This website is intended for healthcare professionals only.
Take a look at a selection of our recent media coverage:
4th July 2022
The combination of an artificial intelligence (AI) system and a radiologist provides better screening accuracy for breast cancer as demonstrated by a higher sensitivity and specificity according to the findings of a retrospective analysis by an international team of radiologists.
The use of screening mammography is designed to identify breast cancer at earlier stages as treatment will be more successful. Moreover, in recent years there has been increased interest in the use of AI systems and a recent study found that the use of an AI system outperformed all of the human readers, with a greater area under the receiver operating characteristic curve margin of 11.5% for screening breast cancer mammograms. Nevertheless, a 2021 systematic review which considered the use of AI for image analysis in breast cancer screening programs concluded that the current evidence for AI does not yet allow judgement of its accuracy in breast cancer screening programmes, and it is unclear where on the clinical pathway AI might be of most benefit. Other work that considered the role of AI for breast cancer screening suggested that an AI system can correctly identify a proportion of a screening population as cancer-free and also reduce false positives and therefore has the potential to improve mammography screening efficiency.
But what if an AI and radiologists worked together, so that the AI could initially triage scans and identify normal cases but those with suspected cancer and where there was diagnostic uncertainty, were referred to the radiologist? This was the question addressed in the retrospective analysis by the research team. The system was designed so that the AI system would flag potential cancerous scans and where it was unsure about the diagnosis, for a second read by a radiologist. The team initially trained the AI system using an internal dataset and then used an external data set and compared the interpretation with that of a radiologist. The performance of both the AI and radiologists was assessed in terms of sensitivity and specificity and the test sets contained a mix of both normal and cancerous scans.
AI and radiologists combined performance
For the external data set the radiologist had a higher sensitivity (87.2% vs 84.6%, radiologist vs AI system) and specificity (93.4% vs 91.3%) and in both cases this difference was statistically significant (p < 0.001 for both).
However, when the AI and radiologists worked together, the radiologist’s sensitivity was 89.7% and the specificity 93.8%. In other words, the combination improved both sensitivity and specificity. The authors calculated that this corresponded to a triaging performance, i.e., the fraction of scans which could be automated) of 60.7%.
Based on these findings, the authors concluded that their system leverages the strength of both the radiologist and the AI system and had the potential to improve upon the screening accuracy of radiologists.
Leibig C et al. Combining the strengths of radiologists and AI for breast cancer screening: a retrospective analysis Lancet Digit Health 2022
6th April 2022
The fracture detection rates are comparable for artificial intelligence (AI) and clinicians according to the findings of a meta-analysis by researchers from the Nuffield Department of Orthopaedics, Rheumatology and Musculoskeletal Sciences, Oxford, UK.
Fractures represent a common reason for admission to hospital around the world. However, research suggests that fortunately, fracture rates have stabilised. For example, one 2019 UK-based study observed that the risk of admission for a fracture between 2004 and 2014 was 47.8 per 10,000 population but that the rate of fracture admission remained stable. Unfortunately, however, fractures are not always detected on first presentation as witnessed by a two-year study in which 1% of all visits resulted in an error in fracture diagnosis and 3.1% of all fractures were not diagnosed at the initial visit. One solution to improve upon the diagnostic accuracy of fractures is the use of artificial intelligence systems and in particular, machine learning, which enables algorithms to learn from data. Related to machine learning is deep learning, which is a more sophisticated approach to machine learning that uses complex, multi-layered “deep neural networks. Deep learning systems hold great potential for the detection of fractures and in a 2020 review, the authors concluded that deep learning was reliable in fracture diagnosis and had a high diagnostic accuracy.
For the present meta-analysis, the Oxford team further assessed and compared the diagnostic performance of AI and clinicians on both radiographs and computed tomography (CT) images in fracture detection. The team searched for studies that developed and or validated a deep learning algorithm for fracture detection and assessed AI vs clinician performance during both internal and external validation. The team analysed receiver operating characteristic curves to determine both sensitivity and specificity.
