NTUH's AI-Assisted Model for Early Detection of Pancreatic Cancer on CT Scans

Taipei, Taiwan Taiwan, Province of China[a]
Early detection is key to ensuring a better prognosis for pancreatic cancer patients.
Pancreatic cancer is a major health concern and the seventh-leading cause of cancer-related death in Taiwan. It is estimated to become the second leading cause in the US before 2030.
NTUH's AI-Assisted Model for Early Detection of Pancreatic Cancer on CT Scans

Pancreatic cancer is a major health concern, with the seventh-leading cause of cancer-related death in Taiwan and estimated to become the second leading cause in the US before 2030. Early detection is key to ensuring a better prognosis for pancreatic cancer patients. National Taiwan University Hospital (NTUH) has developed an artificial intelligence (AI)-assisted model that can accurately detect smaller pancreatic tumors and distinguish them from non-cancerous tissue on CT scans, providing radiologists and clinicians with a valuable tool in early detection of the disease.



Confidence

80%

Doubts
  • It is not clear if this model has been peer reviewed or published in a scientific journal.

Sources

79%

  • Unique Points
    • The MIT team developed computer models capable of identifying three and a half times more high-risk people than current standards by teasing out patterns common among those who went on to develop pancreatic cancer.
    • An AI deep-learning algorithm was introduced to help detect smaller pancreatic tumors, it can accurately distinguish pancreatic cancer tissue from non-cancerous tissue and aid radiologists and clinicians in recognizing suspicious lesions on CT scans
    • The full model accurately predicted disease survival in 87% of patients using 6,363 different biological data points including genetic and molecular information
  • Accuracy
    No Contradictions at Time Of Publication
  • Deception (70%)
    The article is deceptive in several ways. Firstly, the author claims that AI can predict pancreatic cancer and enlarge the group of patients who can benefit from screening. However, they do not provide any evidence to support this claim. Secondly, the author quotes a study published in The Lancet last month which states that their model combines through electronic health records to tease out patterns common among those who went on to develop pancreatic cancer such as changes in blood chemistry or more frequent doctor visits. However, they do not provide any details about how accurate this model is and what the false positive rate is. Thirdly, the author quotes a study published in May of last year which states that their AI system working with data from Denmark and VA Health Care system has similar results to MIT's study. However, they do not provide any evidence to support this claim.
    • The article claims that AI can predict pancreatic cancer but does not provide any evidence to support this claim.
  • Fallacies (85%)
    The article discusses the potential of using artificial intelligence to predict pancreatic cancer and increase early detection rates. The author cites a study by Appelbaum et al., which used machine learning algorithms to identify high-risk patients based on electronic health records. However, there are several fallacies present in this analysis.
    • The article states that the MIT team's model can detect risk for other hard-to-find cancers like ovarian cancer (paragraph 10). This is an example of a hasty generalization fallacy. The author assumes that because the AI system was successful in predicting pancreatic cancer, it will also be effective in identifying high-risk patients with other types of cancer without providing any evidence to support this claim.
    • The article mentions that the MIT team's model misses a lot of patients (paragraph 12). This is an example of a false dilemma fallacy. The author presents only two options: either the AI system works perfectly or it doesn't work at all, ignoring other possibilities and potential limitations.
    • The article states that the MIT team has already adapted their model for kidney cancer (paragraph 13). This is an example of a slippery slope fallacy. The author assumes that because the AI system was successful in predicting pancreatic cancer and adapting it for kidney cancer, it will be effective in identifying high-risk patients with other types of cancers without providing any evidence to support this claim.
  • Bias (100%)
    None Found At Time Of Publication
  • Site Conflicts Of Interest (50%)
    The article discusses the potential for AI to detect pancreatic cancer at an early stage. The author is Dr. Appelbaum who has a financial interest in TriNetX and MIT Computer Science & Artificial Intelligence Laboratory.
    • Author Conflicts Of Interest (0%)
      None Found At Time Of Publication

