CAN ONE SCAN PREDICT THE FUTURE development of lung cancer? We know that screening (with CT scans) individuals — who currently smoke or recently quit — saves lives. Now, scientists are developing a so-called crystal ball of lung cancer using screening and artificial intelligence (AI). Today we explore how AI can predict lung cancer.
“For ethical AI systems, the question is, “Who decides what is ethical?” Well, starting from the developer, researchers, organizations, governments, and international bodies should always act according to their conscience and always in the best interests of humanity. Equal effort must be made to guarantee human safety, freedom, autonomy, and justice.”
― Sri Amit Ray, Ethical AI Systems: Frameworks, Principles and Advanced Practices
Artificial Intelligence Basics
Artificial intelligence (AI) has been increasingly used in medical imaging, including the interpretation of lung CT scans for cancer screening. AI algorithms can analyze and interpret large volumes of medical images with speed and accuracy. The approach can assist radiologists in detecting and diagnosing various conditions, including lung cancer.
The process of using AI for lung cancer screening typically involves the following steps:
- Data collection: A large dataset of lung CT scans and their corresponding clinical information is collected. Experts typically label these scans to indicate the presence or absence of cancer.
- Training the AI model: The collected dataset trains an AI model to learn lung cancer patterns in CT scans. A convolutional neural network is such a model. The model learns to identify and classify suspicious regions or nodules that may indicate the presence of cancer.
- Validation and fine-tuning: The trained model is validated using separate datasets to evaluate its performance. Fine-tuning and optimization techniques are employed to improve the model’s accuracy and reduce false positives and negatives.
- Deployment and integration: Once the AI model demonstrates good performance, it can be deployed into clinical practice. Clinicians may integrate the model into existing radiology workflow systems. The tool can then analyze CT scans automatically and provide radiologists with information about potential cancerous regions for further examination.
Artificial Intelligence and Lung Cancer
While AI can detect lung cancer in CT scans, it is typically a support tool for radiologists. It is not a standalone diagnostic tool. Radiologists still play a crucial role in reviewing AI-generated findings and making the final diagnosis.
AI in lung cancer screening can improve efficiency and accuracy in detecting suspicious lesions. This improvement can lead to earlier detection and better patient outcomes. It is essential to continue refining and validating AI models. We need further research and clinical trials to ensure their safety and effectiveness before widespread implementation.
Artificial Intelligence in Imaging
A new study explores using a validated deep learning model of artificial intelligence (AI). Can AI predict future lung cancer risk using a single low-dose lung CT scan? Let’s look at some basic terms before turning to the groundbreaking study demonstrating this remarkable breakthrough.
Researchers introduced the term artificial intelligence in the 1950s. AI is the science and engineering of making intelligent machines, especially intelligent computers.
Artificial intelligence became important in imaging (such as CT scans) as deep machine learning (DL) became more powerful. DL uses predictive algorithms using large datasets of images. These datasets are independently validated using similarly large datasets.
Deep learning is a form of artificial intelligence. Trainers feed hundreds of thousands of observations from large image data sets into a computer, and the machine develops pattern recognition.
Artificial Intelligence to Predict Lung Cancer
Artificial intelligence can see beyond small nodules/tumors to predict the future development of lung cancer. Deep learning is the leading tool for medical image analysis. I recently learned of a groundbreaking studying applying AI to lung CT imaging.
First, researchers developed a model (and internally validated it) using over 12,000 low-dose CT scans from the US National Lung Screening Trial (NLST). The experts externally validated the model (using two separate data sets with a cumulative more than 23,000 CT screening studies).
Here are the stunning results: Artificial intelligence and a deep learning (DL) model (“Sybil”) appears to predict an individual’s future lung cancer risk after a single baseline computed tomography (CT) chest scan. The AI sees more than nodules; it notices tissue changes in the area of a future nodule.
The findings were impressive out to six years. While the AI performance declined after the researchers removed lung nodules, the AI still could predict future cancer. The model worked for all subsets, including sex, age, and smoking history.
Why the name Sybil for the artificial intelligence tool? Some geeky acronym for something involving deep learning? Nope. Sybil was the prophetess (in Greek and Roman mythology) who could foresee the future.
Key points — AI Can Predict Lung Cancer.
Question. Can artificial imaging applied to medical imaging predict the future development of lung cancer?
Findings. Artificial intelligence and a deep learning (DL) model (“Sybil”) appear to predict future lung cancer risk after a single baseline computed tomography (CT) chest scan. The AI sees more than nodules; it notices tissue changes in the area of a future nodule.
Meaning. Understanding who would benefit from this groundbreaking technology will require significant investment in prospective studies targeting groups with various lung cancer risk profiles.
On another topic, if you know someone with breast cancer, here is my comprehensive online course:
Breast Cancer – What a Patient Needs to Know.
The information I provide in this blog is for educational purposes only and does not substitute for professional medical advice. Please consult a medical professional or healthcare provider for medical advice, diagnoses, or treatment. I am not liable for risks or issues associated with using or acting upon the information in this blog.
Thank you for reading “The Crystal Ball of Lung Cancer: Can a Single Scan Reveal the Future?”
