Alibaba unveils AI model to detect stomach cancer early

Alibaba Group has introduced an AI model called Grape, designed to detect gastric cancer through CT scan analysis.

Grape, developed with Zhejiang Cancer Hospital, achieved 85.1% sensitivity and 96.8% specificity, outperforming radiologists significantly.

The model uses deep-learning algorithms to analyze 3D CT scans and will be used in screenings across Zhejiang and Anhui.

Grape follows other cancer-detection tools by Alibaba’s Damo Academy, including a pancreatic cancer AI approved for fast-tracking by the US FDA.

The team plans to apply Grape’s technology to detect other gastrointestinal cancers in the future.

🔗 Source: South China Morning Post

🧠 Food for thought

1️⃣ Patient acceptance will determine Grape’s real-world impact despite technical superiority

While Grape’s technical performance is impressive, with 85.1% sensitivity and 96.8% specificity that outperforms radiologists, patient acceptance will ultimately determine its success in gastric cancer screening.

Research shows that user-friendly applications, clear instructions, feedback mechanisms, and patient-centered approaches are critical factors for AI acceptance in healthcare 1.

This is particularly relevant considering that fewer than 30% of Chinese patients currently follow recommendations for endoscopies, as noted by Dr. Cheng Xiangdong of Zhejiang Cancer Hospital.

Concerns about data protection and lack of human oversight remain significant barriers to AI adoption in healthcare settings 1, which Alibaba will need to address during implementation.

Successful deployment will require addressing these acceptance factors as Grape moves from research to large-scale screening programs in Zhejiang and Anhui provinces.

2️⃣ Healthcare provider adoption requires strategic implementation beyond technical performance

For Grape to be successfully integrated into clinical practice, Alibaba must consider the factors that influence healthcare professionals’ acceptance of AI technologies.

Research identifies that fear of autonomy loss and integration difficulties are significant barriers to AI adoption among medical professionals 2, challenges that could affect Grape’s implementation despite its superior performance.