Blog
April 30, 2024
5 mins

Advancing Bowel Cancer Awareness: The role of AI in early detection and precision medicine

Colorectal cancer (CRC), also known as Bowel Cancer, has become the third most common malignancy and the second most deadly cancer worldwide, with an estimated 1.93 million new CRC cases diagnosed and 935,173 deaths in 2020 worldwide (NIH). Further, the global new CRC cases are predicted to reach 3.2 million in 2040.

As Bowel Cancer Awareness month in the UK draws to a close, what can we expect from bowel cancer research over the next year? Tailoring treatment plans to individual patients based on their genetic makeup, tumor characteristics, and other factors will become increasingly common. This personalized approach can lead to better outcomes and reduced side effects by targeting therapies more precisely.

At the cutting edge of precision medicine - as with so much these days – is AI. AI algorithms and machine learning models trained on large datasets of CRC patient information can help predict outcomes, identify high-risk individuals, and optimize treatment strategies. 

These technologies have the potential to transform how CRC is detected, managed and treated. AI algorithms can analyze various types of medical imaging, including colonoscopy images, to detect and characterize colorectal lesions, including polyps and tumors. These algorithms can identify suspicious features indicative of CRC with high sensitivity and specificity, assisting pathologists and gastroenterologists in making more accurate diagnoses.

Enhancing pathology interpretation
Digital pathology is pathology 2.0. It takes the pathology of the last two centuries into the internet age.

To understand how AI can change CRC research, we first need to understand the importance of digital pathology – and how it updates routine diagnosis.

Routine diagnosis in pathology involves the application of a stain (essentially a dye) called H&E (hematoxylin and eosin) to tissue samples on microscope slides which highlight and distinguish cellular structures. This helps pathologists spot any abnormalities; however, the technique has its limitations. It is time-consuming and prone to varying interpretation and diagnosis by pathologists, even when considering just a single slide. 

Here’s where digital pathology comes in. These H&E stained slides can be scanned with an automated microscope to produce a composite high-resolution image file of the entire slide - a whole slide image (WSI), opening the door to AI analysis.

AI tools can be trained on WSIs to analyse the H&E stained slides, enabling pathologists to gather more, and better, patient data to inform diagnosis, treatment and monitoring.

Rather than replacing the pathologist’s expertise, AI serves as a powerful aid, helping them work more efficiently and accurately.

Digital techniques can help minimise varying interpretations and free people from repetitive lab work, as well as evaluate images and identify details that the human eye could miss. It can also be incredibly cost-effective – a 2020 Deloitte research report estimated that in Europe, “AI could save up to 53 million hours of routine analyses for clinical technicians, linked to potential savings up to £755m (€883m) per year”.

Transforming early detection and treatment of CRC with digital pathology

So what does this mean for colorectal cancer? The use of digital pathology holds great promise for improving the accuracy, efficiency, and personalized management of CRC, particularly in diagnostics. AI algorithms can identify histological features associated with different tumor subtypes, stages, and grades, providing valuable diagnostic and prognostic information.

AI-driven diagnostic systems have the potential to improve the accuracy of colorectal cancer detection by reducing false positives and false negatives. By analyzing large datasets of annotated medical images, AI algorithms can learn to distinguish between benign and malignant lesions with high sensitivity and specificity.

One AI tool already being used in the lab is Owkin’s MSIntuit® CRC, the first CE-marked AI diagnostic that prescreens for MSI - an important genetic marker to determine colorectal cancer treatment - directly from H&E WSIs.

Normally, every colorectal cancer patient would need to send a sample for molecular testing to check for MSI during diagnosis. But MSIntuit® CRC can screen out a proportion of those patients who don’t need this testing – saving time and money for pathologists and potentially time for patients.

Clinicians and AI working together

AI systems excel at detecting patterns, identifying anomalies and accurately predicting outcomes. That’s what has allowed us to initially create this kind of technology - and as we keep innovating, we have incorporated more and more superpowers of AI into our models (such as the ability to recreate any missing areas of WSIs).

While the promise of AI in CRC diagnostics is clear, successful implementation requires rigorous validation, regulatory approval, and integration into clinical workflows. Collaboration between multidisciplinary teams of clinicians, researchers, and industry partners is essential to ensure the effectiveness, safety, and scalability of AI-driven diagnostic solutions in real-world healthcare settings.

By leveraging advanced computational techniques and big data analytics, AI has the potential to revolutionize CRC screening, diagnosis, and treatment, ultimately leading to reduced healthcare burdens, earlier detection, more accurate diagnosis, and improved patient outcomes in the fight against colorectal cancer.

References

Watts, T. (2024), Raconteur, An AI diagnostic revolution – pushing the digital frontiers of pathology https://www.raconteur.net/healthcare/ai-diagnostic-digital-pathology

American Cancer Society, Colorectal Cancer Statistics: How Common Is Colorectal Cancer? https://www.cancer.org/cancer/types/colon-rectal-cancer/about/key-statistics

Authors
Owkin
Testimonials
No items found.
Advancing Bowel Cancer Awareness: The role of AI in early detection and precision medicine

No items found.
No items found.
No items found.