Revolutionizing Disease Diagnosis Through AI Driven Digital Pathology

Pathology is the science of diagnosing diseases, primarily through collecting and analyzing samples of tissues, cells and body fluids. The traditional process of pathology involves the evaluation of biopsy, wherein tissues are preserved with chemical fixatives (such as formalin) and further transferred to histology labs. Following this, the specimen undergoes a set of processes, such as treatment, embedment, sectioning, and staining. Further, the prepared histological slides are examined and evaluated under a microscope by a trained pathologist.

The process of AI-based digital pathology allows scanning of slides via computer monitors, by replacing the conventional microscopic approaches. Further, by converting glass slides to images, samples can be transmitted from diagnostic centers to pathologists within a fraction of time.

AI-based digital pathology enables identification of optimal treatment plans based on patient profiles, by utilizing digital methods for patient classification and selection for diagnostic tests.

AI has lately had an unprecedented influence on medicine and can have a significant impact in pathology. Given the massive amount of data generated by pathology, AI may present an opportunity for all pathology subdomains to innovate and offer a revolutionary care delivery model in both imaging and non-imaging applications. Owing to the advantages over conventional approaches in the field of pathology, AI based digital pathology market is anticipated to grow at a CAGR of around 8.3%, till 2035, according to Roots Analysis.

Artificial Intelligence in Digital Pathology

Driven by the ongoing digitalization of the healthcare sector, there has been an increase in the use of AI in pathology, in recent years. AI, along with its subfields of machine learning and deep learning, is quickly becoming a key technology in the healthcare sector, with the potential to transform lives and improve patient outcomes, across a wide range of medical specialties.

At present, multiple AI-based approaches have been developed and are being used to assist pathological diagnosis and research. One of these approaches includes the use of deep learning program using artificial neural networks (ANNs). In a manner similar to a biologically complex neural network of the human brain, ANNs may independently determine whether their interpretation or prediction is accurate. In addition, convolutional neural networks (a type of deep multi-layer neural network) are specialized for visual images. Convolutional networks serve as a pre-processing step that enables computer vision and machine vision models to process, examine and categorize digital pictures, or portions of images, into predefined categories.

It is worth highlighting that the new standard of treatment will incorporate AI-based digital pathology together with clinical data, biomarkers and multi-omics data. In addition, to facilitate a more effective pathology workflow, AI-based digital pathology offers a detailed and individualized picture, thereby, allowing pathologists to address the progression of complicated diseases for improved patient treatment.

Workflow of AI-based Digital Pathology

Figure below highlights information on steps involved in the usual workflow of AI-based digital pathology process.

Further, steps involved in the workflow of AI approaches in digital pathology have been briefly described below:

1. Preparation of Tissue Sample: This process is very similar to the conventional approach. A pathologist examines a biopsy to determine its color, size and consistency. At this point, the specialist can detect symptoms of malignancy and select which areas of a specimen should be inspected under the microscope. Further, the chosen region is prepared by following multi-step processes, such as treatment of the tissue with chemicals in order to maintain its structure, mounting the specimen on a glass slide, staining to improve contrast and protecting the tissue with coverslips.

2. Converting into Virtual Sample / Whole slide imaging (WSI): WSI or virtual microscopy is a technique that is used to enable digital pathology. Its central component, which is a WSI scanner, captures a picture of the glass slide and generates a precise electronic replica known as a virtual slide. It is worth noting that virtual slides, unlike glass slides, are easy to replicate, save, categorize and distribute. Furthermore, they may be linked to electronic health records, thereby providing a complete picture of a patient’s health.

3. Saving a Virtual Slide: The scanner pre-processes the virtual slide automatically and stores it to on-premises or cloud storage. In order to minimize the file size, a compression approach is frequently employed before saving the slide.

4. Viewing and Editing of Slide: In the digital process, instead of using a traditional microscope, a pathologist uses a computer display to analyze enlarged tissue samples. A slide viewing and management software is used to zoom out a tissue segment and observe its smallest features. In addition, this software allows the pathologist to view the slide from different angles, add annotations and even compare multiple images at one time.

5. Sharing Data: Using specialized digital pathology software applications, slides are converted to an electronic format, thereby allowing them to be exchanged using the internet. These slides can be shared to gain a second opinion, as well as with patients, research facilities and other stakeholders.

6. Reporting Results: Some image viewing systems provide reporting capabilities. However, this work is often accomplished by enabling interaction with the laboratory information or laboratory information management system (LIS/LIMS) and hospital information system (HIS).

Applications of AI-based Digital Pathology Solutions

Figure below is a pictorial summary of different applications of AI-based digital pathology solutions.

Future Perspectives

The digital revolution of pathology is projected to accelerate in the coming years, considering multiple growth drivers, including growing number of laboratories adopting high throughput digital scanning and software technologies to assist diagnostic practice. In addition, factors, including shortage of skilled pathologists in remote areas, increasing pathology workloads due to ageing populations, higher rate of cancer screening programs, rising complexity of pathology testing and time constraints, and requirement for pathology labs to outsource expertise in the field, also contributes significantly towards the need for AI-based digital pathology solutions.

Moreover, the same driving forces are pushing the development of AI-based digital pathology to assist pathologists with diagnostic issues that they confront in the present scenario. By incorporating AI-based digital pathology technologies into clinical processes, possible savings may be realized, in terms of turn-around times, as well as patient outcomes, which are enabled through better detection and repeatability. Such advancements are expected to play a significant role in increasing the overall quality of AI-based digital pathology solutions. Given the aforementioned characteristics, we anticipate that the AI-based digital pathology industry will experience substantial growth over the next decade.