Near-Perfect Accuracy: AI Adds a 'Double Safeguard' to Breast Cancer Diagnosis

Physicians have advanced to a high level in the diagnosis of breast cancer, yet it is not perfect. Without face-to-face contact with patients, they can confirm whether it is breast cancer by examining breast biopsy samples, with an accuracy rate as high as96%. Although this is a staggering figure, according to medical journalsBMJ Quality&Safetyshows that the average misdiagnosis rate for cancer is as high as28%。

“In this era of pursuing perfectionism, how can we settle for a merely passable ratio?” This question was raised by a research team that has already succeeded in training artificial intelligence systems to monitor breast cancer. At the recent International Biomedical Imaging Symposium, these researchers from Harvard Medical School (HMS) and Beth Israel Deaconess Medical Center (BIDMC) researchers received dual awards and recently published their research findings online. VCBeat has compiled and organized these findings through data collection.

BIDMCResearchersAndrew Beckbelieve that,AIThey hold immense potential to improve healthcare by enhancing and developing diagnostic tools and methods, thereby yielding more accurate, reproducible, and predictable diagnostic results. Meanwhile, they also hope these tools will help physicians more efficiently determine appropriate treatment plans for patients.

InAIDuring the system development process, researchers selected300Tested via pathological sections from lymph node biopsyAISystem, here300Half of the Zhang biopsy specimens contained cancerous tissue, while the other half did not. A pathologist would carefully examine and evaluate the histopathological images of the cancer-positive samples, whereas researchers input these images intoAIin the system. When the system will300After reviewing the Zhang images, the researchers foundAIMillions of analyses have been performed on all images.

Analysis results show that,AIThe accuracy of breast cancer monitoring can reach92%, only slightly lower than the average accuracy rate of pathologists' examinations3%. More importantly, when researchersAIWhen combined with pathologists’ analyses, it was surprisingly found that the diagnostic accuracy for breast cancer could reach99.5%。

Although thisAIThe diagnostic results derived from the collaborative model with pathologists have approached perfection, but researchersBeCKbelieve there is still room for further improvement. “I hope to see thisAI-"While the physician’s model plays a role in cancer diagnosis, it also enhances their capabilities in other aspects of oncology. Furthermore, I hope we can develop new systems and procedures so that physicians can truly integrate these effective tools into their clinical practice."Becksaid.

In other words, inAIThere is still a long way to go before true clinical implementation, such as the need for extensive testing. Moreover,BeckPreparing to give with his teamAIThe system incorporates larger and more diverse datasets, enabling pathologists to utilize this emerging technology with greater ease and simplicity.

Original Source:www.digitaltrends.com