Performance of artificial intelligence-based breast cancer screening in a community setting: a real-world evaluation study. (July 2022)

• Record Type: Journal Article
• Authors: Ramprasad, Jenisa; Krishnan, Lakshmi; Gangadharan, Charitha; Deshpande, Gargi; Madhu, Himanshu; Kakileti, Siva Teja; Manjunath, Geetha
• Is Part Of: Lancet oncology. Volume 23(2022)Supplement 1
• Page Start: S20
• Publication Date: 2022-07
• Subjects: Oncology -- Periodicals; Neoplasms -- Periodicals; Cancérologie -- Périodiques; Oncologie; Oncology; Periodicals; Electronic journals; 616.994005
• DOI: 10.1016/S1470-2045(22)00419-3
• Languages: English
• ISSNs: 1470-2045

Abstract

Abstract: Background: Despite improvements in treatment strategies, breast cancer survival rates in India remain low because of poor awareness and advanced stages at presentation. Early diagnosis and treatment can improve the survival rates of patients with breast cancer. A population-based breast cancer screening programme with mammography is often not feasible in resource-constrained settings. Thermalytix, a novel breast cancer screening technology, applies artificial intelligence algorithms over thermal images to generate an automated interpretation report indicating breast health. The Thermalytix testing equipment is portable, requires minimal skills to operate, and operation is 20 times less expensive than a mammography. The test has been used to screen women in community settings in India. This study was done to assess the performance of the Thermalytix test in identifying women suspicious of breast cancer in community settings. Methods: Women who underwent Thermalytix tests in community-based screening camps in the Indian states of Karnataka and Maharashtra between Aug 12, 2017, and May 31, 2021, were included in this study. The Thermalytix test captured five thermal images in multiple views from each participant and the tool automatically generated a triage report. A more detailed report with annotations of probably abnormal regions and an overall breast health score ranging from 0 to 5 (B-score) was also generated. A B-score of 4 or 5 on triage indicates a high level of concern and prompts a recommendation to follow-up with breast ultrasonography. A B-score of 3 prompts a recommendation to repeat the Thermalytix test after 3 months. A B-score of 1 or 2 indicates thermal patterns of low level, usually prompting a recommendation to continue routine screening. Participants with B-scores of 4 or 5 on the Thermalytix test were followed-up with a telephone call to obtain information about their subsequent tests, such as mammography, ultrasonography, or histopathological examination. Women with a follow-up report were categorised as having "Normal", "Abnormal–Benign", or "Abnormal–Malignant" findings (as per their Breast Imaging Reporting and Data System score of 1, 2 or 3, or 4 or 5, respectively). Findings: 13 932 women were included in the analysis. 775 women complained of a breast lump, 625 (4·5%) of whom were given B-scores of 4 or 5 after Thermalytix testing and were referred for follow-up breast ultrasonography. 117 (18·7%) of these 625 women underwent standard follow-up investigations. Abnormalities were detected in 78 of the 117 followed-up women (74 benign and four malignant breast lesions), indicating that Thermalytix has a positive predictive value of 66·7% in detecting benign and malignant breast lesions. Interpretation: In resource-constrained countries where systematic screening programmes with mammography is not scalable or feasible, the portable, radiation-free Thermalytix test is a promising option for affordable and accessible breast cancer screening. This screening test is not resource-intensive and can be set up in remote locations and operated with minimal training. The Thermalytix test can be used to triage women for referral for further evaluation in secondary or tertiary health-care centres, which makes it a logistically and economically feasible option for breast cancer screening in low-income and middle-income countries. Funding: This programme was sponsored by Niramai Health Analytix and was executed in collaboration with National Health Mission, Maharashtra, India, and Bruhat Bengaluru Mahanagara Palike, Municipal Corporation of Bangalore City, Karnataka, India.