MEDICAL INFORMATICS IV
SURGICAL INSIGHT-GUIDED DEEP LEARNING FOR COLORECTAL LESION MANAGEMENT
Yayıncı:
İstanbul Üniversitesi Yayınları
Colonoscopy stands as a pivotal diagnostic tool in identifying gastrointestinal diseases, including potentially malignant tumors. The procedure, however, faces challenges in the precise identification of lesions during visual inspections. The recent strides in AI and machine learning technologies have opened avenues for enhanced medical imaging analysis, including in the field of colonoscopy. In this study, we introduce an AI-driven convolutional neural network (CNN) model designed to optimize diagnosis and intervention during colonoscopy procedures. The model transcends the conventional focus on polyp detection, encompassing all potentially malignant lesions. Leveraging surgical histopathology results as the gold standard, we trained our model to achieve heightened accuracy in lesion detection and characterization. Our model exhibited remarkable precision and recall rates of 0.79604 and 0.78086, respectively, alongside a mean Average Precision (mAP50) of 0.83243. Furthermore, it demonstrated a sensitivity of 70.73% and a specificity of 92.00% during real-time external validation, showcasing its robustness in identifying lesions accurately. The model stands as a promising tool in the clinical setting, offering real-time applicability and compatibility with existing medical equipment. By facilitating more accurate diagnoses, it harbors the potential to enhance decision-making in gastrointestinal surgery and improve patient outcomes. Its impressive diagnostic metrics indicate that it could be a substantial step forward in the early detection and treatment of gastrointestinal diseases.
Yayınevi Bilgileri
Alıntı Yap
Benzer Yayınlar
Kitap Bölümleri
Bu Sayfayı Yazdırın
Endnote
Mendeley