Clinical-Grade Interpretable Artificial Intelligence Tool for Automated Detection of Lymph Node Metastasis in Prostate Cancer
Bibliographic record
Abstract
Lymph node metastasis (LNM) is a critical prognostic factor for prostate cancer and is associated with increased mortality and poor clinical outcomes, necessitating modifications to therapeutic strategies. Manual histopathological evaluation of lymphatic tissue on glass slides is labor intensive, subject to interobserver variability, and prone to error. Deep learning approaches offer substantial promise in enhancing the accuracy and efficiency of LNM detection; however, their efficacy is contingent upon the availability of extensive annotated data sets. In this study, we developed a novel artificial intelligence (AI)-driven model leveraging a limited data set of annotated samples. By identifying and incorporating the most informative instances from unlabeled data into the training process, the model optimizes its learning trajectory through iterative error correction. Validation was performed on independent data sets from 3 academic medical centers, comprising 787 whole slide images and >2000 lymph node tissues. On a combined test set of 165 positive and 622 negative cases, the model achieved an area under the receiver operating characteristic curve of 0.94 (95% CI, 0.92-0.96), with slide-level sensitivity and specificity of 96% (95% CI, 92%-99%) and 92% (95% CI, 89%-94%), respectively. Importantly, the AI algorithm identified micrometastases in 17 cases that were initially missed by pathologists. Although pathologists exhibited a 9% miss rate in micrometastasis detection, the AI model demonstrated a significantly lower miss rate of 3% using the institutional data set, highlighting its potential for clinical deployment. This fully autonomous and reproducible method also significantly reduced slide examination times compared with both general and genitourinary pathologists (P < .001). The proposed method demonstrated interpretability by identifying metastasis regions on whole slide images labeled as positive. Ablation studies substantiate the robustness of the proposed methodology for LNM detection.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.004 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.003 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.000 |
| Research integrity | 0.001 | 0.000 |
| Insufficient payload (model declined to judge) | 0.005 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".