Fracture detection rates of AI and clinicians
A total of 42 studies with a median number of 1169 participants were included, 37 of which included fractures detected on radiographs and 5 with CT. A total of 16 studies compared the performance of the AI against expert clinicians, 7 to experts and non-experts and one compared AI to non-experts.
When evaluating AI and clinician performance in studies of internal validation, the pooled sensitivity was 92% (95%CI 88 – 94%) for AI and 91% (95% CI 85 – 95%) for clinicians. The pooled specificity values were also broadly similar with a value of 91% of AI and 92% for clinicians.
For studies looking at external validation, the pooled sensitivity for AI was 91% (95% CI 84 – 95%) and 94% (95% CI 90 – 96%) for clinicians on matched sets. The specificity was slightly lower for AI compared to clinicians (91% vs 94%).
The authors concluded that AI and clinicians had comparable reported diagnostic performance in fracture detection and suggested that AI technology has promise as a diagnostic adjunct in future clinical practice.
Kuo RYL et al. Artificial Intelligence in Fracture Detection: A Systematic Review and Meta-Analysis Radiology 2022
27th September 2021
Diffuse intrinsic pontine glioma (DIPG) is a childhood brain tumour that is both highly aggressive and difficult to treat. It is an essentially incurable brain tumour with an incidence of 1 to 2 cases per 100,000 and occurs mainly between the ages of 6 to 9 years with an estimated median overall survival of 11.2 months. The lack of progress with treatments was highlighted in a 2020 review of trials in DIPG which commented on how there was a consistency of non-significant improvement to prognosis in these trials.
The underlying cause of this incurable brain tumour is thought to be related to genetic mutations, one of which occurs in the ACVR1 gene that encodes the serine/threonine kinase, ALK2 and is present in around 21% of DIPG samples. There are currently no known treatments for DIPG with this mutation although some data from patient-derived in vitro tumour models suggests that inhibition of ALK2 reduces cell viability. The clinical development of ALK2 inhibitors is therefore a potentially important avenue to explore. Nevertheless, drug discovery can be a time-consuming process although one recent approach which is becoming used more frequently, is big data analytics and in particular, the use of artificial intelligence (AI) systems.
The use of AI for drug discovery essentially seeks to explore whether currently available treatments can be repurposed and this was the approach taken by a team from the Division of Molecular Pathology, Institute of Cancer Research, London, UK, to identify potential drug candidates for ACVR1-mutant DIPG cases. The researchers turned to BenevolentAI which makes use of a knowledge graph that incorporates millions of documents and information on diseases, biological tissues, genes and proteins. Using BenevolentAI, the researchers looked for compounds that were likely to have an inhibitory effect on the ALK2 protein but also one that could penetrate into the central nervous system.
In their search for a treatment for the incurable brain tumour, one likely candidate emerged: vandetanib and which is already approved for symptomatic medullary thyroid cancer in patients with unresectable locally advanced or metastatic disease. However, drug levels of vandetanib in the cerebrospinal fluid have been found to be low, and a search for compounds that would enable vandetanib to remain within the brain long enough to induce a therapeutic effect uncovered everolimus. The researchers found that in vivo studies in mice demonstrated that combining vandetanib with everolimus, which is again approved for several cancers, increased the brain concentration of vandetanib by 56%.
In further animal studies, the combination extended survival by 14 % compared to controls and also significantly reduced tumour burden over the course of a four week treatment. To examine the clinical value in patients, the researchers used both drugs in four children with DIPG due to ACVR1 mutations.
The study lead, Professor Chris Jones, said “DIPG is a rare and aggressive childhood brain cancer, and survival rates have not changed over the past 50 years, so we desperately need to find new treatments for this disease.”
The authors concluded by calling for an early phase clinical trial to determine the utility of this approach for children with this incurable brain tumour.
Carvalho DM et al. Repurposing vandetanib plus everolimus for the treatment of ACVR1-mutant diffuse intrinsic pontine glioma. Cancer Discovery 2021.