    83%

    • Unique Points
      • National Taiwan University Hospital (NTUH) has developed an artificial intelligence (AI)-assisted model for interpreting computed tomography (CT) scans to detect pancreatic cancer more accurately and at earlier stages
      • `Pancreatic cancer was the seventh-leading cause of cancer-related death in Taiwan in 2022, and it has been estimated that it could become the second-leading cause of cancer-related deaths in the US before 2030a
      • aEarly detection is key to ensure a better prognosis for pancreatic cancerǷ
      • ◨The five-year survival rate after surgical resection is about 8% if the tumor detected is 2cm or smaller, but the survival rate drops to less than % if it is detected when the tumor grows bigger than 2cmρ
      • People with early-stage pancreatic cancer usually have no symptoms and about δ of the 2cm or smaller tumors in the pancreas cannot be detected by a CT scan׷
      • 溘An AI deep-learning algorithm was introduced to help detect smaller pancreatic tumors, it can accurately distinguish pancreatic cancer tissue from non-cancerous tissue and aid radiologists and clinicians in recognizing suspicious lesions on CT scansα
      • 溘The sensitivity rate for detecting pancreatic tumors 2cm or smaller was about %׷
    • Accuracy
      • The article discusses the potential of using artificial intelligence to detect pancreatic cancer at an early stage.
      • A group of patients with a known genetic risk have a better prognosis and can survive up to 80% if their cancer is detected early enough.
    • Deception (100%)
      None Found At Time Of Publication
    • Fallacies (75%)
      The article contains several logical fallacies. The author uses an appeal to authority by citing the success of the AI-assisted model and its medical device license from FDA as evidence for its effectiveness. However, this does not necessarily mean that it is accurate or reliable in detecting pancreatic cancer accurately at earlier stages.
      • The sensitivity rate for detecting pancreatic tumors 2cm or smaller was about 80 percent
      • Endoscopic ultrasound-guided biopsies can assist in the diagnostic evaluation of benign or malignant tumors
    • Bias (85%)
      The article is biased towards the success of NTUH's pancreatic cancer team in developing an AI-assisted model for interpreting CT scans. The author uses language that deifies the team and their achievements, such as calling them a 'breakthrough device', earning awards, obtaining patents and having FDA medical device license. This creates a sense of heroism around the topic which is not objective.
      • NTUH said endoscopic ultrasound is also an important tool that their pancreatic cancer team use for clarifying whether there is actually a pancreatic tumor, its precise location and size
        • The AI-assisted diagnosis system has obtained a Food and Drug Administration (FDA) medical device license
          • The study on the AI-based method to provide early detection of pancreatic cancer also earned the Alexander R. Margulis Award last year for the best original scientific article published in Radiology.
            • The treatment of pancreatic cancer is more complicated than other cancers
            • Site Conflicts Of Interest (100%)
              None Found At Time Of Publication
            • Author Conflicts Of Interest (50%)
              The author has a conflict of interest on the topic of National Taiwan University Hospital (NTUH) as they are reporting on a breakthrough in pancreatic cancer treatment developed by NTUH. The article also mentions Liao Wei-chih who is an associate professor at NTUH and received the Alexander R. Margulis Award for his work in AI, which could indicate that he has a financial stake or personal relationship with the university.

              70%

              • Unique Points
                • Molecular Twin Precision Oncology Platform launched by Cedars-Sinai in 2021
                • A majority of cancer patients are allowing researchers to include their clinical information and samples from blood, tumor and other sources so that they can continue to build the Molecular Twin platform
              • Accuracy
                • The article discusses the potential of using artificial intelligence to detect pancreatic cancer at an early stage.
                • A group of patients with a known genetic risk have a better prognosis and can survive up to 80% if their cancer is detected early enough.
              • Deception (50%)
                The article is deceptive in several ways. Firstly, the author claims that Molecular Twin outperforms the standard test for predicting pancreatic cancer survival when it has not been compared to any other tests. Secondly, the author states that using Molecular Twin will create tests that can be used even in locations that lack access to advanced resources and technology, but there is no evidence provided to support this claim. Thirdly, the article uses sensationalist language such as 'viability' and 'pairing patients with the most effective therapies', which could mislead readers into believing that Molecular Twin has already been proven to be an effective tool for pancreatic cancer treatment.
                • The article uses sensationalist language such as 'viability' and 'pairing patients with the most effective therapies', which could mislead readers into believing that Molecular Twin has already been proven to be an effective tool for pancreatic cancer treatment.
                • The author claims that Molecular Twin outperforms the standard test for predicting pancreatic cancer survival, but no comparison has been made with any other tests.
                • The author states that using Molecular Twin will create tests that can be used even in locations that lack access to advanced resources and technology, but there is no evidence provided to support this claim.
              • Fallacies (85%)
                The article contains several fallacies. The first is an appeal to authority when it states that the Molecular Twin Precision Oncology Platform developed at Cedars-Sinai can be used to study any tumor type, including pancreatic cancer. This statement implies that because a tool was developed by experts in one field, it must be effective for all other fields without further testing or validation. The second fallacy is an inflammatory rhetoric when the article states that Molecular Twin outperformed the only Food and Drug Administration-approved pancreatic cancer test, a blood test called CA 19-9. This statement implies that Molecular Twin is significantly better than any other available option without providing evidence of its superiority in all cases. The third fallacy is an appeal to emotion when the article states that using Molecular Twin technology could one day guide and improve treatment for all cancer patients, expanding the availability of precision medicine. This statement implies a personal or emotional connection to the topic without providing any evidence.
                • The Molecular Twin Precision Oncology Platform developed at Cedars-Sinai can be used to study any tumor type, including pancreatic cancer.
              • Bias (85%)
                The article contains a statement that the Molecular Twin Precision Oncology Platform developed at Cedars-Sinai can be used to study any tumor type, including pancreatic cancer. This is an example of monetization bias as it implies that this technology will generate revenue for Cedars-Sinai and its stakeholders.
                • The Molecular Twin Precision Oncology Platform developed at Cedars-Sinai can be used to study any tumor type, including pancreatic cancer.
                • Site Conflicts Of Interest (50%)
                  The article discusses a new tool developed by Cedars-Sinai Cancer and its researchers to improve pancreatic cancer patient care. The authors of the study are Dan Theodorescu, MD, Ph.D., Arsen Osipov, MD and Jennifer Van Eyk, Ph.D.
                  • The article mentions that Cedars-Sinai Cancer is a partner in developing this new tool.
                  • Author Conflicts Of Interest (50%)
                    The author has a conflict of interest on the topic of precision medicine as they are affiliated with Cedars-Sinai Cancer and have developed a molecular twin precision oncology platform. The article does not disclose this conflict.