
"A human would say it's normal. But AI detected these subtle patterns, and it was very confident. It found the cancer. We just detected lung cancer in this guy a year or two earlier than we could have otherwise!"
This is what Mozziyar Etemadi, a biomedical engineer from Northwestern University's Feinberg School of Medicine in Chicago, told Nature magazine. He was thrilled when the AI algorithm his team trained to detect cancer identified early signs of the disease on a patient’s CT scan.
This is just one example of how AI can contribute to the fight against cancer. Medical organizations recognize this and are working on implementing AI-based medical solutions in their practices.
So, what can AI do in detecting, predicting, and treating cancer? Let's find out!
Benefits of AI in Cancer Diagnosis and Treatment
AI applications in healthcare date back to the 1970s. Over time, the technology has become more advanced, with new subtypes emerging. The latest subtype, Generative AI (Gen AI), is gaining popularity in this field. This technology excels at analyzing large datasets, and unlike classic AI, Gen AI can create original content such as reports, clinical documentation, and realistic synthetic data that can be used for educational purposes or to train algorithms.
For more information on the differences between AI and Generative AI and on Gen AI use cases in healthcare and pharmaceuticals, visit our blog.
Overall, AI and its subtypes offer many advantages in healthcare. Here are five key benefits of implementing AI in cancer detection and treatment:
- Personalized Therapy: AI, supported by big data, allows doctors to study diverse information about the patient and cancer cells to develop personalized treatment. Such therapy will have a stronger impact on cancer cells while causing less harm to healthy cells. For example, the National Institutes of Health (NIH) has developed an AI tool that can analyze individual tumor cells and predict their response to a given drug.
- Reducing False Positives and Negatives: Using AI in cancer diagnostics increases accuracy, reducing the number of false positives and negatives. For example, when doctors examine mammograms, they deliver false-positive news to one in ten patients. With AI, a Google research team developed software that reduces false positives in mammogram readings by 6% and false negatives by 9%.
- Non-Invasive Tumor Classification: Doctors often perform surgery only to find that a tumor is benign, which could have been avoided. With AI in cancer detection, such cases will become less frequent. Researchers are already experimenting with these solutions. A team from Harvard University and the University of Pennsylvania developed a deep learning algorithm for tumor classification. It can identify and characterize isocitrate dehydrogenase (IDH) mutations on MRI images of gliomas without invasive procedures.
- Early Disease Detection: AI can detect changes in medical images that are invisible to the human eye and identify cancer at an early stage before it spreads to other parts of the body. Often, this early detection is a matter of life and death. For example, pancreatic cancer is typically detected in late stages when survival is only 20%. If detected earlier, survival increases to 50%.
- Speeding Up Diagnosis: Tissue analysis using traditional staining on a slide can take several days, while analysis using AI takes just seconds.
Now let's look at some practical applications of artificial intelligence in predicting, diagnosing, treating, and researching cancer.
How AI is Applied in Cancer Prediction
1. Using AI to Predict Cancer with Medical Data and Imaging: AI can not only detect existing cancer but also identify people at high risk of developing the disease before it occurs, allowing doctors to closely monitor high-risk patients and intervene immediately if necessary.
A research group from the Radiological Society of North America found that AI algorithms analyzing mammograms to predict cancer risk outperformed the standard Breast Cancer Surveillance Consortium (BCSC) risk model in predicting breast cancer. The researchers tested five AI algorithms, and all performed better than BCSC in predicting cancer development over the next five years.
AI can help detect various types of cancer. In addition to the breast cancer study mentioned above, there is a deep learning model for predicting lung cancer. It analyzes low-dose CT scans and determines whether a patient will develop cancer in the next year with 86%-94% accuracy.
Using AI to predict cancer based on medical imaging is becoming popular among cancer researchers. But there are also AI-based tools that can predict disease solely based on patient medical records.
How Artificial Intelligence is Applied in Cancer Diagnosis
2. Using AI to Detect Cancer in Medical Images: There are many applications of AI in medical imaging. In 2023, the FDA approved 122 AI and ML tools solely for radiology. Detecting and classifying cancerous tumors is one of the most notable use cases.
For example, Providence Health System teamed up with the University of Washington and Microsoft to create Prov-GigaPath—an AI model that can detect cancer in medical images and tissue samples. This algorithm was trained on over a billion pathology image fragments from 30,000 Providence patients, an impressively large dataset for training.
Prov-GigaPath is an open-access model that people around the world can use.
