Oral cancer is a common type of cancer. Most patients are in the advanced stages and the cancer cells may have spread to other parts of the body. About 50% of patients survive less than 5 years. Recently, an AI platform was jointly developed by researchers from the Faculty of Dentistry, the Faculty of Dentistry, the Department of Pathology at Queensland Hospital, and the School of Medicine and Dentistry at James Cook University in Queensland. As long as the input data have a cancer risk of 20 years, its accuracy is over 80%!

Oral cancer is difficult to diagnose and is easy for patients to abandon

The initial symptoms of oral cancer are mostly some leukoplakia in the oral cavity, and the clinical diagnosis is oral leukoplakia OL and oral lichenoid mucositis OLM. Patients generally require regular follow-up visits and multiple biopsy check-ups. After years of continuous monitoring, many patients experience fatigue and even refuse follow-up visits.

The patient had lingual lichen planus mucositis.

(Image source: HKU Faculty of Medicine website)

The patient had buccal leukoplakia.

(Image source: HKU Faculty of Medicine website)

How to use?

The current global OL / OLM cancer risk ratio ranges from 0.4% to 40.8%, according to the University of Hong Kong School of Dentistry. The research team has developed an online platform to predict the risk of oral cancer lesions.The forecast period is up to 20 years, Currently free to use by medical staff and the general public! Each is required to provide 26 types of information, including personal background information, clinical and pathological description of the disease, and treatment received.The platform presents a curve showing the predicted danger level each time.When the patient reaches a certain level of risk, the medical staff can begin to monitor the patient in an encrypted manner; With it the patient can understand the tendency of the condition and remind himself to follow up regularly or perform further tests.

(Screenshot of the platform website)

According to the University of Hong Kong, doctors and dentists can compare the risk of cancer in two or more OL / OLM patients, depending on how high and low their curvature. In general, as the risk goes from 0.5 (blue dashed line) to 1.0, the risk of cancer decreases and decreases.

(Image source: HKU Faculty of Medicine website)

Can be predicted risk levelMaintain accuracy for 17 years from the date of entryHowever, the predicted risk curve changes as the data change, i.e., the level of smoking and alcohol consumption, recurrence, and the amount of oral epithelial change during treatment monitoring, so the data may need to be updated periodically.

What is the principle? Is that accurate?

Everyone’s most concerned about how accurate the forecast is! PlatformAdopt the “DeepSurv” Artificial Intelligence Algorithm, The algorithm can predict the risk of cancer based on personal statistics, clinical and pathological data. The research team used patient data from OL and OLM in Hong Kong to train AI and test its accuracy, in Newcastle, UK. Showing results,Data from 716 patients used for evaluation in Hong Kong, “DeepSurv” can accurately predict the cancer risk level of 95% of patients; Data from 382 patients in the UK, the forecast accuracy is 82%!

The team hopes the new platform will help prevent oral cancer and improve early diagnosis, but the development platform is primarily research-based and needs to be further improved.

Dr. Zhou Yuxiong, Associate Professor of Clinical Oral and Maxillofacial Surgery, HKU Faculty of Dentistry; Dr. Kai Soihohui, Assistant Professor of Research; Dr. John Audio, PhD student; Dr. Cheng Livu, Clinical Associate Professor of Oral Science;
(Image source: HKU Faculty of Dentistry website)

HKU AI Platform

Cost:Sow birth

Chinese versionhttps://pred-facdent-hku-deepsurv-cn.herokuapp.com/

English versionhttps://opmd-pred-facdent-hku-deepsurv.herokuapp.com/

Source:University of Hong Kong Faculty of Dentistry

[പൊതുവോട്ട്]Participate in the selection of etnet’s most popular column and program now to get rich rewards!
Quick vote