Fast and effective routes to cancer diagnosis have never been more needed. Thanks to COVID-19, hospitals are burdened by a huge backlog of routine procedures. The scale of the impact of the pandemic on hospital care, and in particular on cancer treatment, is now becoming increasingly apparent.
We need innovative strategies to stratify patients’ risk of cancer and to prioritize patients for diagnostic investigations – and here’s where technologies incorporating high-level artificial intelligence (AI) could play a key role. AI applications in healthcare have progressed rapidly in the past few years and new, innovative ways of implementing AI are starting to make a real difference within diagnostics. These methods are already being used across the world; for example, AI now assists with detecting lung cancer – one of the most common cancers – from CT scans (1).
AI applications also have an important role in supporting the diagnosis of rare cancers, such as brain cancers – a traditionally difficult task. Patients most often present to primary care with nonspecific symptoms indicative of more probable non-cancer diagnoses. Referring every patient for expensive brain scans is neither possible nor cost effective. The best-performing symptom-based referral guidelines for suspected brain tumor only expect to identify a brain tumor approximately 3 percent of the time (2), so developing translatable technology that can be implemented within the clinic to improve triage for brain imaging is a major unmet need. Because smaller tumors are more often and more easily managed surgically, with less harm to the patient, early cancer detection is a key goal for improving patient outcomes.
Spectroscopic liquid biopsy is an innovative strategy for assessing blood samples – and, because it is quick and cost-effective, it could be a major game-changer in the diagnosis of cancer and other diseases. Blood samples are readily available and convenient for patients, so can be ordered earlier than current diagnostic pathways in the investigation of new-onset nonspecific symptoms. The low-cost technology, based on the interaction of infrared light with molecules present in the patient sample, generates a biological signal which can then be classified using an AI algorithm to detect cancer. In the brain tumor population, this allows the detection of disease within a symptomatic population – identifying which patients need urgent imaging and which do not.
Advances in AI have allowed us to maximize the opportunity that computational approaches offer for the detection of cancer and other diseases. If the technology is harnessed appropriately, spectroscopy-based liquid biopsy and AI have the potential to not just triage patients effectively, but ultimately increase survival rates and improve quality of life.
Newsletters
Receive the latest analytical science news, personalities, education, and career development – weekly to your inbox.
References
- E Svoboda, “Artificial intelligence is improving the detection of lung cancer,” Nature, 587, S20 (2020). PMID: 33208974.
- K Zienius et al., “Direct access CT for suspicion of brain tumour: an analysis of referral pathways in a population-based patient group,” BMC Fam Pract, 20, 118 (2019). PMID: 31431191.