Integrating artificial intelligence (AI) in healthcare transformed the detection and treatment of diseases, including cancer. This essay examines how AI aids in detecting and battling cancer. It discusses bumps in the road and provides a balanced perspective on AI’s role in modern medicine. The essay reviews doubts and questions we all ask ourselves. Is AI as a future of detecting and fighting cancer a tool supporting discovery? Or is it slowing research down?
Let’s talk!
Google DeepMind united with Google Brain, a robotic research department. Its sole purpose was to teach AI from an empirical experience. It was based on the same mechanism humans use to learn from their own life experiences.
To delve further, deep learning is a specialized subset of machine learning implemented in this process. It leverages artificial neural networks to extract patterns and insights from vast datasets. These networks mimic the structure of the human nervous system. Interconnected nodes adjust their weights as learning progresses.
This dynamic adaptation enhances AI’s ability to classify data efficiently. It’s significantly improving performance across various domains. These domains include image recognition, natural language processing, and speech recognition.
MORE ABOUT Established types of deep learning techniques
-> Convolutional neural networks (CNNs) excel at identifying objects in images, even when those objects are partially obscured or distorted. Through further examination, this branch of AI employs convolution layers. It utilizes pooling layers to extract features from images and videos. Then, it uses these features to classify or detect objects or scenes.
-> Deep reinforcement learning allows an AI to learn how to behave in an environment through interactions. The AI, similar to a human being, receives rewards or punishments based on its actions. It interacts with an environment imperatively by making choices that maximize cumulative rewards. The process enables AI to learn sophisticated strategies and can directly learn rules from sensory inputs. This approach uses deep learning’s ability to extract complex features from unstructured data, like distorted or unclear image scans.
-> A recurrent neural network or RNN is a deep neural network, trained on sequential or time series data. This model can make sequential predictions or conclusions based on sequential inputs. Traditional deep-learning networks assume that inputs and outputs are independent of each other. Nevertheless, the output of recurrent neural networks depends on the prior elements within the sequence. But this network has its own limitations, it’s not obsolete. RNNs were popular for sequential data processing because of their ability to handle temporal dependencies. Yet, the network can’t maneuver within accelerating gradient problems.
So why medicine?
There are ongoing challenges and occasional AI-related glitches. Despite these, personal accounts of perseverance and innovation often drive groundbreaking discoveries in the field.
One of these was the story of Eric Lefkofsky, founder of the famous AI platform, Tempus. His wife was diagnosed with breast cancer, which motivated him to focus on perfecting oncologic diagnostic tools. Alongside Ryan Fukushima, they started building a world-class team. They focused on creating the first version of a platform. This platform is capable of ingesting real-time healthcare data.
Paramount techs this duo introduced are IMPRESSIVE!
- ONE quickly accesses patient information. This includes report status, results on actionable biomarkers, and MSI/TMB status. Unlike most GPs who use Google to find matching diagnoses, or mislabel ovarian cancer pain as a period problem, Tempus AI builds a data cohort. This cohort includes rich molecular biomarkers and therapies. Coding experience is not required and there is no disregard for any information.
- NOW reads discrete genomics data, which can power decision-making, advanced analytics, and clinical research.
- PIXEL offers AI-enabled insights from medical images that link lesions across time points. They create longitudinal tracking reports to calculate lesion response automatically. Then automates therapy response criteria and generates a comprehensive report.
- ASSAYS is a genomic profiling service. It encompasses a broad range of sequencing options. These options include tests for molecular profiling, allowing for data-driven patient decisions.
- ALGOS gains extra insights across multiple cancer types through algorithmic testing options. The software scans prognostic biomarkers for immune checkpoint inhibitor candidates and measures homologous recombination deficiency. One of the strongest suits of this technology is its ability to refine diagnosis for cancers with uncertain origins (while this issue is particularly problematic for publicly funded hospitals around the world).
Advancements in Drug Development
Collaboration between AI researchers and pharmaceutical scientists is essential for advancing drug discovery and development. By integrating their skills, they can develop sophisticated machine-learning models designed to predict the efficacy of potential drug candidates.
These AI-driven algorithms accelerate the drug discovery process and enhance the accuracy and efficiency of clinical trials. AI analyzes vast datasets during these trials. It recognizes patterns and detects potential adverse effects. AI also supports pharmaceutical companies in making informed decisions about which drug candidates to rank. This ultimately streamlines the overall drug development process, reducing both time and costs.
Beyond its role in gene sequencing, AI addresses the complexities of cancer’s molecular landscape. Effective cancer therapies must target multiple pathways while minimizing resistance. AI models identify conserved binding sites, analyze mutation patterns, and predict adaptive resistance mechanisms. This enables the design of drugs that remain effective against evolving cancer cells.
