Alibaba DAMO Academy Unveils DAMO EAGLE AI for Non-Invasive Esophageal Cancer Detection
Alibaba has unveiled its latest cancer screening AI tool, DAMO EAGLE. Developed jointly by Alibaba DAMO Academy, Sichuan Cancer Hospital, and Sun Yat-sen University Cancer Center, this model detects esophageal cancer—including early-stage and precancerous lesions—using standard CT scans, eliminating the need for endoscopy or contrast agents. Validated across 80,000 cases in three countries, the research was published on September 22 in Nature Medicine, a leading international journal.

Achieving 90% Sensitivity and 99.2% Specificity in a Challenging Field
China accounts for nearly half of global esophageal cancer cases, with most diagnoses occurring at late stages, often missing the window for curative treatment. Early detection through minimally invasive endoscopic surgery offers a five-year survival rate above 95%. While upper gastrointestinal endoscopy is the current standard, its invasive nature and complexity hinder widespread adoption. Plain chest CT scans, however, are cost-effective and widely accepted, naturally covering the esophagus. If AI can identify esophageal risks during lung cancer screening, it would significantly boost efficiency. The difficulty lies in the esophagus’s structure as a collapsible hollow organ, with early lesions hidden in mucosal layers and obscured by heart and blood vessel movements, making identification on plain CT scans exceptionally hard for radiologists.
Yao Jiawen, an algorithm expert at DAMO Academy, explained that the team innovatively correlated endoscopic reports with precise lesion locations on enhanced CT images to train the AI on plain CT data. This approach enables the detection of subtle early lesions invisible to the human eye. According to the study, DAMO EAGLE achieves 90% sensitivity for esophageal cancer in opportunistic screening scenarios and 52.5% sensitivity for weak signals in precancerous lesions. In real-world applications, it maintains a specificity of 99.2%, performing consistently on low-dose CT scans. This allows seamless integration into existing lung cancer screening workflows without adding burden to patients.
Advancing the 'Plain CT + AI' Multi-Cancer Screening Strategy
Wang Qifeng, Chief Physician of Radiotherapy at Sichuan Cancer Hospital, noted that dietary habits like consuming hot and salty foods contribute to high esophageal cancer rates in western China, particularly Sichuan. Limited medical resources and low patient compliance create significant gaps in endoscopic screening. "AI does not replace endoscopy; it identifies high-risk individuals who then undergo targeted endoscopy, improving accuracy and detection rates," he stated. Crucially, this approach shifts prevention efforts earlier, offering more opportunities for early intervention.
With the recent release of AI models for pancreatic, gastric, colorectal, esophageal cancers, and aortic dissection, DAMO Academy has successfully established the "plain CT + AI" multi-screening technical route. Having published five papers in Nature Medicine, this technology holds the potential to screen for China’s top seven deadliest cancers using a single plain CT scan.
Related article
Tencent WorkBuddy Launches on HarmonyOS PC App Market
On July 27, Zhang Jun, Tencent’s public relations director, announced the official launch of the company’s AI office assistant, WorkBuddy, on the HarmonyOS PC app market. This marks the debut of the first desktop AI agent within the HarmonyOS ecosyst
ByteDance’s 50,000 AI Chip Buy Signals New Era for Domestic GPU Industry
As the user base of AI large models continues to expand, the self-reliance and controllability of computing infrastructure has become a strategic focus for major technology companies. On June 15, it was reported that ByteDance is currently in deep di
Denmark Fights AI Cheating With Oral Defenses and Live Screen Monitoring
Danish high schools have witnessed a sharp rise in AI-assisted cheating during this summer’s exam period. As generative AI becomes integrated into classrooms, students increasingly rely on it for homework and exams, creating significant challenges fo
Related Special Topic Recommendations
Comments (0)
0/500
Alibaba has unveiled its latest cancer screening AI tool, DAMO EAGLE. Developed jointly by Alibaba DAMO Academy, Sichuan Cancer Hospital, and Sun Yat-sen University Cancer Center, this model detects esophageal cancer—including early-stage and precancerous lesions—using standard CT scans, eliminating the need for endoscopy or contrast agents. Validated across 80,000 cases in three countries, the research was published on September 22 in Nature Medicine, a leading international journal.

Achieving 90% Sensitivity and 99.2% Specificity in a Challenging Field
China accounts for nearly half of global esophageal cancer cases, with most diagnoses occurring at late stages, often missing the window for curative treatment. Early detection through minimally invasive endoscopic surgery offers a five-year survival rate above 95%. While upper gastrointestinal endoscopy is the current standard, its invasive nature and complexity hinder widespread adoption. Plain chest CT scans, however, are cost-effective and widely accepted, naturally covering the esophagus. If AI can identify esophageal risks during lung cancer screening, it would significantly boost efficiency. The difficulty lies in the esophagus’s structure as a collapsible hollow organ, with early lesions hidden in mucosal layers and obscured by heart and blood vessel movements, making identification on plain CT scans exceptionally hard for radiologists.
Yao Jiawen, an algorithm expert at DAMO Academy, explained that the team innovatively correlated endoscopic reports with precise lesion locations on enhanced CT images to train the AI on plain CT data. This approach enables the detection of subtle early lesions invisible to the human eye. According to the study, DAMO EAGLE achieves 90% sensitivity for esophageal cancer in opportunistic screening scenarios and 52.5% sensitivity for weak signals in precancerous lesions. In real-world applications, it maintains a specificity of 99.2%, performing consistently on low-dose CT scans. This allows seamless integration into existing lung cancer screening workflows without adding burden to patients.
Advancing the 'Plain CT + AI' Multi-Cancer Screening Strategy
Wang Qifeng, Chief Physician of Radiotherapy at Sichuan Cancer Hospital, noted that dietary habits like consuming hot and salty foods contribute to high esophageal cancer rates in western China, particularly Sichuan. Limited medical resources and low patient compliance create significant gaps in endoscopic screening. "AI does not replace endoscopy; it identifies high-risk individuals who then undergo targeted endoscopy, improving accuracy and detection rates," he stated. Crucially, this approach shifts prevention efforts earlier, offering more opportunities for early intervention.
With the recent release of AI models for pancreatic, gastric, colorectal, esophageal cancers, and aortic dissection, DAMO Academy has successfully established the "plain CT + AI" multi-screening technical route. Having published five papers in Nature Medicine, this technology holds the potential to screen for China’s top seven deadliest cancers using a single plain CT scan.
Tencent WorkBuddy Launches on HarmonyOS PC App Market
On July 27, Zhang Jun, Tencent’s public relations director, announced the official launch of the company’s AI office assistant, WorkBuddy, on the HarmonyOS PC app market. This marks the debut of the first desktop AI agent within the HarmonyOS ecosyst
ByteDance’s 50,000 AI Chip Buy Signals New Era for Domestic GPU Industry
As the user base of AI large models continues to expand, the self-reliance and controllability of computing infrastructure has become a strategic focus for major technology companies. On June 15, it was reported that ByteDance is currently in deep di
Denmark Fights AI Cheating With Oral Defenses and Live Screen Monitoring
Danish high schools have witnessed a sharp rise in AI-assisted cheating during this summer’s exam period. As generative AI becomes integrated into classrooms, students increasingly rely on it for homework and exams, creating significant challenges fo





Home






