By George Bakas, Data Scientist

Deep Learning (DL) is a subset of Machine Learning (ML) which falls under the category of Artificial Intelligence (AI). But how DL have an impact on the fight against cancer?

..a DL model can “learn” features! To begin with, DL is based on artificial neural networks, that are constructed in layers and receive and process information. In contrast to traditional ML, which can be only taught information or features of images that are extracted and coded by humans, DL is highly flexible and scalable. In this scope, a DL model can “learn” features by itself which makes it extremely powerful when it comes to image analysis.

..challenges in the field of histopathology: lack of experienced physicians, limitation of global healthcare resources! Digital pathology is an era that has erupted during the past years. The ability to digitize Whole Slide Images (WSI) has given rise to the ability of end users to process the images in very interesting ways. As a result, Computer vision and specifically Deep Learning models have been implemented in many ways in order to identify objects within images. Digital Histopathology combined with artificial intelligence promises to increase both the accuracy and the availability of high-quality healthcare to patients. Typical limitations in the field of histopathology are the lack of experienced physicians and the limitation of global healthcare resources.

..pathological diagnosis, classification and disease prognosis and prediction! An important fact that contributes to the development of collaboration between AI and histopathology is the ever-increasing amount of available data which is now provided in digital form. AI technologies have the ability to handle a large volume of data generated through a patient’s life cycle to improve pathological diagnosis, classification and disease prognosis and prediction.

..do not replace but rather support medical staff! Machine learning based AI algorithms must be used effectively to process and understand the underlying data within an image (more than what a human can interpret) as well as identify human misclassification. The most important advantage that the field of digital histopathology acquires combined with AI methodologies is the reduction of errors in the diagnosis and more efficient classification of patient samples. The underlying automated methodologies, do not replace but rather support medical staff, in order to avoid mistakes or highlight important information that may be missed within the image.

.. digital pathology passes through AI! Is deep learning a candidate that can eliminate faults and misidentifications in the fight against cancer? None can answer directly, but it definitely has already enhanced and updated the future of digital pathology.