TEAM QUANTUM

Scientists at the Institute of Advanced Study in Science and Technology (IASST), Guwahati, have developed a deep-learning-based (part of artificial intelligence) method to evaluate hormone status for prognosis of breast cancer. This approach will help in early detection of the cancer.

Breast cancer is the most common invasive cancer, accounting for 14 per cent of cancers among Indian women, both in rural and urban India. The post-cancer survival rate related to breast cancer was reported to be 60 per cent.

Cancer detection has always posed a major challenge. Medical professionals use several techniques for detection, such as X-ray, computer tomography (CT) scan, positron emission tomography (PET), ultrasound, and magnetic resonance imaging (MRI), besides pathological tests such as urine and blood examination.

For accurate detection of cancer, pathologists use histopathology biopsy images — examination of the microscopic tissue structure of the patient. The accuracy of the manual identification of cancer from microscopic biopsy images has always been a major concern for pathologists and medical practitioners.

Moreover, this method is subjective and may vary depending on the expertise of the analyser and the quality of the images. Hence automated image analysis of cells and tissues became an active research field in medical informatics.

Automated identification of cancerous cells has been around for decades, but recent developments in computer and microscopy hardware are attracting more attention.

Lipi B Mahanta, the head of a research group at the Central Computational and Numerical Sciences Division of IASST, has developed a novel deep-learning-based quantitative evaluation of the hormone estrogen or progesterone with the help of immunohistochemistry (IHC). IHC is a lab process that uses antibodies to check for certain antigens (markers) in a sample tissue. It basically ‘colours’ the antibodies that attach to the antigen, so the antigen can be viewed under a microscope.

Then, in comes deep learning, helping interpret the images to call out (or segment) the cancerous cells. The team developed the algorithm using the data available at the city’s premier B Borooah Cancer Institute.

“The proposed architecture, namely IHC-Net, can semantically segment the exact positive and negative nuclei from tissue images,” says a press release of the Department of Science and Technology, the parent body of IASST.

Mahanta has been using computational biology to come up with algorithmic solutions to detect cancer early. In an earlier study, she developed deep learning tools for detecting cervical cancer.

These algorithms can classify and call out cancerous cells in smear images with 98 per cent accuracy. In yet another study, Mahanta applied the same technique for the detection of oral cancer.

comment COMMENT NOW