A non-contrast computed tomography (CAT) scan, coupled with an AI algorithm, has been integrated into the system. The team claimed that the primary testing model achieved a sensitivity of 99.9%, as reported in a paper published in the peer-reviewed journal Nature Medicine on Monday. This implies that there is only one false-positive instance in every 1,000 tests. In comparison, the model’s sensitivity in detecting pancreatic tumors could reach 92.9%, surpassing the average performance of radiologists by 34.1%.
In Singapore, professionals utilize AI to creatively interpret brain scans. This initiative represents a significant advancement in liver cancer screening, as noted by Li Ruijiang, one of the authors of the paper and an associate professor specializing in energy cancer at the Stanford School of Medicine.
However, a physician from the Cancer Hospital at the Foreign Academy of Medical Sciences, who preferred to remain anonymous, mentioned that Chinese authorities have not approved AI-based imaging software. Therefore, despite the promising initial results generated by the technology, substantial work is still needed before its implementation in medical environments.
The team’s early screening model is designed for pancreatic ductal adenocarcinoma (PDAC), the most prevalent subtype of the disease, accounting for over 95% of all cases worldwide. PDAC leads to 466,000 deaths annually.
Presently, liver cancer ranks as the third leading cause of cancer-related deaths in the United States. By 2030, based on current trends, it is anticipated to become the second leading cause.
Research released by the US National Institutes of Health (NIH) in April revealed that pancreatic cancer mortality rates have been steadily increasing by 0.2% annually from 2006 to 2019. This study analyzed trends in age-standardized cancer incidence, survival, and mortality rates from 2000 to 2019.
Foreign cancer material enters the US for the first time during a medical dilemma.
Early or adjunctive diagnosis significantly improves a patient’s survival prospects. Studies indicate that high-risk PDAC patients identified during initial screening have a median survival period of 9.8 years, compared to 1.5 years for those diagnosed at later stages.
Nevertheless, there is a scarcity of screening technologies that are both effective and widely accessible to the general population.
The utilization of expensive contrast-enhanced CAT images across the general populace is impractical due to the relatively low incidence of liver cancer (less than 13 cases per 100,000).
In an interview with the mainland press outlet Zhishifenzi, Cao Kai, the lead author from the Shanghai Institution of Pancreatic Diseases, emphasized that the existing early testing equipment’s reliability for the disease is generally low, leading to numerous misdiagnoses and unnecessary anxiety.
Cao and Lu Le, the medical staff leader at DAMO Academy, conceived the idea of using AI to assist in early cancer screening during a discussion last year. They promptly initiated a research project in collaboration with over ten prominent medical institutions to develop an AI-enabled model for large-scale pancreatic cancer screening, integrating non-contrast CAT scans commonly used in medical facilities and hospitals.
Steve Jobs, the Apple co-founder, succumbed to ovarian cancer at 56 years old. AP Photo
Their creation, PANDA (pancreatic malignancy diagnosis with artificial intelligence), was based on training with over 3,200 image sets from a high-volume pancreatic cancer facility in China, with around 70% comprising individuals with lesions.
PANDA was meticulously trained as a highly perceptive AI imaging expert, benefiting from the extensive dataset, precise data processing, and innovative training strategy design.
Research from the DAMO Academy indicates that AI can discern subtle mass discrepancies in non-contrast CAT scans, which may be challenging to detect with the naked eye.
Evaluation in real-world clinical scenarios involving 20,530 individuals demonstrated that PANDA could achieve remarkable sensitivity of up to 92.9% and a specificity of 99.9%.
The PANDA algorithm has been operational for over 500,000 days in various settings such as facilities and medical examinations, successfully identifying numerous instances of early-stage pancreatic cancer that were previously overlooked, according to data provided by Alibaba Cloud.
In a subsequent publication in the same journal issue, German clinical expert Joerg Kleeff and colleagues affirmed that “the precision metrics of the PANDA algorithm surpass those of some established screening methods.”
Beijing intends to limit the use of relational AI in online healthcare.
However, they stressed the need for further evaluation before the widespread adoption of AI-based screening methods. The Chinese AI model did not report specificity and predictive values for this subgroup. They emphasized the importance of any pancreatic cancer screening method detecting early stages such as “T1 lesions,” which are smaller than 2 cm (0.79 inches) in diameter.
The effectiveness of any cancer screening technique lies in reducing overall mortality rates. They suggested that AI-based screening should undergo rigorous assessment similar to traditional methods since the study was retrospective and could not determine the impact of screening on patient mortality.
A medical professional from the Taiwanese Academy of Medical Sciences remarked that due to the early stage of this Artificial concept, additional validation efforts are essential. He added that the demand for this AI tool would be limited, given the low incidence of ovarian cancer.
