“Outfitted with deep neural networks, mobile devices can potentially extend the reach of dermatologists outside of the clinic. It is projected that 6.3 billion smartphone subscriptions will exist by the year 2021 and can therefore potentially provide low-cost universal access to vital diagnostic care.”
Then, in October, Japanese researchers reported using artificial intelligence to analyze endocytoscopic images of polyps and automatically identify colorectal cancer in real time.4 The system compared roughly 300 of a polyp’s features to more than 30,000 endocytoscopic images fed into it, and diagnosed its pathology in under a second.
Even IBM’s famed Watson system is moving from being a novelty game show act on Jeopardy! to a potent physician’s assistant, capable of digesting crucial data from a cancer patient’s report — including blood test results, tumor pathology and imaging information, and the nature of genetic mutations — and, after scouring reams of medical literature, yielding treatment recommendations tailored to a specific patient.
In that way, Watson helps solve the problem of information overload for practicing oncologists.
As Alessandra Curioni-Fontecedro, MD, of the division of oncology at University Hospital Zurich in Switzerland noted in a 2017 article, “It has been estimated that a physician should read 29 hours per working day in order to stay updated about new medical research. Moreover, every year, the medical literature increases by doubling the amount of information every 3 years.”5
Another use could be helping practitioners to identify more effective drug combinations for treating cancer. Two studies published last year reported on computational approaches developed by Weill Cornell Medicine researchers, which were able to predict the response viability of drug combinations that proved to be “in agreement with experimental data.”6
“We found that drug synergy and combinatorial effectiveness can be predicted from a relatively small subset of combinations based only upon single drug efficacies,” the authors of one of the two studies wrote. And, they added, “as additional large combinatorial screens become available, this methodology could prove to be impactful for the identification of drug synergy within the larger universe of possible drug combinations.”7
Processing vast amounts of cancer-related information may be one of the greatest strengths of artificial intelligence, Dr Bejnordi said, becoming increasingly valuable in the era of “big data.”
He expects the existing hesitation of practitioners to rely on artificial intelligence is likely to fade as they come to see it as a powerful diagnostic aid rather than as a threat.
“It’s definitely a tool for pathologists to use. It’s going to really lower the time it takes for them to read these images. And as pathologists see this potential I think there will be more openness and they will trust it more and, in the long run, make use of these algorithms.”
- Ehteshami Bejnordi B, Veta M, Johannes van Diest P, et al. Diagnostic assessment of deep learning algorithms for detection of lymph node metastases in women with breast cancer. JAMA. 2017;318(22):2199-210. doi: 10.1001/jama.2017.14585
- Tiwari P, Prasanna P, Wolansky L, et al. Computer-extracted texture features to distinguish cerebral radionecrosis from recurrent brain tumors on multiparametric MRI: a feasibility study. AJNR Am J Neuroradiol. 2016;37(12):2231-6.
- Esteva A, Kuprel B, Novoa RA, et al. Dermatologist-level classification of skin cancer with deep neural networks. Nature. 2017;542:115-8. doi: 10.1038/nature21056
- Mori Y, Kudo S, Misawa M, et al. Diagnostic yield of “artificial intelligence”-assisted endocytoscopy for colorectal polyps: a prospective study. Oral presentation at: 2017 United European Gastroenterology week; October 28-November 1, 2017; Barcelona, Spain.
- Curioni-Fontecedro A. A new era of oncology through artificial intelligence. ESMO Open. 2017;2(2):e000198. doi: 10.1136/esmoopen-2017-000198
- Du W, Goldstein R, Jiang Y, et al. Effective combination therapies for B-cell lymphoma predicted by a virtual disease model. Cancer Res. 2017;77(8):1818-30. doi: 10.1158/0008-5472.CAN-16-0476
- Gayvert KM, Aly O, Platt J, Bosenberg MW, Stern DF, Elemento O. A computational approach for identifying synergistic drug combinations. PLoS Comput Biol. 2017;13(1):e1005308. doi: 10.1371/journal.pcbi.1005308