As in other domains, artificial intelligence is becoming increasingly important in medicine. In particular,deep learning-based pattern recognition methods can advance the field of pathology byincorporating clinical, radiologic, and genomic data to accurately diagnose diseases and predictpatient prognoses. In this review, we present an overview of artificial intelligence, the brief historyof artificial intelligence in the medical domain, recent advances in artificial intelligence applied topathology, and future prospects of pathology driven by artificial intelligence.

The Gleason grading system, currently the most powerful prognostic predictor of prostate cancer, is based solely on the tumor’s histological architecture and has high inter-observer variability. We propose an automated Gleason scoring system based on deep neural networks for diagnosis of prostate core needle biopsy samples. To verify its efficacy, the system was trained using 1133 cases of prostate core needle biopsy samples and validated on 700 cases. Further, system-based diagnosis results were compared with reference standards derived from three certified pathologists. In addition, the system’s ability to quantify cancer in terms of tumor length was also evaluated via comparison with pathologist-based measurements. The results showed a substantial diagnostic concordance between the system-grade group classification and the reference standard (0.907 quadratic-weighted Cohen’s kappa coefficient). The system tumor length measurements were also notably closer to the reference standard (correlation coefficient, R = 0.97) than the original hospital diagnoses (R = 0.90). We expect this system to assist pathologists to reduce the probability of over- or under-diagnosis by providing pathologist-level second opinions on the Gleason score when diagnosing prostate biopsy, and to support research on prostate cancer treatment and prognosis by providing reproducible diagnosis based on the consistent standards.

The Gleason score contributes significantly in predicting prostate cancer outcomes and selecting the appropriate treatment option, which is affected by well-known inter-observer variations. We present a novel deep learning-based automated Gleason grading system that does not require extensive region-level manual annotations by experts and/or complex algorithms for the automatic generation of region-level annotations. A total of 6664 and 936 prostate needle biopsy single-core slides (689 and 99 cases) from two institutions were used for system discovery and validation, respectively. Pathological diagnoses were converted into grade groups and used as the reference standard. The grade group prediction accuracy of the system was 77.5% (95% confidence interval (CI): 72.3–82.7%), the Cohen’s kappa score (κ) was 0.650 (95% CI: 0.570–0.730), and the quadratic-weighted kappa score (κ_quad) was 0.897 (95% CI: 0.815–0.979). When trained on 621 cases from one institution and validated on 167 cases from the other institution, the system’s accuracy reached 67.4% (95% CI: 63.2–71.6%), κ 0.553 (95% CI: 0.495–0.610), and the κ_quad 0.880 (95% CI: 0.822–0.938). In order to evaluate the impact of the proposed method, performance comparison with several baseline methods was also performed. While limited by case volume and a few more factors, the results of this study can contribute to the potential development of an artificial intelligence system to diagnose other cancers without extensive region-level annotations.

Chang H, et al. J Gynecol Oncol. 2017 Dec;29:e27. https://doi.org/10.3802/jgo.2018.29.e27

We propose a novel transfer learning framework for pathological image analysis, the Response-based Cross-task Knowledge Distillation (RCKD), which improves the performance of the model by pretraining it on a large unlabeled dataset guided by a high-performance teacher model. RCKD first pretrains a student model to predict the nuclei segmentation results of the teacher model for unlabeled pathological images, and then fine-tunes the pretrained model for the downstream tasks, such as organ cancer sub-type classification and cancer region segmentation, using relatively small target datasets. Unlike conventional knowledge distillation, RCKD does not require that the target tasks of the teacher and student models be the same. Moreover, unlike conventional transfer learning, RCKD can transfer knowledge between models with different architectures. In addition, we propose a lightweight architecture, the Convolutional neural network with Spatial Attention by Transformers (CSAT), for processing high-resolution pathological images with limited memory and computation. CSAT exhibited a top-1 accuracy of 78.6% on ImageNet with only 3M parameters and 1.08 G multiply-accumulate (MAC) operations. When pretrained by RCKD, CSAT exhibited average classification and segmentation accuracies of 94.2% and 0.673 mIoU on six pathological image datasets, which is 4% and 0.043 mIoU higher than EfficientNet-B0, and 7.4% and 0.006 mIoU higher than ConvNextV2-Atto pretrained on ImageNet, respectively.

Accurately segmenting cancer lesions is essential for effective personalized treatment and enhanced patient outcomes. We propose a multi-resolution selective segmentation (MurSS) model to accurately segment breast cancer lesions from hematoxylin and eosin (H&E) stained whole-slide images (WSIs). We used The Cancer Genome Atlas breast invasive carcinoma (BRCA) public dataset for training and validation. We used the Korea University Medical Center, Guro Hospital, BRCA dataset for the final test evaluation. MurSS utilizes both low- and high-resolution patches to leverage multi-resolution features using adaptive instance normalization. This enhances segmentation performance while employing a selective segmentation method to automatically reject ambiguous tissue regions, ensuring stable training. MurSS rejects 5% of WSI regions and achieves a pixel-level accuracy of 96.88% (95% confidence interval (CI): 95.97–97.62%) and mean Intersection over Union of 0.7283 (95% CI: 0.6865–0.7640). In our study, MurSS exhibits superior performance over other deep learning models, showcasing its ability to reject ambiguous areas identified by expert annotations while using multi-resolution inputs.