Manus AI with CBCT Achieves 95% Detection and Diagnosis of Jaw Lesions, Outperforming Other Chatbots

Manus AI using cone beam CT (CBCT) data achieved 95% detection and correct diagnosis of jaw lesions in 97 patients. Gemini 2.5 Pro detected 80% but correctly diagnosed 56%. Copilot and Claude were least accurate. Significant differences between chatbots.

• In 97 patients with jaw lesions, Manus chatbot reading raw CBCT data detected and correctly diagnosed 95% of lesions, while the same bot using panoramic images was less accurate.
• Gemini 2.5 Pro using panoramic images detected 80% of lesions but gave a correct diagnosis in only 56%, indicating many false positives or misdiagnoses.
• Integration of raw 3D CBCT data substantially improved performance, suggesting that diagnostic chatbots benefit from advanced imaging data over 2D views.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.