Cover photo

Medical Artificial Intelligence in Japan

I've been diving deeper into AI recently, especially as part of Network School, where I’ve had the chance to meet incredible people from AI, machine learning and entrepreneurs. AI is a subject I didn’t know much about before, and I’m still learning. Whenever I bring it up to my mom, she always tells this story about a movie she saw in the past where, in the end, the robot created by a professor ends up killing him. It never fails to crack me up! But in all seriousness, while AI is advancing rapidly and offering incredible new solutions, it’s also bringing new challenges. Overall, I believe AI has immense potential to be beneficial.

Just some days ago, I came across two fascinating pieces of news about AI in Japan's medical field. As I researched further, I discovered something even more surprising: Japan's AI and medical integration actually began in the mid-2010s, about 15 years ago, during Shinzo Abe's tenure as Prime Minister. Abe prioritized the adoption of AI in healthcare as part of a broader innovation strategy, addressing the challenges posed by Japan's aging population. His administration laid the foundation for AI advancements across multiple sectors, including healthcare.

So what about today's government? From some small research done, looks like they still have it in focus. Japan has been advancing initiatives to integrate AI and Internet of Things (IoT) technologies to enhance healthcare delivery. This includes developing AI-driven tools to assist with diagnostics, automate record-keeping, and improve patient care. The government’s vision includes "AI hospitals," aimed at addressing challenges like the shortage of medical professionals and providing better, more efficient care. For more info, please check here.

In recent years, Japan has also invested in more advanced applications, such as AI-assisted liquid biopsies for cancer treatment and personalized medicine. The goal is to use AI to analyze vast amounts of patient data, improving the precision and speed of treatments. These innovations aim to reduce the burden on healthcare workers while improving the quality of care for patients​. Check it out here.

AI helping to detect pancreatic cancer as early as the first stage:
This is a huge development, as pancreatic cancer is notoriously hard to diagnose early and is often discovered at stage 4, by which point treatment options are limited.

Pancreatic cancer is the fourth leading cause of cancer-related deaths globally, and it's projected to become the second most lethal cancer by 2030. The five-year survival rate is tragically low, so the fact that AI could assist in early detection is incredibly promising. It offers hope in a field where early symptoms are often too vague to prompt timely diagnosis. The research team led by Japanese provided the below facts:

Pancreas Image- Photo Credit

  • Only 1.8% and 2.3% of cases are caught at early stages (IA and IB), as most patients are asymptomatic* at that stage. Stage IA refers to a tumor that is 2 cm or smaller, while Stage IB indicates a tumor larger than 2 cm. The image below provides a clearer visualization of the differences in tumor size between these stages. *Asymptomatic means no symptoms.

    Photo Credit
  • Currently, there are no blood biomarkers that can detect pancreatic cancer in its early stages. Today, the following tests that are commonly used to detect pancreatic cancer: an MRI scan or MRCP (magnetic resonance cholangiopancreatography), an endoscopic ultrasound, or a CT scan. Another common test involves measuring CA 19-9, a tumor marker closely associated with pancreatic cancer. In healthy individuals, CA 19-9 levels typically fall within the normal range of 0 to 37 U/mL (units per milliliter). However, in those with pancreatic cancer, these levels are often significantly elevated, making it a useful indicator in the diagnostic process.

  • According to a nationwide study in Japan, abdominal ultrasound (AUS) screenings detected only 152 cases of pancreatic cancer out of 3,005,393 individuals screened—a detection rate of just 0.005%. This highlights the urgent need for more effective and less invasive diagnostic tools to catch pancreatic cancer at an early stage.

    So far, only a few studies have explored the full range of miRNAs (small RNA molecules) in pancreatic cancer patients, making it difficult to confirm their potential as reliable early detection markers. To explain briefly, microRNAs (miRNAs) don't code for proteins, but they regulate gene expression by either reducing the stability of messenger RNA (mRNA) or slowing down protein production. These miRNAs are quite stable in challenging conditions, like blood or pancreatic juice, because they are often enclosed in small vesicles or bound to proteins, protecting them from being broken down. This stability makes miRNAs promising candidates for liquid biopsies.

    The research aimed to determine if miRNA profiles combined with CA19-9, a tumor marker commonly used for pancreatic and other cancers, could help detect pancreatic cancer at various stages.

    • They analyzed 425 blood samples: 213 from pancreatic cancer patients (collected from 14 centers) and 212 from healthy people (collected from 3 centers).

    • The study used next-generation sequencing (NGS) to look at all miRNAs and used an automated machine learning (AutoML) method to create models based on 100 highly expressed miRNAs and CA19-9.

    • The models were tested in a separate group of patients to see if they worked.

    • Results showed that a combination of 100 miRNAs and CA19-9 could be useful biomarkers for detecting pancreatic cancer, including early-stage cases without symptoms.

      In the study, AI was used through an automated machine learning (AutoML) platform called DataRobot. This AI tool automatically created over 60 diagnostic models by analyzing data from 100 highly expressed miRNAs. The models were evaluated based on their ability to accurately predict pancreatic cancer.

    • By using AI, the researchers were able to find the best-performing models, combining miRNA profiles and CA19-9 levels to distinguish pancreatic cancer patients from healthy individuals with high accuracy.


      References:
      Yamada, T., et al. (2024). Early detection of pancreatic cancer using artificial intelligence. British Journal of Cancer, 130, 1-10. https://doi.org/10.1038/s41416-024-02794-5

    • Suzuki, H., et al. (2020). The role of artificial intelligence in critical care: A narrative review. Journal of Intensive Care, 8(1), 52. https://doi.org/10.1186/s40560-020-00452-5

    • Ministry of Health, Labour and Welfare. (2018). Pancreatic cancer screening and diagnosis report. https://www.mhlw.go.jp/content/10601000/000337597.pdf

    • Yamada, T., et al. (2024). Early detection of pancreatic cancer using artificial intelligence. British Journal of Cancer, 130, 1-10. https://doi.org/10.1038/s41416-024-02794-5

    • National Cancer Institute. (n.d.). Stage I pancreatic cancer. Cancer.gov. https://www.cancer.gov/publications/dictionaries/cancer-terms/def/stage-i-pancreatic-cancer

Loading...
highlight
Collect this post to permanently own it.
Bee Curios logo
Subscribe to Bee Curios and never miss a post.
#medical#ai#pancreas#cancer#japan