inroads on this ambitious plan, Chacko is starting with cancer, where molecular signatures give insights on what therapies can be most effective in treating a tumour.

Harnessing the vast library of cellular markers that are unique to different cancer subtypes and tumour microenvironment, Tan Kai Wei, a fifth year MD-PhD student in Chacko’s lab, is determined to develop a probe that can accurately detect who is most likely to benefit from an expensive yet highly effective new type of cancer therapy known as immunotherapy.

He hopes that a scan will overcome the limitations imposed by the current gold standard, which is a biopsy. “Tumours are heterogeneous by nature. Biopsies are limited by the amount of tissue sampled and can be skewed by sampling bias” said Tan. “This probe, however, can be detected in whole body imaging allowing us to scan the entire body and overcome the sampling bias.”

This kind of imaging-guided approach is already starting to make a difference in the clinic. For example, adding a tracer that lights up cancer’s glucose uptake allows doctors to identify the most metabolically active node to biopsy in a suspected lymphoma patient.

The binding of the peptide probe (cyan) to the protein target human Programmed Death Ligand 1 (orange) allows identification of cancer patients for immunotherapy based on their PD-L1 expression.

“In the past, doctors used to look for the largest node to biopsy because that’s all we knew,” said Goh. “But sometimes large nodes may not be where the cancer is most easily identifiable because they contain tissue that has already died.”

To boost the development of novel probes, Chacko has assembled a multidisciplinary team of leading scientists across Singapore to create the Cancer ImmunoTherapy Imaging (CITI) programme. CITI’s task is the efficient translation of novel in vivo diagnostic imaging probes from discovery to clinic by bringing together key academic, clinical, industry and government stakeholders.

“The goal is to develop novel diagnostics to image the immune system, especially in the context of cancer. But these probes will not be limited to cancer as the immune system is the underpinning of a variety of diseases including infectious diseases, cardiovascular diseases and neurological disorders,” said Chacko.

She is also planning to use the probes developed from the CITI programme to better understand how immunology plays a role in dengue infection. As there are currently no drugs approved for dengue, she hopes to use imaging to monitor response to therapy.


Boosting counting power

While scientists like Chacko are working in the lab to find new signatures that help to decrypt the body’s internal messages, clinician scientists like ophthalmologist Wong Tien Yin are focusing on automating how images are processed.

For the last two decades, Wong has been working on a large-scale project where the blood vessels in the retina are imaged to detect various blinding eye conditions, such as diabetic retinopathy, glaucoma and age-related macular degeneration.

His latest innovation is the addition of a deep-learning artificial intelligence (AI) software system that was developed to help with the analysis. The system was greenlighted for clinical use in October 2019.

“Technology [like this] will replace transient, subjective and descriptive assessments from a physician with objective, quantifiable and retrievable data,” said Wong.

Anatomical pathologist Dr Cheng Chee Leong agrees. With the digitisation of microscopy and application of imaging analytics, pathological investigations can potentially become more objective, precise and reproducible.

“The digital imaging of histological slides has allowed us to unlock the image that a single pathologist sees under the microscope and turn it into a portable, sharable asset whose contents, such as morphological and cellular features, are amenable to computer analysis,” said Cheng, a senior consultant with SGH’s Department of Anatomical Pathology.

And he is very aware that this presents a unique opportunity. With relatively few histological slide imaging analytic studies done in Asian populations, Singapore could help to set the standards.

“What we do need to do is to actually be very active in the digitisation of histological slides so that we can put ourselves at the forefront,” added Cheng.

With digitisation in place, images will be ready for an algorithm to analyse them, an advance that will produce results much faster and with similar accuracy than a human specialist can. For example, to assess the damage to the heart muscle from a heart attack on an MRI, a trained reader may require about 45 minutes to one hour. “But a well-validated algorithm could do it in minutes,” said Derek Hausenloy, a professor with the Cardiovascular and Metabolic Disorders Programme at Duke-NUS.


The final, but steep, climb to wide-spread adoption

To fulfil the potential of imaging as seen by researchers like Chacko, a significant amount of work is still required. Supplanting or even complementing existing technologies requires a greater understanding of the clinical implications of the information encoded in an image.

“In order for a digital image [technology] to qualify as diagnostic, we need to ensure that we have the skills and understanding of how to validate these images compared with what we see using our more than a 100-year-old approach of looking through the microscope,” said Cheng. “We have to ensure that the information is on a par.”

Even if the information is on a par, concerns about patient privacy and data security also remain to be resolved because the images would need to be tagged with the necessary clinical information to be an effective resource for developing new protocols.

“We are still working out how to do this securely and safely,” said Dr Charles Goh. “Properly anonymised and annotated datasets are still only a tiny fraction of the imaging data that is out there.”

Another obstacle to the wide-spread use of imaging as a screening tool is cost to the public purse. Unlike treatments that can be billed to those who needed them, scanners and other imaging equipment are not only costly but also require dedicated space, maintenance and manpower.

And then, there’s the health cost of screening, both in terms of the anxiety and unnecessary interventions triggered by a false alarm as well as the exposure to radiation.

“If there is radiation exposure, then there is a real cost in terms of the potential for secondary malignancy, especially with screening that is done at the population level,” said Goh.

“We need to learn what is normal and what is not. So, it is not just the harm of the image itself, but also about the unnecessary interventions and patient anxiety,” added Goh.


Many ways to the top

With the help of AI, some of the challenges are already being addressed. For example, AI could shorten the duration of scans and reduce the accompanying exposure to radiation by improving the image quality of low-dose images through artefact reduction and noise suppression.

“We can use AI reconstruction techniques so we get readable images at lower doses of radiation,” said Goh.

Another way to overcome some of the barriers is through combining scanning modalities to glean more information from a single scan. For example, combining an MRI with PET. The MRI provides the fine detail of the tissue while the PET picks up the signal of a particular tracer. When viewed in a combined image, it provides insights on precisely where the disease is manifesting and how the surrounding tissues are responding.

“This is a nice combination of technology that can give new insights into heart disease and heart failure,” said Derek Hausenloy.

As well as combining technologies, combining imaging tools with the development of a therapeutic early on can also make an attractive value proposition.

“The biggest value of imaging is not just in diagnosis but also in driving the development of new therapies with the greatest safety and efficacy,” said Chacko. “If you do this rationally and efficiently, if you image wisely, we can drive down healthcare costs and best impact our patients.”