3. Using AI to Detect Cancer in Blood Tests: Blood tests using AI can help doctors more accurately detect cancer. A study published in Cancer Cell International states that blood profiling, where AI algorithms analyze plasma cfDNA and mRNA profiles, is a superior method for detecting and monitoring cancer compared to conventional CT scans.
Researchers from the Kimmel Cancer Center at Johns Hopkins University developed a new AI-based technology for diagnosing lung cancer using blood tests. This approach was tested on blood samples from 796 people in the US, Denmark, and the Netherlands. The researchers combined this blood test with protein biomarkers, patients' clinical risk factors, and CT scans. As a result, they accurately detected cancer in 91% of patients with early-stage disease and 96% of patients with late-stage cancer.
4. Using AI for Self-Diagnosis Applications: AI in cancer diagnosis can help people get initial feedback on their conditions through self-diagnosis without needing a doctor's appointment. Although diagnoses made with such tools are not definitive and do not replace a visit to the doctor, AI-based self-diagnosis solutions can promote early cancer detection among a broader population.
For example, SkinVision, a digital health startup from Amsterdam, developed a mobile app that helps users check skin abnormalities for cancer. The app can take a picture of a suspicious spot on the skin using a smartphone camera and send it for evaluation. The AI algorithm analyzes the color, texture, and shape of the lesion and provides feedback to the user within 30 seconds. The algorithm is said to offer 95% accuracy in detecting skin cancer.
How AI is Applied in Cancer Treatment
5. Using AI in Immunotherapy: Immunotherapy strengthens the immune system, enabling it to destroy cancer cells. For example, the UK-based biotech company Etcembly uses its proprietary Gen AI solution, EMLy, to scan the genetic makeup of T-cell receptors (part of the body's immune system). EMLy, based on a large language model, can analyze millions of lines of genetic code and identify which receptors have the greatest potential to bind with and destroy cancer cells without harming healthy cells.
6. Using AI in Drug Development: Traditional clinical trials for new drugs can take 10-15 years to complete. AI can significantly speed up this process, including drug development and discovery. For example, the Cancer Research Institute partnered with IDIBELL and the UK-based biotech company Vivan Therapeutics to overcome drug resistance with the help of AI and produce cancer treatments focusing on the cancer-instigating protein KRAS.
An industrial Gen AI startup, Zapata AI, is also working on destroying KRAS. They used Gen AI running on quantum hardware to produce over a million candidate drugs. Then the researchers filtered the results to eliminate molecules with less potential. The 15 remaining molecules were tested through cell-based essays. The successful compounds showed superior properties and were very different from the existing KRAS exhibitors.
7. Using AI in Genome Sequencing: AI algorithms allow doctors to better understand the genomic sequences of tumors and develop personalized treatment plans for patients. For example, the Dana-Farber Cancer Institute, in collaboration with MIT, developed a machine learning model called OncoNPC, which can identify the origin of metastases based on genomic data from the tumor with high accuracy. In these cases, the accuracy level surpassed 95%.
With these promising results, the researchers were encouraged to apply the model in real life. OncoNPC analyzed 900 unknown tumors and classified 40% of them with high confidence.
How Artificial Intelligence is Applied in Cancer Research and Monitoring
8. Using AI to Support Cancer Research: Generative AI is good at processing unstructured data, which humans and traditional search methods struggle to handle. Large language models, such as BioGPT, can transfer unstructured doctor’s notes into organized, easy-to-process information. These models can also search published medical literature, looking for connections between different factors, and organize everything into knowledge graphs.
In an attempt to facilitate cancer research, the Moffitt Cancer Center teamed up with Deloitte, Oracle, and Nvidia to produce an AI-powered platform that can identify and document health conditions that impact cancer care.
9. Using AI to Improve Patient Care: Artificial intelligence’s role in cancer treatment expands to offer platforms that assist cancer patients and help them communicate with their physicians.
For instance, Hurone AI used Amazon’s machine learning platform Bedrock to build its own AI solution that connects patients diagnosed with cancer to qualified doctors. The tool prompts patients to answer specific questions so that doctors can get a better idea of their state. The AI also saves all interactions in the corresponding EHR entries.
Barriers to adopting AI in cancer detection and treatment, or why pathologists are afraid to use the technology
AI brings about many challenges. Here are the most relevant ones for the healthcare sector.
