An artificial intelligence (AI)-based lymph node metastasis diagnostic model (LNMDM) has demonstrated accuracy, generalizability, and clinical potential to help avoid pathologist misdiagnosis in the management of bladder cancer . Lancet Oncology.
“LNMDM achieved satisfactory results and is therefore a worthy addition to other AI clinical tools used for automated analysis of pathology slides,” said the study authors.
Across the five validation sets, the diagnostic sensitivity area under the curve (AUC) for LNMDM ranged from 0.978 (95% CI, 0.960-0.996) to 0.998 (95% CI, 0.996-1.000). The diagnostic sensitivity of LNMDM (0.983; 95% CI, 0.941-0.998) was significantly higher than that of junior pathologists (0.906; 95% CI, 0.871-0.934) and senior pathologists (0.947; 95% CI, 0.919-0.968). rice field. Furthermore, the AI model increased the sensitivity of junior pathologists from 0.906 without AI to 0.953 with AI (P. <.0001) and 0.946 to 0.986 for senior pathologists (P. = .00012).
LNMDM continued to yield an AUC of 0.943 (95% CI, 0.918-0.969), a sensitivity of 0.943 (95% CI, 0.891-0.975), and a specificity of 0.819 (95% CI, 0.763-0.866) for breast. I was. Cancer set by multiple cancer examination. For the prostate cancer dataset, the AUC was 0.922 (95% CI, 0.884-0.960), the sensitivity was 0.955 (95% CI, 0.889-0.988), and the specificity was 0.833 (95% CI, 0.785-0.875).
LNMDM detected tumor metastasis in 13 patients in the validation set who were classified as node-negative by the pathologist.
LNMDM maintained 100% sensitivity, allowing pathologists to exclude 80% to 92% of negative slides across different validation sets.
“LNMDM achieved satisfactory results and is therefore a worthy addition to other AI clinical tools used for automated analysis of pathology slides,” said the study authors. “LNMDM was able to detect tumor metastases, especially micrometastases, in lymph nodes. Cross-center, cross-instrument, and multiple cancer datasets, LNMDM maintained satisfactory performance.”
Investigators in a retrospective multicenter diagnostic study evaluated LNMDM in patients with bladder cancer who underwent radical cystectomy and pelvic lymphadenectomy and had full-slide images of lymph node sections available.
Patients treated at two hospitals were included in each hospital’s internal validation set, and patients treated at the other three hospitals were included in the external validation set. Overall, the researchers used these five validation sets to compare performance between LNMDMs and pathologists. The researchers also collected findings for multiple cancer tests from the breast cancer dataset and the prostate cancer dataset.
The primary endpoint of the trial was the observed diagnostic sensitivity in the four prespecified groups. A validation set, a single lymph node test set, a multiple cancer test set, and a subset comparing the performance of a pathologist and her LNMDM. Secondary endpoints included specificity, precision, positive predictive value (PPV), and negative predictive value (NPV).
Investigators included a total of 998 patients with a median age of 64 years (interquartile range, 56–72 years). In the overall population, most patients were male (88%), no node involvement (73%), urothelial carcinoma (93%), and no neoadjuvant chemotherapy (86%).
In the single lymph node test set, LNMDM produced an AUC of 0.995 (95% CI, 0.991-0.998). In this dataset, the precision was 0.964 (95% CI, 0.945-0.978), the sensitivity was 0.960 (95% CI, 0.916-0.985), and the specificity was 0.966 (95% CI, 0.943-0.981). Furthermore, PPV was 0.912 (95% CI, 0.857-0.951) and NPV was 0.985 (95% CI, 0.967-0.994).
The investigators performed a post hoc subgroup analysis to test whether LNMDM could identify positive lymph nodes after neoadjuvant chemotherapy. The lowest AUC for all validation sets was 0.994. Moreover, in the subset of patients who did not receive neoadjuvant chemotherapy, LNMDM produced His AUCs ranging from 0.978 (95% CI, 0.960–0.996) to 1.000 (95% CI, 1.000–1.000).
reference
Wu S, Hong G, Xu A, et al. An Artificial Intelligence-Based Model for Lymph Node Metastasis Detection in Whole-Slide Images of Bladder Cancer: A Retrospective, Multicenter, Diagnostic Study. Lancet on call. Published online March 6, 2023. doi:10.1016/S1470-2045(23)00061-X
