An era of artificial intelligence: A pathologist’s view
In these changing times when we are coming across COVID-19 pandemic effects and everyone is looking forward to an end of this pandemic at the earliest, the next question arises, what is next. The most likely scenario is that in coming few months we will be able to get rid of this disease with positivity graph already coming down in many places. The non COVID work has also started accelerating in most of the medical institutions. The institutes now not only want to come up back to their original pre-COVID days but also want to come up in a big way with all the finest edge in technology which will sweep the medical industry eventually.
Artificial intelligence (AI) is one such field, which is already making news in the arena for the last many years and most so common in recent years. Its application in various fields is well known. For medical industry too it is going to be revolutionary. Being a laboratory person I can foresee its application well in the pathology laboratories. All our automated analytical analyzers are based on some sort of equipment intelligence wherein the machine tells that whether the sample’s integrity is existing for testing or not.
Instrument’s brain lies in the form of microprocessors and sensors. The sophisticated software system does the rest by analyzing the data according to customer needs. Broadly speaking, there are three kinds of concepts behind AI, deep learning, machine learning, and artificial intelligence. In deep learning there are multiple hidden layers in the artificial neural network. There is no need to define specific features in data input. It discovers the features from the raw data provided earlier, so the input goes via multiple layers of learning algorithm to give the output. Machine learning allows the software to come out with output even without being specifically programmed. The most important part of AI is to provide large amount of data and complexity so that machine adopts it well in its learning algorithms.
The two most widely applications of AI is in the field of histopathology and cytopathology wherein digital pathology is being combined with AI. There is speculation whether in coming time, AI will replace pathologists and a substitution has come for the eyes of pathologists.
While the coming time will only be able to answer that however there in no exaggeration in saying that, AI at least is coming fast closer to pathologist’ eyes.
AI involves learning the various cellular details and tissue diagnosis details like architecture, cell size, membrane, nuclear details, background details like necrosis, associated features like mitosis etc, then keeping it in the memory and storing those data in the form of algorithm. In future when the slide have such features, the diagnosis is made. Who could have thought few years back that without pathologist intervention a cancer diagnosis can be made?
It is not only about the diagnosis but also grading tumors. Many of the tumors involve counting of mitotic figures per high power field to grade the tumor which by manual microscopy is tedious and time consuming, and with the help of AI can be made in no time.
So it is going to be very useful in tumor proliferation score or Gleason score in prostatic carcinoma. AI involves image standardization procedures where a digitized image is compared with reference image via several algorithms.
The several segmentation algorithms such as edge, region, color, and gradient exist. In histopathology we come across various cells, nuclei, glands the features of these are related to area size, gray value distribution within the areas and length of circumference. That is how the specific objects or cells like cancer cell, mitosis, neutrophils, or glands are identified. Texture, structure, and statistical analysis help in zeroing in to specific diagnosis or identification. The day is not far when AI will revolutionize the working in pathology laboratory.
