Cancer persists as a major hurdle for medical science, evidenced by the
Unfortunately, this situation is expected to worsen, with research forecasting 28 million new cancer diagnoses annually by 2040 and over 35 million by 2050.
In addition to fundamental research, oncology, as a medical specialty, is leading the fight against cancer, a disease that remains a significant global health concern.
Oncology encompasses the full cycle of care, including prevention, early detection, diagnosis, treatment and palliative support.
While substantial progress has been made, the inherent complexity of cancer, with its numerous and diverse manifestations, continues to pose a considerable challenge for even the most skilled practitioners.
Recognising this, artificial intelligence is becoming an increasingly essential complementary tool to human expertise, offering the potential to significantly improve the accuracy and efficiency of cancer detection, diagnosis and treatment.
Using algorithms for answers
AI algorithms are exceptionally accurate at analysing certain medical images, enabling the tools to detect tumors at an early stage.
In addition, they can process extensive patient data, including genetic profiles and treatment records, to suggest personalised therapies designed for each patient, including the potential results of innovative clinical trials.
For example, without the support of AI, diagnostic imaging remains a challenging task, as radiologists must analyse complex and often low-resolution scans to identify early-stage tumours.
Professor Rachid El Fatimy
This process is inevitably prone to human error. In fact, research indicates that 50 per cent of cancers are still only detected at an advanced stage.
The case is the same for many pathologists, who may struggle to distinguish between malignant and benign abnormalities, given the subtle morphological differences.
Furthermore, with the vast and ever-growing volume of medical data, including genomic profiles, laboratory results and patient histories, clinicians face significant logistical challenges in synthesising information for timely decision-making.
Machine learning algorithms, trained on extensive datasets of medical images and clinical records, can also detect patterns that may elude human observers.
Indeed, a 2024 Nature Medicine study found that AI holds great potential in accurately distinguishing between glioma and solitary brain metastases, enabling earlier detection and treatment.
In addition, by studying scans like MRIs or CTs, AI can map a tumour’s traits with no needle required. This approach is not only more time efficient, but it is also more cost-effective.
One study found that an AI tool can analyse brain tumour scans in just three seconds, compared with the five minutes typically needed by skilled neuroradiologists.
If implemented across the UK, it could save the NHS £1.5 million over the next three years.
In breast cancer detection, AI has similarly enhanced diagnostic accuracy.
AI increases the chance of the disease being detected, with one study suggesting that the use of AI tools can result in a much higher detection rate of 17.6 per cent.
The implications of such advancements extend beyond accuracy; earlier diagnosis translates to improved treatment outcomes and higher survival rates.
AI has also helped to reveal two different subtypes of prostate cancer, allowing clinicians to more accurately classify tumors.
This advancement will not only help to save lives, but it will revolutionise the way prostate cancer is diagnosed and managed, enabling more tailored treatment to take place.
This shift from conventional, generalised treatment approaches to highly personalised therapy strategies minimises patient risk and optimises clinical outcomes.
Despite its benefits, the role of AI in medicine is often misunderstood.
It is not a miracle solution or a temporary trend. Rather, it is a radical transformation of medical practice and a method of augmenting human capabilities.
However, this integration requires careful attention to the biases inherent in AI tools and the complex ethical implications of using machines to make vital decisions.
For example, although AI can analyse data at remarkable speed, human interpretation and final processing decisions remain essential.
By effectively combining human expertise with the analytical power of machine learning, AI promises more accurate diagnoses, faster interventions and better outcomes for patients worldwide.
The data dilemma
However, the effectiveness of AI in oncology hinges on the quality and breadth of the data it processes.
Medical data is often inconsistent, fragmented and subject to errors. Imaging scans, for example, can vary based on the type of machine and the expertise of the technician.
Genetic sequencing data may contain inaccuracies, while electronic health records are frequently incomplete or dispersed across multiple institutions.
This variability poses a fundamental challenge: AI models trained on flawed or incomplete data sets will yield unreliable predictions.
Another major concern is bias in AI training data.
Many AI-driven diagnostic tools are developed using datasets predominantly sourced from high-income regions, such as North America and Europe, which limits their generalisability to diverse populations.
Efforts to mitigate these biases are underway.
For example, initiatives such as The Cancer Imaging Archive (TCIA) are expanding datasets to include a more representative cross-section of patients from across the world.
Ensuring fairness in AI requires ongoing monitoring of not only the datasets used for training, but also the methodologies employed in model development and validation.
This involves implementing robust standards for data collection, ensuring diverse demographic representation in clinical trials and continuously monitoring any AI systems for bias.
Effective collaboration between researchers, healthcare providers and governments to ensure equitable access to new advancements is also essential.
Finally, the human touch must remain, as whilst AI can support the industry, clinicians are essential in this human- centric field of care.
By addressing these challenges, the healthcare industry can unlock AI’s full potential, allowing clinicians to deliver accurate, equitable cancer care.
