Optimal EUS-guided FNA cytology preparation when rapid on-site evaluation is not available

Application of artificial intelligence to digital-rapid on-site cytopathology evaluation during endoscopic ultrasound-guided fine needle aspiration: A proof-of-concept study

BACKGROUND

During endoscopic ultrasound-guided fine needle aspiration (EUS-FNA), cytopathology with rapid on-site evaluation (ROSE) can improve diagnostic yield and accuracy. However, ROSE is unavailable in most Asian and European institutions because of the shortage of cytopathologists. Therefore, developing computer-assisted diagnostic tools to replace manual ROSE is crucial. Herein, we reported the validation of an artificial intelligence (AI)-based model (ROSE-AI model) to substitute manual ROSE during EUS-FNA.

METHODS

A total of 467 digitized images from Diff-Quik (D&F)-stained EUS-FNA slides were divided into training (3642 tiles from 367 images) and internal validation (916 tiles from 100 images) datasets. The ROSE-AI model was trained and validated using training and internal validation datasets, respectively. The specificity was emphasized while developing the model. Then, we evaluated the AI model on a 693-image external dataset. We assessed the performance of the AI model to detect cancer cells (CCs) regarding the accuracy, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).

RESULTS

The ROSE-AI model achieved an accuracy of 83.4% in the internal validation dataset and 88.7% in the external test dataset. The sensitivity and PPV were 79.1% and 71.7% in internal validation dataset and 78.0% and 60.7% in external test dataset, respectively.

CONCLUSION

We provided a proof of concept that AI can be used to replace manual ROSE during EUS-FNA. The ROSE-AI model can address the shortage of cytopathologists and make ROSE available in more institutes.

Impact of rapid on-site evaluation on diagnostic accuracy of EUS-guided fine-needle aspiration of solid pancreatic lesions: experience from a single center

BACKGROUND

The use of ROSE in EUS-FNA pancreatic lesions is still controversial in many centers. In this study, we aimed to demonstrate the contribution of ROSE to the diagnostic accuracy, as well as its assistance to the pathologist/cytopathologist.

METHODS

162 EUS-FNA biopsies were included in the study. EUS-FNA cytology results were reported according to the six-tiered system of Papanicolaou Cytopathology Society and compared to their final diagnosis with histopathology and/or clinical follow-ups regarding malignancy. The diagnostic yield, the difference in diagnostic accuracy, and the contribution of ROSE to providing the pathologist with adequate tissue uptake (number of slides and cell blocks) for further examination were compared in the ROSE and non-ROSE patient groups.

RESULTS

In the non-ROSE group, the diagnostic accuracy according to the final diagnoses was 96% and the sensitivity was 94.44%, specificity 100%, PPV 100%, NPV 87.50%; while diagnostic accuracy was 97.09%, sensitivity 97.47%, specificity 95.83%, PPV 98.77%, NPV 92% in patients with ROSE. There was no significant difference in diagnostic accuracy between those with and without ROSE (p: 0.078). In diagnostic cases, the number of passes, slides and cell blocks were significantly higher in patients with ROSE than those without ROSE (p: 0.003, p: 0.007, p: 0.012, respectively). ROSE was independently associated with diagnostic yield when evaluated by number of passes, slides, cell blocks in regression analysis (p: 0.001, OR: 5.07, confidence interval: 1.89-13.5).

CONCLUSION

ROSE may increase the acquisition of sufficient tissue, but it does not have an advantage in diagnostic accuracy. ROSE may raise the number of slides, which may assist the pathologist for the diagnosis. If the lesion is solid and/or contains a solid component, diagnostic yield is higher in patients where ROSE is available. Therefore, ROSE still maintains its applicability in terms of increasing the diagnostic efficiency and making the final diagnosis.