1. Biased training data Unfortunately, bias is common among AI models. They can discriminate against certain ethnic groups and even against hospitals. Algorithms that work well for one care center may perform poorly when transferred elsewhere. A research team from the University of Chicago demonstrated how a machine learning-powered cancer detection application taught itself to consider the medical institution submitting the image as a factor in determining whether the scan shows signs of cancer.
Bias is also painfully common when diagnosing skin cancer in people of color. Many AI models are still trained predominantly on white people, as the medical community fails to gather enough representative data for diverse skin colors.
2. Fear of Replacement: Some pathologists are afraid that by working with AI on cancer treatment, they are simply training their replacement. They hear quotes, such as, “An AI algorithm can learn from a much larger library than a radiologist can. In some cases, a million images or more.” And they start worrying and thinking that AI can surpass them in everything they do.
The reality is that AI can be great at one task or at a few tasks, but it will not replicate pathologists’ scope of work. So, AI won’t replace doctors, but doctors who use AI might replace the ones who don’t.
3. Data Collection and Management Challenges: Healthcare data is typically stored in heterogeneous and unstructured ways, and it’s challenging to standardize terminology across medical facilities.
Initiatives, such as patient-reported outcome measures (PROMs), allow collecting standardized data from the start when patients experience distress. However, putting pressure on physicians to collect more data can lead to burnout and other increased workload problems.
PotentiaMetrics, a healthcare analytics company, employed JustSoftLab to develop an AI-enabled web portal for collecting, managing, and presenting patient data. The platform gathered and maintained patient information from the moment of diagnosis and throughout their survivor journey. Users would enter their data through a web-based questionnaire and generate reports. Also, the platform helped physicians craft personalized treatment plans based on the detailed information they received.
4. Lack of Training Data: More training data leads to higher system accuracy, and this is a challenge in the healthcare sector. The existing medical image datasets are significantly smaller than natural image sets. For example, the LUNA dataset of CT images contains only 888 instances, and the Indian Diabetic Retinopathy Image Dataset (IDRiD) includes approximately 600 images.
One way to address this issue is to synthesize medical datasets using generative AI. But it will consume considerable computational power, and a doctor still needs to review the results.
5. Ethical Issues and Considerations: Artificial intelligence algorithms for cancer prediction and treatment are often a black box. Pathologists and even researchers who developed and trained the model can’t explain how it delivers its outcome. For instance, when AI-powered software identifies the optimal cancer treatment for a particular patient, it doesn’t explain how it inferred this information.
Hospitals can use explainable AI for cancer detection and treatment, where algorithms reveal the reasoning behind their decision making. This will make it easier for doctors to act upon the recommendations and explain them to patients. However, turning AI into a white box will take away some of its predictive power. So, it’s a tradeoff that healthcare organizations will have to consider making.
There are also concerns about medical data ownership and obtaining patient consent when using their information. For example, the University of Chicago shared medical records with Google to help them develop an AI-powered predictive EHR environment. As a result, both parties were slammed with a lawsuit for patient data misuse.
To summarize
Healthcare organizations need to be cautious when using AI to detect and prevent cancer. Artificial intelligence is a powerful tool that can save patients’ lives and physicians’ time. But using AI can also have devastating consequences if it’s not trained and deployed correctly. Contract reliable AI consultants with healthcare experience to help you build and deploy the technology.
Whether working alone or with a team of external AI experts, your company could significantly increase your chances of success by following these steps:
Make sure your AI tool for cancer detection and treatment doesn’t contain any rooted bias
Schedule regular audits to remove any bias the algorithm may acquire while learning
Invest in data collection and organization. With this solid basis, it will be easier to initiate other AI projects.
Consider using explainable AI to fight cancer if the black-box concept is challenging to implement
Address your employees’ fear of being replaced by intelligent algorithms
Supply your employees with a detailed guide on how to use AI-powered tools and who bears responsibility for the final decision
At JustSoftLab, we offer an AI concept development service that allows healthcare organizations to experiment with AI before embarking on large-scale projects. You can learn more about this process on our blog, as well as explore our guide on implementing AI and estimating the costs of such initiatives.
#AI #ArtificialIntelligence #CancerTreatment #HealthcareInnovation #MedTech #Oncology #CancerResearch #Immunotherapy #Genomics #MedicalAI #HealthcareTech #DigitalHealth #PrecisionMedicine #CancerDetection #AIinHealthcare #FutureOfMedicine #SmartHealthcare #TechInMedicine #CancerCare #MedicalInnovation #JustSoftLab