Purpose: The molecular classification of breast cancer is crucial for effective treatment. The emergence of digital pathology has ushered in a new era in which weakly supervised learning leveraging whole-slide images has gained prominence in developing deep learning models because this approach alleviates the need for extensive manual annotation. Weakly supervised learning was employed to classify the molecular subtypes of breast cancer. Materials and Methods: Our approach capitalizes on two whole-slide image datasets: one consisting of breast cancer cases from the Korea University Guro Hospital (KG) and the other originating from The Cancer Genomic Atlas dataset (TCGA). Furthermore, we visualized the inferred results using an attention-based heat map and reviewed the histomorphological features of the most attentive patches. Results: The KG+TCGA-trained model achieved an area under the receiver operating characteristics value of 0.749. An inherent challenge lies in the imbalance among subtypes. Additionally, discrepancies between the two datasets resulted in different molecular subtype proportions. To mitigate this imbalance, we merged the two datasets, and the resulting model exhibited improved performance. The attentive patches correlated well with widely recognized histomorphologic features. The triple-negative subtype has a high incidence of high-grade nuclei, tumor necrosis, and intratumoral tumor-infiltrating lymphocytes. The luminal A subtype showed a high incidence of collagen fibers. Conclusion The artificial intelligence (AI) model based on weakly supervised learning showed promising performance. A review of the most attentive patches provided insights into the predictions of the AI model. AI models can become invaluable screening tools that reduce costs and workloads in practice.

Purpose: The molecular classification of breast cancer is crucial for effective treatment. The emergence of digital pathology has ushered in a new era in which weakly supervised learning leveraging whole-slide images has gained prominence in developing deep learning models because this approach alleviates the need for extensive manual annotation. Weakly supervised learning was employed to classify the molecular subtypes of breast cancer. Materials and Methods: Our approach capitalizes on two whole-slide image datasets: one consisting of breast cancer cases from the Korea University Guro Hospital (KG) and the other originating from The Cancer Genomic Atlas dataset (TCGA). Furthermore, we visualized the inferred results using an attention-based heat map and reviewed the histomorphological features of the most attentive patches. Results: The KG+TCGA-trained model achieved an area under the receiver operating characteristics value of 0.749. An inherent challenge lies in the imbalance among subtypes. Additionally, discrepancies between the two datasets resulted in different molecular subtype proportions. To mitigate this imbalance, we merged the two datasets, and the resulting model exhibited improved performance. The attentive patches correlated well with widely recognized histomorphologic features. The triple-negative subtype has a high incidence of high-grade nuclei, tumor necrosis, and intratumoral tumor-infiltrating lymphocytes. The luminal A subtype showed a high incidence of collagen fibers. Conclusion The artificial intelligence (AI) model based on weakly supervised learning showed promising performance. A review of the most attentive patches provided insights into the predictions of the AI model. AI models can become invaluable screening tools that reduce costs and workloads in practice.

Purpose: The molecular classification of breast cancer is crucial for effective treatment. The emergence of digital pathology has ushered in a new era in which weakly supervised learning leveraging whole-slide images has gained prominence in developing deep learning models because this approach alleviates the need for extensive manual annotation. Weakly supervised learning was employed to classify the molecular subtypes of breast cancer. Materials and Methods: Our approach capitalizes on two whole-slide image datasets: one consisting of breast cancer cases from the Korea University Guro Hospital (KG) and the other originating from The Cancer Genomic Atlas dataset (TCGA). Furthermore, we visualized the inferred results using an attention-based heat map and reviewed the histomorphological features of the most attentive patches. Results: The KG+TCGA-trained model achieved an area under the receiver operating characteristics value of 0.749. An inherent challenge lies in the imbalance among subtypes. Additionally, discrepancies between the two datasets resulted in different molecular subtype proportions. To mitigate this imbalance, we merged the two datasets, and the resulting model exhibited improved performance. The attentive patches correlated well with widely recognized histomorphologic features. The triple-negative subtype has a high incidence of high-grade nuclei, tumor necrosis, and intratumoral tumor-infiltrating lymphocytes. The luminal A subtype showed a high incidence of collagen fibers. Conclusion The artificial intelligence (AI) model based on weakly supervised learning showed promising performance. A review of the most attentive patches provided insights into the predictions of the AI model. AI models can become invaluable screening tools that reduce costs and workloads in practice.

Background/Aims: Accurately diagnosing diffuse gastric wall thickening is challenging. Hypertrophic gastritis (HG), while benign, mimics the morphology of Borrmann type 4 advanced gastric cancer (AGC B-4). We compared the features of endoscopy and endoscopic ultrasonography (EUS) between them. Methods: We retrospectively reviewed patients who underwent EUS for gastric wall thickening between 2000 and 2021, selecting HG and pathologically confirmed advanced gastric cancer cases. Ulceration and antral wall thickening were determined via endoscopy, while EUS assessed the 5-layered gastric wall structure, measuring the proper muscle (PM) layer and total wall thickness. Results: Male dominance was observed in AGC B-4, and the hemoglobin and albumin levels were significantly lower. The rate of antral wall thickening and presence of ulceration were significantly higher in AGC B-4 cases. Destruction of the PM layers was observed only in AGC B-4 cases, and the PM was significantly thicker in AGC B-4 cases. Forceps biopsy had an excellent success rate in ulcer-present AGC B-4 cases, but only a 42.6% success rate was observed for cases without ulcers, necessitating additional diagnostic modalities. A PM thickness of 2.39 mm distinguished between AGC B-4 and HG effectively. The multivariable analysis showed that a thickened PM layer and the presence of ulceration were significant risk factors for the diagnosis of AGC B-4. Conclusions Endoscopic findings of a thickened gastric wall, including antral involvement, and presence of ulcer were significant risk factors for the diagnosis of AGC B-4. EUS findings of destroyed wall layers and a thickened PM of >2.39 mm were the key points of differentiation between HG and AGC B-4.