1 May 2023 — Artificial intelligence (AI) could help doctors diagnose lung cancer earlier, according to a study by researchers at the Royal Marsden NHS Foundation Trust in collaboration with the Cancer Institute of London and Imperial College London. may be useful for
The LIBRA study, supported by The Royal Marsden Cancer Charity, the National Institute for Health and Care Research (NIHR), RM Partners, and Cancer Research UK, used data from CT scans of approximately 500 patients with large lung nodules. and AI algorithms. We then tested the AI model to see if it could accurately identify cancerous nodules1.
Lung nodules are abnormal growths that are common and mostly benign. However, some lung nodules can be cancerous, and large nodules (such as 15-30 mm in size) pose the highest risk.
Researchers hope that the technology could ultimately speed up lung cancer detection by helping to quickly track high-risk patients to treatment and streamlining the analysis of patient scans.
The authors used a measure called AUC (“area under the curve”) to assess how effective the new model was in predicting cancer. An AUC of 1 indicates a perfect model, but 0.5 is expected if the model is guessing randomly. Results published in The Lancet’s eBioMedicine show that the AI model was able to identify cancer risk for each nodule with an AUC of 0.87. Performance improved on the Brock score, the test currently used in the clinic, where he scored 0.672.
The new model also performed comparable to the Herder score, another test currently used in the clinic, with an AUC of 0.83. However, the artificial intelligence model uses only two variables, which may simplify and speed up nodule risk calculations in the future, as opposed to a Herder score of 7 and a Brock score of 93,4. .
The new model could also help clinicians make decisions about patients who do not currently have a clear referral route. Using Herder, a patient is classified as low risk if the score is less than 10% of hers and as high risk and requires intervention if the score is greater than 70% of his. A wide range of testing or treatment options can be considered for patients in the intermediate risk group (10-70%). Combined with Herder, the researchers’ model was able to identify this group of high-risk patients and would have suggested early invention for 18 of 22 (82%) nodules diagnosed with cancer.
To analyze the CT scan data, researchers used a technique called radiomics. This makes it possible to extract information about a patient’s illness from medical images that are not easily visible to the human eye.
Lung cancer is the leading cause of cancer death worldwide, accounting for just over a fifth (21%) of cancer deaths in the UK5. Patients diagnosed with early-stage disease can be treated much more effectively, but recent data show that over 60% of lung cancers in the UK are diagnosed at stage 3 or 4. As such, initiatives to speed up detection are urgently needed6.
Dr Benjamin Hunter, Clinical Oncology Registrar at The Royal Marsden NHS Foundation Trust and Clinical Research Fellow at Imperial College London, funded by Cancer Research UK, said:
“According to these initial results, our model appears to accurately identify large cancerous lung nodules. We hope that tracking will improve early detection and potentially lead to more successful cancer treatments. We’ll see if we can predict it accurately.”
The Principal Investigator of the LIBRA study, Dr. Richard Lee, is Consultant Physician in Respiratory Medicine and Early Diagnosis at the Royal Marsden NHS Foundation Trust, Team Leader of the Early Diagnosis and Detection Team at the Cancer Institute, London, and The Funded by the Institute of Cancer Research. The Royal Marsden Cancer Charity said:
“Although in its early stages, this study is an example of the significant scientific and clinical research that we are conducting at Royal Marsden and the ICR Center for Early Diagnosis and Detection. We want to push the boundaries of speeding up disease detection using
“People diagnosed with lung cancer early are much more likely to survive five years than those whose cancer is discovered later. This study is the first to develop a radiomics model specifically focused on large pulmonary nodules, which may one day help clinicians identify high-risk patients. there is. ”
Keith Hewett, 64, from Watford, was diagnosed with lung cancer in 2018 and underwent surgery at a local hospital. He was then referred to Dr. Richard Lee at Royal Marsden for follow-up. Last year, a CT scan found nodules in Keith’s lungs, and after further investigation, he was diagnosed with cancer again. Keith, who has a similar medical history to the patients used in the study, said:
“After the initial diagnosis, I had CT scans every three months at Royal Marsden, and just about to get every six months, Dr. Lee noticed a change in the scans. I wasn’t sure what it was, but agreed it needed further investigation. I got
“It turned out I had 3 cancerous nodules in my lungs and I had surgery at the Royal Brompton. My care at The Royal Marsden was excellent with great attention to detail and comfort in their care.” I felt it.
“If there’s new technology that can help CT scans make it clearer whether something is cancer or not, that would be great. You’ll want to know as soon as possible if you’re affected.”
More information: https://www.icr.ac.uk/
References:
- This study was supported by the Center for Early Diagnosis and Detection, which aims to facilitate early diagnosis of cancer. The center was established at The Royal Marsden in partnership with The Institute of Cancer Research (ICR) and is funded by the Royal Marsden Cancer Charity and the National Institute for Health and Care Research Biomedical Research Center (NIHR BRC) . Marsden and ICR.
- The Brock score is calculated using nine patient characteristics: age, sex, family history, history of emphysema, nodule size, nodule density, nodule location, number of nodules, and spinous processes. See https://www.uptodate.com/contents/calculator-solitary-pulmonary-nodule-malignancy-risk-in-adults-brock-university-cancer-prediction-equation
- The Herder score is calculated using similar patient characteristics (e.g. age, smoking status, cancer history) and nodule characteristics (e.g. size, location, spinous process status), but the PET scan It also includes the findings of https://www.brit-thoracic.org.uk/quality-improvement/guidelines/pulmonary-nodules/pn-risk-calculator/
- This new model focuses on nodule size (ratio of nodule surface to volume) and its texture (gray length co-occurrence matrix (GLCM) correlation) on CT scans.
- https://www.cancerresearchuk.org/health-professional/cancer-statistics/statistics-by-cancer-type/lung-cancer#heading-One
- https://crukcancerintelligence.shinyapps.io/EarlyDiagnosis/
