Although the phrases "deep learning algorithms” and “image visualization techniques" invoke ideas of high-tech surveillance, these technologies may soon be more applied in medicine as compared to military scenarios. A team of IIT Kharagpur researchers in collaboration with the Breast Oncology team at Tata Medical Centre (TMC), Kolkata has just devised a technology that makes possible the automated identification of aggressive breast cancers more precisely than a pathologist.
At the Biomedical Imaging Informatics (BMI) Laboratory of Indian Institute of Technology Kharagpur, researchers have developed advanced deep learning (DL) technology for computerized recognition of candidate hotspots and subsequent proliferation rate scoring by quantifying Ki-67, a non-histone protein sensitive to radiotherapy and chemotherapy from breast cancer immunohistochemical (IHC) images. This research work has been published in a very highly reputed peer reviewed journal, Nature Scientific Reports
“Our main aim is to provide the expertise of top-level pathologists where they are not available as well as to assist pathologists for faster and accurate detection of hotspots and scoring Ki-67 index. Moreover, such an advantageous technique provides automated stratification of differential proliferation rates (less, moderate and high), which are helpful for breast cancer prognostication”, said Chandan Chakraborty, research leader of the new study and Professor in-charge of BMI Laboratory. He added that this technology will be very beneficial to those living in developing countries especially in rural/remote areas.
“This software can allow accurate identification of the aggressive cancers anywhere, even in the remotest part of the country, allowing faster referral and quicker treatment", says Dr. Sanjoy Chatterjee, senior clinical oncologist, TMC Kolkata, India. The study team from TMC Kolkata, also included senior breast surgeon Dr. Rosina Ahmed and lead pathologist Dr. Indu Arun.
“Currently a manual method is used by the pathologist which is dependent on a steep learning curve and gaining experience under mentorship. We need an automatic scoring mechanism which will provide high throughput, more objective and reproducible results in comparison with the manual evaluation”, said researcher Monjoy Saha, DST-INSPIRE fellow at BMI Lab.
This software will not only improve the throughput of Ki-67 reporting in high volume tertiary care centres but also allow a standardized reporting system for much lesser margin for error. The research team of IIT Kharagpur in conjunction with the breast unit in TMC Kolkata wishes to take this research further into clinical use to improve breast cancer outcomes.
This research study has been financially supported by Ministry of Human Resource and Development, Govt. of India, under the ‘Signals and Systems for Life Sciences (SSLS)’ – a mega-initiative in healthcare by IIT Kharagpur.