Professional scientists from Harvard Medical recently developed a new AI cancer detection tool, called CHIEF (Clinical Histopathology Imaging Evaluation Foundation), which was made to detect cancer cells/ tumors, predict the spreading/life of the tumor, and determine the right cancer treatments, at a lab at Harvard Medical School. This impressive piece of technology was recently set in motion in September 2024, after months of the AI being trained by scientists, using different images and slides to teach it to detect cancer in those images.
CHIEF is different from other AI tools because of its ability to predict what a cancer patient needs to survive and determine/ predict how certain treatments may change the DNA sequence of a person, along with being able to detect cancer cells and find out how a tumor originated.
The AI model can detect 11 different types of cancer and can be trained to detect more. It can also detect cancer in different body types and different patients across the globe.
CHIEF was able to 96% accurately detect cancer in the esophagus, stomach, and many other organs. CHIEF also achieved 90% accuracy in being able to identify cancer in the colon, lung, and many other organs, in which doctors were not able to identify cancer.
CHIEF is one of the first and most accurate cancer AI tools to predict patient survival and useful treatments for certain tumors and cancer cells. The AI tool can tell if a tumor should be treated through surgery, chemotherapy, radiation, immunotherapy, etc. It can predict how long the patient may live and decide if the tumor has a chance of spreading. The tool can also identify characteristics of a tumor, such as how fast it will grow/ spread, what its structure is, etc.
CHIEF achieves this by interpreting images and slides of tumor tissues in the area that needs to be looked over. Then, it detects a cancer cell and predicts if the tumor can be treated and how it can be treated. Since different cancer treatments change the sequencing of a person’s DNA for the better or the worse, and different treatments also cause different mutations in genes, CHIEF can predict what mutations can harm a person and what mutations will be beneficial to them. CHIEF can also see how many more immunity cells a person has, to be able to fight cancer and handle various types of treatments.
Doctors are not able to do what CHIEF does, as quickly as it does. Doctors can send samples of a person’s tumor to a DNA sequencing lab for them to test out different treatments of it; however, CHIEF can identify this within seconds, since it’s able to run those same tests, without a lab. CHIEF bases part of its predictions after interpreting how the cancer originated. The treatments that CHIEF has recommended have worked successfully.
This tool would be able to save millions of lives. Ashish A, a family member of someone who recently passed away from lung cancer, stated, “My father was diagnosed with stage 4 cancer a few years ago, and recently passed away because his cancer spread. He went through a lot of chemotherapy, but his cancer kept reappearing because doctors had caught on to the cancer too late. A device like this could have possibly let us know that he had cancer at stage 1 or 2, which could’ve been more easily treated.”
Training this phenomenal AI tool took a lot of effort. The model was trained by interpreting 15 million random images of specific parts of each organ to detect cancer. It later was trained by looking at 6,000 broader images and was trained to detect cancer from any type of image. This is why CHIEF can also detect different types of cancer on specific parts of the body, or even on a whole-body image. The findings for cancer, no matter the size of the image, can point out the area of the body that needs to be closely observed.
Based on all of the actions that CHIEF can perform with accuracy, Scientists explained that CHIEF says their new AI tool will be helpful to clinics, hospitals, and many other medical facilities across the world.
This tool has already set up a future of continued success in the healthcare industry. Kun-Hsing Yu, an Assistant Professor of Biomedical Informatics at the Blavatnik Institute at Harvard Medical School, stated, “If validated further and deployed widely, our approach, and approaches similar to ours, could identify early on cancer patients who may benefit from experimental treatments targeting certain molecular variations, a capability that is not uniformly available across the world.”