23rd September 2021
Globally, in 2020, there were an estimated 2.3 million women diagnosed with breast cancer leading to 685,000 deaths. Fortunately, improvements in survival over recent decades have been attributed to population-based breast cancer screening with mammography. In fact, a recent UK study suggested that screening reduces cancer mortality by 38% among women screened at least once.
The use of artificial intelligence (AI) systems for image recognition in breast cancer screening could lead to improvements in the detection of cases, either as a standalone system or as an aid to radiologists. Indeed, there is some evidence to support the value of AI with one retrospective analysis of an AI screening algorithm concluding that it showed better diagnostic performance than a radiologist. Nevertheless, in a 2019 review, it was concluded that while AI systems have good accuracy for breast cancer detection, methodological concerns and evidence gaps exist that limit translation into clinical breast cancer screening settings.
In light of these concerns, a team from the Division of Health Sciences, University of Warwick, UK, were commissioned by the UK National Screening Committee to undertake a systematic review to determine whether there was sufficient evidence to support the introduction of AI for mammographic image analysis in breast screening. They conducted literature searches up to May 2021 and included studies that reported the test accuracy of AI algorithms either alone or in combination with radiologists, to detect breast cancer in digital mammograms in screening practice or in test sets. The team included cancer confirmed by histological analysis of biopsy samples in cases where women were referred for further tests after screening as the reference standard or from symptomatic presentation during follow-up.
The review identified a total of 12 studies including 131,822 women undergoing breast cancer screening. In studies with a standalone AI system, the algorithm calculated a cancer risk score, categorising women at either high (recall) or low (no recall) risk. When used to assist the radiologist, the AI system simply provided a level of suspicion. In two large retrospective studies including 76,813 women, that compared the AI system with the clinical decisions of a radiologist, 96% of systems were less accurate than a single radiologist and all were less accurate than a double read.
Overall, the authors reported considerably heterogeneity in study methodology, some of which resulted in high concerns over the risk of bias and applicability. In their study, they commented that “evidence is insufficient on the accuracy or clinical effect of introducing AI to examine mammograms anywhere on the screening pathway”.
In the conclusion, the authors noted how AI systems for breast cancer screening are a long way from having the quality and quantity required for implementation into clinical practice.
Freeman K et al. Use of artificial intelligence for image analysis in breast cancer screening programmes: systematic review of test accuracy. BMJ 2021
9th October 2020
Now a team from Weill Cornell Medicine, New York, has created an artificial intelligence (AI) system that can use routine test data results to determine if a patient has COVID-19. Normally, clinicians order a battery of blood tests in addition to a PCR test, including routine laboratory tests and a chest X-ray and these results are generally available within 1 – 2 hours. Researchers therefore hypothesised if the results of the routine laboratory test could be used to predict if someone was infected with COVID-19 without the PCR test. The included patient demographics such as age, sex, race into a machine learning model and incorporated the results for 27 routine tests. The laboratory results were made available two days before the PCR test result. The dataset included a total of 5893 patients admitted to hospital between March and April 2020 and they excluded individuals under 18 years of age and those who PCR result was inconclusive and patients without laboratory test results within two days prior to the PCR test.
A total of 3356 patients who were tested for COVID-19 were included with a mean age of 56 years of whom, 1402 were positive and 54% emergency department admissions. Using a machine learning technique known as a gradient boosting decision tree, overall, the algorithm identified COVID-19 positivity with a sensitivity of 76% and a specificity of 81%. However, limiting the analysis to emergency department patients, increased the sensitivity to 80% and the specificity to 83%. Moreover, the algorithm correctly identified those who had a negative COVID-19 test result. A recognised limitation was the testing was specific those admitted to hospital with moderate to severe disease and thus requires further work to identify milder cases.
Nevertheless, the authors concluded that their algorithm is potentially of value in identifying whether patients have COVID-19 before they receive the results of a PCR test.
Yang HS et al. Routine laboratory blood tests predict SARS-CoV-2 infection using machine learning. Clin Chem 2020; https://doi.org/10.1093/clinchem/hvaa200