Co-author: Surajeet Dutta
As the world continues to grapple with the COVID-19 crisis, pharmaceutical research laboratories have become the center of attention. Companies, including startups, are leaving no stones unturned to harness advanced technologies to arrest the crisis.
With no dearth of data, the ability to integrate data from multiple sources is critical. Laboratories need the tools that can make the best use of data to assist in decision-making and informed and uninformed searches on data repositories.
For this, artificial intelligence (AI), which is the foundation for building any intelligent system that uses cognitive computing, is the key. It recognizes patterns and presents trends that can help us make decisions in real time.
Synergy between AI and Lab Informatics
Laboratory information systems have to move beyond traditional data storage given the limitations of relational database management systems (RDBMSs). With exponentially increasing data volumes, complex search or retrieval queries using Structured Query Language (SQL) are getting cumbersome.
With lockdown enforced all around the globe and people restricted to their homes, a solution has to be in place for continuing routine business operations remotely. To address such a situation there will be a rise in demand for voice-controlled tasks, image processing for identification of sample tags, and character identification in medical orders, making suggestions based on the history of a particular patient, and the like. Most of the laboratories store precise data but need to retrieve information based on imprecise criteria. All such new requirements can be solved by leveraging the many applications of AI.
Real-time, Data-Driven Decision Making
Recent developments in AI and machine learning (ML) have shown a ray of hope to scientists. They can now use digital analytics and tools to scan through huge volumes of multi-sourced data and select the appropriate one. Solutions leveraging these technologies accumulate theoretical data from experimental workflows and render it as usable information when required by scientists. As a result, because of digital analytics, scientists are no longer confined only to their labs and can connect and share tools with their peers both within and outside the premises.
AI in Digital Pathology
Digital pathology is a new way for scientists and physicians to interact with pathological data. Clinical laboratories are now digitizing by converting glass slides to e-slides and digital slides. This means that samples in the form of glass slides are now converted into pixels in high throughput. As a result, laboratories now have access to a significantly larger pool of data attached to metadata, such as a pathology report that provides information about the diagnosis and patient outcome. These images have been used to begin to train algorithms. Alongside, two important things occurred in the field of computer science:
- Immense improvement in computing capabilities as well as the proliferation of cloud
- A shift from ML to newer deep learning technologies within the AI umbrella
Unlike the traditional applications of AI and ML, which require an expert pathologist to annotate images and train the algorithm, larger datasets allow laboratories to use convolutional neural networks to distinguish the important features required for an accurate diagnosis.
Auto-diagnosis is Now a Possibility
Digital pathology uses computerized and auto diagnosis. The virtual microscopy generates multiple images of the same tissue for parallel analysis. With a multitude of annotated images from different tissue samples, AI can recognize the regular and abnormal features in tissues and identify a pattern that may point to a type of disease. This is a huge stride toward auto diagnosis personalized medication as we can integrate radiology with genomics, proteomics, and demographics for more accurate prognoses.
Scientists and bioengineers have already started working on a similar area where they have merged AI with standard microscopy and infrared light for assembling digital biopsies for the identification of vital molecular characteristics of cancer samples without using dyes and labels. Standard biopsies reveal cellular details using staining methods, which take a relatively long time. Besides, any diagnosis by manually studying the shape and size of the cells can lead to errors as the observation may vary from pathologist to pathologist.
ML-enabled computer programs combine white light and infrared signals that bounce off the biopsy sample when viewed under a virtual microscope. AI can gather details that cannot be seen by human eyes.
Going Beyond Boundaries for Early Detection
Due to ethical reasons and accessibility limitations, the study of human placenta in situ was never possible. With the advancement in technology, scientists are now bringing engineering technologies like AI, IoT, and ML into microbiology to develop a unique 3D model for studying complicated processes. The model is powered by a microfluidic sensing chip, which helps physicians and scientists to study placenta-related medical conditions and detect anomalies before they mature.
According to 2018 data, 228 million cases of malaria were reported worldwide, becoming the most severe public health problem (prior to COVID-19).
Malaria does not transmit from a pregnant woman to the fetus in utero. However, the infected red blood cells (RBC) flowing through the blood vessels in the placenta can affect both the woman and the fetus. This phenomenon results in resulting in about 10,000 maternal and 200,000 newborn deaths annually. The developing and subtropical countries are the worst hit, especially sub-Saharan Africa. This can be a major breakthrough in the area of placenta study.
Scientists have been researching on the reproducibility crisis for long. Next-generation laboratories can help in expediting the research by integrating data access and collection with the scientists’ workflows. Uninterrupted access to information enables scientists to make data-driven decisions in real time. Digital assistants and devices automatically capture and analyze data for the accuracy of the records as per the standardized, validated protocols and for generating reproducible results. As the reliability of the findings will increase, so will accelerate the scientific advances and the launch of new products in the market. With the advent of new technologies, growing competition, and changing dynamics, AI has become essential. The emergency of AI-enabled next-generation laboratories is not a distant-future dream but the current-age imperative.