Additionally, AI is transforming medicinal chemistry by predicting the efficacy and toxicity of potential drug compounds. Traditional drug discovery methods are labor-intensive, requiring extensive experimentation to assess a compound’s potential effects on the human body. This process is often slow, costly, and subject to a high degree of variability. AI-driven models significantly expedite this process by rapidly screening vast chemical libraries and identifying promising candidates with greater precision.
Finally, the critical application of AI in drug discovery is the identification of drug-drug interactions. These interactions occur when multiple drugs are administered simultaneously to treat the same or different conditions in one patient. This can lead to altered effects or adverse reactions. AI algorithms can analyze vast pharmacological datasets to predict and mitigate these risks, ensuring safer and more effective treatment regimens.
Thus, any cons?
Despite its advantages, the use of AI in oncology is not without limitations and risks. Several challenges need to be addressed to maximize its potential. AI systems need vast amounts of high-quality data to function effectively. Inadequate or biased datasets can compromise accuracy and perpetuate disparities in care. Training data that lack diversity results in algorithms that are less effective for underrepresented populations, exacerbating existing health inequities.
TRUSTed DATA?
- Ancestry bias is prevalent in many AI platforms, as so-called gold-standard genomic datasets significantly underrepresented non-European populations. This lack of diversity leads to disparities in research outcomes and limits our comprehensive understanding of human diseases, including cancer.
- Genomic data bias shows that AI can generate inaccurate predictions. This happens when AI is applied to patients from diverse backgrounds. These errors arise primarily due to the lack of representative training data, which can result in misdiagnoses and suboptimal treatment recommendations.
- Demographic bias encompasses factors such as sex, age, language, and disability status, as well as socioeconomic variables including income, educational attainment, and access to healthcare. When datasets do not represent a broad range of demographic backgrounds, AI algorithms suffer from diminished predictive accuracy. This bias can lead to delayed diagnoses of life-threatening conditions such as skin cancer.
- Additionally, systemic prejudices and racial disparities contribute to worsening health outcomes across diverse patient populations. Addressing these biases requires the implementation of equitable data collection strategies. It also involves developing AI models that prioritize fairness and inclusivity.
- Methodological bias or mismanagement of data can significantly impact research reliability. Arguments challenging the value-free ideal of science have raised an important question. How can we distinguish between legitimate values that enhance research? How do we identify biases that compromise its integrity?
- Bias can emerge at any stage of the research process, including study design, data collection, analysis, and publication. Identifying and mitigating these biases is crucial. This ensures that scientific advancements, particularly in AI-driven drug discovery, remain robust, credible, and ethically sound.
Biases embedded in AI models, whether due to training data limitations or algorithmic design, can skew outcomes. This skewing leads to disparities in treatment recommendations. Addressing these challenges requires ongoing efforts to enhance AI Interpretability, implement bias-mitigation strategies, and ensure ethical AI deployment in healthcare settings.
CONCLUSION
Artificial intelligence is transforming cancer detection and treatment, offering unprecedented opportunities to improve accuracy, efficiency, and personalization in oncology. However, its integration into clinical practice is accompanied by significant challenges, including data bias, financial barriers, and ethical considerations.
Stakeholders must emphasize transparency to fully harness AI’s potential. Equity and interdisciplinary collaboration are also crucial. These actions ensure that technological advancements benefit all patient populations. By balancing innovation and ethical responsibility, AI can become a fundamental tool in the fight against cancer. It can ultimately save lives and improve global health outcomes.
REFERENCES:
The Economist (2024). Artificial intelligence is taking over drug development. [online] The Economist. Available at: https://www.economist.com/technology-quarterly/2024/03/27/artificial-intelligence-is-taking-over-drug-development?utm_medium=cpc.adword.pd&utm_source=google&ppccampaignID=18156330227&ppcadID=&utm_campaign=a.22brand_pmax&utm_content=conversion.direct-response.anonymous&gad_source=1&gclid=Cj0KCQiA4-
Brazil, R. (2024). How AI is transforming drug discovery. [online] The Pharmaceutical Journal. Available at: https://pharmaceutical-journal.com/article/feature/how-ai-is-transforming-drug-discovery
Pfizer (2021). Artificial Intelligence: on a Mission to Make Clinical Drug Development Faster and Smarter | Pfizer. [online] Pfizer.com. Available at: https://www.pfizer.com/news/articles/artificial_intelligence_on_a_mission_to_make_clinical_drug_development_faster_and_smarter
National Cancer Institute (2023). Trusting the Data—A Look at Data Bias | CBIIT. [online] datascience.cancer.gov. Available at: https://datascience.cancer.gov/news-events/blog/trusting-data-look-data-bias
IBM (2021). Recurrent Neural Network (RNN). [online] Ibm.com. Available at: https://www.ibm.com/think/topics/recurrent-neural-networks
IBM (2021a). Convolutional Neural Networks. [online] Ibm.com. Available at: https://www.ibm.com/think/topics/convolutional-neural-networks
About The Author
