Prediction of BAP1 Expression in Uveal Melanoma Using Densely-Connected Deep Classification Networks

Abstract

Uveal melanoma is the most common primary intraocular malignancy in adults, with nearly half of all patients eventually developing metastases, which are invariably fatal. Manual assessment of the level of expression of the tumor suppressor BRCA1-associated protein 1 (BAP1) in tumor cell nuclei can identify patients with a high risk of developing metastases, but may suffer from poor reproducibility. In this study, we verified whether artificial intelligence could predict manual assessments of BAP1 expression in 47 enucleated eyes with uveal melanoma, collected from one European and one American referral center. Digitally scanned pathology slides were divided into 8176 patches, each with a size of 256 × 256 pixels. These were in turn divided into a training cohort of 6800 patches and a validation cohort of 1376 patches. A densely-connected classification network based on deep learning was then applied to each patch. This achieved a sensitivity of 97.1%, a specificity of 98.1%, an overall diagnostic accuracy of 97.1%, and an F1-score of 97.8% for the prediction of BAP1 expression in individual high resolution patches, and slightly less with lower resolution. The area under the receiver operating characteristic (ROC) curves of the deep learning model achieved an average of 0.99. On a full tumor level, our network classified all 47 tumors identically with an ophthalmic pathologist. We conclude that this deep learning model provides an accurate and reproducible method for the prediction of BAP1 expression in uveal melanoma.

Piloting a Deep Learning Model for Predicting Nuclear BAP1 Immunohistochemical Expression of Uveal Melanoma from Hematoxylin-and-Eosin Sections

“…from a conventional stain, such as the H&E). To the best of our knowledge, Sun et al 30 is the first group that developed a DenseNet 24 model to predict nBAP1 expression on BAP1-IHC stained UM patches. Although this work was groundbreaking in UM, a potential weakness of the authors’ study was the risk of “information leakage” caused by splitting patches from the same slide into both the training and test sets.…”

“…Table 4 shows the corresponding AUCs. It can be seen that when trained with the same subset of samples, our method outperforms that based on the BAP-1 stained slide, 30 and that the performance gain can be even up to 5% (from model (1024)-4).…”

“…For comparison, we also re-implemented the method from Sun et al, 30 which was developed for the classification of patches from BAP1-stained UM slides, and applied it to the our dataset with the same configuration of sample splitting. Table 4 shows the corresponding AUCs.…”

Successful Identification of Nasopharyngeal Carcinoma in Nasopharyngeal Biopsies Using Deep Learning

“…With tremendous success in recognition of natural images, convolutional neural networks have been quickly adopted for medical image analysis [7]. Therefore, analysis of pathology images using artificial intelligence (AI) is becoming more feasible recently [8][9][10][11][12][13][14][15][16][17][18][19]. Of note, a few articles described the success of AI in identifying metastatic breast cancer in lymph nodes [10,[12][13][14]16,19].…”

Uveal Melanoma

“…Diagnostic procedures are addressed by Sun et al who present an innovative artificial intelligence-based method to assess BAP1 expression by immunohistochemistry [19]. Le Guin et al show that the specific GNAQ Q209R mutation is restricted to circumscribed choroidal hemangioma and very rare in uveal melanoma [46].…”