Diagnosi del cancro basata sull'intelligenza artificiale: un sistema di intelligenza artificiale rivoluzionario

Artificial intelligence advances have made it possible to not only diagnose serious illnesses such as cancer, but to also predict their course. Harvard Medical School (Nuova finestra)recently revealed a multi-purpose AI model named Chief that can be used as a ChatGPT to diagnose and predict outcomes for different cancer types.

The majority of AI cancer systems focus on specific tasks. For example, identifying cancer cells and calculating a tumor’s genetic mutations. The Chief model is much more advanced. The Chief model can be used to diagnose many types of cancer, predict the patient’s survival rate and provide advice on suitable treatment.

A personalized vision of treatment

The model uses image analysis to predict the genetic profile of cancerous tissues with high accuracy. Chief was tested on 19 (Nuova finestra), cancer types (Nuova finestra),. Its flexibility is comparable to that of ChatGPT. The “tumor-microenvironment” can be identified. The tissue that surrounds a cancer is crucial to the response of treatment.

Chief can tell if a patient is more likely to respond well to radiotherapy, surgery, or chemotherapy. Its ability to accurately predict the response of a cancer to conventional treatments is one of its greatest breakthroughs. But it also has the ability to identify tumor traits that were not previously recognized to be linked with patient survival. The ability to personalize treatment can help guide the development of new targeted treatments for patients who are resistant to conventional options. This tool identified new cancer characteristics that were linked with patient survival. It demonstrates the power of AI-based methods to evaluate cancers effectively and to identify patients who are less responsive to standard treatment.

Eventually bridging resource disparities in the fight against cancer

Researchers at Harvard Medical School (Nuova finestra) developed a large knowledge base by training “Chief”. They used 15,000,000 tissue images. Researchers then trained the model with 60 000 slides of tissue from different organ types. This allowed it to analyse both complete and specific images. It can then associate the details of the image with their global context. This is crucial for accurately assessing cancer characteristics. It was then tested on 19 400 images of patients from around the world, where it outperformed existing AI methods in terms of detecting cancer, identifying tumour origin and even predicting mutations that are associated with treatment responses by 36%.

Researchers plan to train “Chief” on rare diseases, pre-cancerous tissue images that could help with early detection of cancer and prevention. The researchers also plan to incorporate more molecular information to differentiate cancers of varying degrees of aggression and predict the benefits of treatment as well as the side effects. To bridge resource gaps and improve cancer care, the ultimate goal of “Chief”, is making it accessible to all.

Flussi di lavoro personalizzati utilizzando Automazione Make.com