How AI is improving cancer diagnostics

How AI is improving cancer diagnostics

Divan van Ruler & Angela de Loureiro
24 March 2022

How AI is improving cancer diagnostics

Artificial intelligence and machine learning techniques are breaking into biomedical research and health care. This includes cancer research, where the potential applications are vast. Cancer, as we know it, is first described in an Egyptian text dating back to as early as 3000 BC. The nature of cancer has remained consistent throughout the ages. However, diagnostics and treatment have evolved over time as complications, such as drug resistance, emerged. 

Today AI and machine learning are unlocking new and improved methods of discovering and treating cancer. In this post, we’ll be discussing how AI is improving cancer diagnostics.

Lung cancer

Lung cancer is the leading cause of cancer-related deaths worldwide, as well as in South Africa. The latest World Health Organization (WHO) fact sheet on cancer states that lung cancer accounts for 2.21 million new cancer cases worldwide in 2020, coming second only to new breast cancer cases. Lung cancer is responsible for 1.8 million deaths worldwide in 2020, far outnumbering other cancer types. To effectively combat lung cancer, early detection is paramount. However, with no official screening program and low awareness of lung cancer, early detection is a difficult task.

The application of machine learning to lung cancer is already underway. Models can differentiate between Non-Small Cell Lung Cancer (NSCLC) and Small Cell Lung Cancer (SCLC) based on histological data. The NSCLC and SCLC subgroups can be further subdivided, with machine learning models being able to differentiate between members of the Non-Small Cell Lung Cancer group. Without the aid of machine learning, this highly laborious differentiation would require highly skilled pathologists.

Colorectal cancer

Like lung cancer, colorectal cancer is another cancer of great concern both locally and internationally. The WHO fact sheet lists colorectal cancer as the third largest cause of new cancer cases (1.93 million). It is the second-largest contributor to cancer-related deaths (916 000)2 globally. In 2017, colorectal cancer was the 2nd most common cancer among South African men and 3rd most common cancer among South African women. Thus, early detection is crucial for the successful treatment of colorectal cancers. However, colorectal cancer often shows no symptoms in many people during its early stages, which can complicate diagnosis. When symptoms do present at later stages, they can often vary depending on tumour size and location.

While possible, detecting colorectal cancer early on presents different challenges than those discussed in lung cancer. Models for machine learning typically infer indirect colorectal cancer diagnoses. Models use data from biomarkers instead of a scan of the colon itself. Complete blood count, as well as whole-genome sequencing of plasma cell-free DNA, is useful in building models to detect colon cancer.

Breast cancer

Breast cancer might not rank amongst the worst outlooks, but it is still one of the most common cancers in women.

Generally, it should become routine for an annual breast scan to ensure that the breasts are fine. AI can assist during these routine procedures to identify breast cancer.

No matter how the diagnosis is concluded, be it with ultrasound, mammogram, or MRI, AI ensures that extra level of confidence in the final decision. It is critical to not overlook the potential threat. AI provides that needed confidence to conclude whether it’s malignant or benign.

Future innovation in cancer diagnostics

The need for a cost-effective way to screen large amounts of people in a non-invasive manner has risen. The interest in machine learning and data science over the last decade has increased the amount of research conducted on the application of machine learning in the diagnosis of cancer and show no signs of slowing down. New models are constantly being developed, compared, and tested to ensure their robustness.

In future, models may even be expanded to include pathologies outside the scope of cancer, such as a model with the ability to differentiate between lung cancer and tuberculosis based on a patient’s CT scan. Machine learning can also be expected to contribute to the field of drug discovery in an increasingly large way as drug resistance is becoming a major issue in cancer treatment. The process of drug discovery is complex and can take years to bring a new medication that is safe to use to market. Machine learning can greatly cut down on development time by identifying prognostic biomarkers, validating drug targets, and analysing clinical trial data.

Conclusion

Cancer is going to be around for centuries to come. The common theme throughout is the importance of early detection. AI is not only paving the way for early detection and treatment – but adds in that extra level of confidence when a diagnosis is made. AI is not here to take over but to complement and assist our medical team to better support us as a nation.

How Integrove can help you

Here at Integrove, we have a team of data scientists who are passionate about solving problems. Our data scientists are experts in using machine learning, AI, and deep learning to establish a baseline, meaning what constitutes “normal” behaviour for a system, by monitoring relevant attributes or features. Our experience includes working with medical practitioners and engineers to create the best solutions for your unique business.

Contact Us