Weakly Supervised Multiple Instance Learning for Benign–Malignant Classification of Parotid Gland Tumors on Whole-Slide Images
摘要
Background: Accurate differentiation between benign and malignant parotid gland tumors (PGTs) is important for diagnosis and treatment planning. Although histopathologic examination of hematoxylin and eosin (H&E)-stained slides remains the diagnostic reference standard, whole-slide image (WSI) assessment is labor-intensive and may be challenging in histologically heterogeneous lesions. Weakly supervised multiple instance learning (MIL) enables WSI-level classification using case-level diagnostic labels without requiring region-level annotation.
Methods: We retrospectively analyzed 101 patients with PGTs, including 58 benign and 43 malignant cases, represented by 189 digitized WSIs. Patient-level stratified five-fold cross-validation was performed to ensure that all WSIs from an individual patient were assigned exclusively to one fold. Following automated tissue detection and quality-controlled patch extraction, 256 × 256-pixel patches were encoded using either ResNet-50 or the pathology-pretrained CTransPath backbone. Three MIL strategies—CLAM-SB, attention-based MIL (ABMIL), and mean pooling—were evaluated. Model performance was assessed using the area under the receiver operating characteristic curve (AUC), accuracy, sensitivity, specificity, F1-score, and area under the precision–recall curve.
Results: CTransPath + CLAM‑SB yielded the highest mean AUC among the evaluated configurations (0.83), with an accuracy of 0.81 sensitivity of 0.82, specificity of 0.80, and F1‑score of 0.80. Under the CLAM‑SB framework, CTransPath showed a higher mean AUC than ResNet‑50 (0.83 vs. 0.81). Using CTransPath features, CLAM‑SB achieved higher mean AUC than ABMIL (0.80) and mean pooling (0.77). However, pairwise differences between models did not reach statistical significance. The optimized CLAM‑SB configuration with a bag loss weight of 0.3 further increased the mean AUC to 0.86.
Conclusions: Weakly supervised MIL enabled benign–malignant classification of PGTs from routine H&E WSIs without region-level annotation. In this retrospective single-center cohort, CTransPath combined with CLAM-SB showed the most favorable overall performance, with a consistent trend toward better discrimination across models. Given the limited sample size, these findings should be regarded as preliminary and warrant validation in larger, multicenter cohorts with external validation to confirm their generalizability and clinical utility.
参考文献
[1] Lee DH, Jung EK, Lee JK, Lim SC: Comparative analysis of benign and malignant parotid gland tumors: A retrospective study of 992 patients. Am J Otolaryngol 2023, 44(2):103690. DOI:10.1016/j.amjoto.2022.103690
[2] Faur AC, Buzaș R, Lăzărescu AE, Ghenciu LA: Current Developments in Diagnosis of Salivary Gland Tumors: From Structure to Artificial Intelligence. Life (Basel) 2024, 14(6). DOI:10.3390/life14060727
[3] Johnson F, Bozzato A, Mansour N, Mantsopoulos K, Psychogios G, Zengel P, Hofauer B: Sonography of Salivary Gland Tumors and Disorders. Ultraschall Med 2025, 46(4):318-344. DOI:10.1055/a-2481-7248
[4] Huang AT, Tang C, Bell D, Yener M, Izquierdo L, Frank SJ, El-Naggar AK, Hanna EY, Weber RS, Kupferman ME: Prognostic factors in adenocarcinoma of the salivary glands. Oral Oncol 2015, 51(6):610-615. DOI:10.1016/j.oraloncology.2015.03.005
[5] Muller S, Tilakaratne WM: Update from the 5th Edition of the World Health Organization Classification of Head and Neck Tumors: Tumours of the Oral Cavity and Mobile Tongue. Head Neck Pathol 2022, 16(1):54-62. DOI:10.1007/s12105-021-01402-9
[6] Locati LD, Ferrarotto R, Licitra L, Benazzo M, Preda L, Farina D, Gatta G, Lombardi D, Nicolai P, Vander Poorten V et al: Current management and future challenges in salivary glands cancer. Front Oncol 2023, 13:1264287. DOI:10.3389/fonc.2023.1264287
[7] Zoccali F, Cialente F, Colizza A, Ralli M, Greco A, de Vincentiis M: Clinico-histopathological review of 255 patients who underwent parotidectomy for pleomorphic adenoma: a 10-year retrospective study-a proposal for an optimal diagnostic and therapeutic algorithm for patients with recurrent pleomorphic adenoma. Eur Arch Otorhinolaryngol 2023, 280(7):3329-3335. DOI:10.1007/s00405-023-07897-y
[8] Cheng PC, Chiang HK: Diagnosis of Salivary Gland Tumors Using Transfer Learning with Fine-Tuning and Gradual Unfreezing. Diagnostics (Basel) 2023, 13(21). DOI:10.3390/diagnostics13213333
[9] Bera K, Schalper KA, Rimm DL, Velcheti V, Madabhushi A: Artificial intelligence in digital pathology - new tools for diagnosis and precision oncology. Nat Rev Clin Oncol 2019, 16(11):703-715. DOI:10.1038/s41571-019-0252-y
[10] Rossi ED, Faquin WC, Baloch Z, Barkan GA, Foschini MP, Pusztaszeri M, Vielh P, Kurtycz DFI: The Milan System for Reporting Salivary Gland Cytopathology: Analysis and suggestions of initial survey. Cancer Cytopathol 2017, 125(10):757-766. DOI:10.1002/cncy.21898
[11] Alsanie I, Shephard A, Azarmehr N, Vargas P, Pring M, Rajpoot NM, Khurram SA: Exploring the feasibility of AI-based analysis of histopathological variability in salivary gland tumours. Sci Rep 2025, 15(1):29171. DOI:10.1038/s41598-025-15249-5
[12] Zbären P, Nuyens M, Loosli H, Stauffer E: Diagnostic accuracy of fine-needle aspiration cytology and frozen section in primary parotid carcinoma. Cancer 2004, 100(9):1876-1883. DOI:10.1002/cncr.20186
[13] Madabhushi A, Lee G: Image analysis and machine learning in digital pathology: Challenges and opportunities. Med Image Anal 2016, 33:170-175. DOI:10.1016/j.media.2016.06.037
[14] Wilson CM, Ospina OE, Townsend MK, Nguyen J, Moran Segura C, Schildkraut JM, Tworoger SS, Peres LC, Fridley BL: Challenges and Opportunities in the Statistical Analysis of Multiplex Immunofluorescence Data. Cancers (Basel) 2021, 13(12). DOI:10.3390/cancers13123031
[15] Classe M, Lerousseau M, Scoazec JY, Deutsch E: Perspectives in pathomics in head and neck cancer. Curr Opin Oncol 2021, 33(3):175-183. DOI:10.1097/cco.0000000000000731
[16] Wang S, Yang DM, Rong R, Zhan X, Fujimoto J, Liu H, Minna J, Wistuba, II, Xie Y, Xiao G: Artificial Intelligence in Lung Cancer Pathology Image Analysis. Cancers (Basel) 2019, 11(11). DOI:10.3390/cancers11111673
[17] Beck AH, Sangoi AR, Leung S, Marinelli RJ, Nielsen TO, van de Vijver MJ, West RB, van de Rijn M, Koller D: Systematic analysis of breast cancer morphology uncovers stromal features associated with survival. Sci Transl Med 2011, 3(108):108ra113. DOI:10.1126/scitranslmed.3002564
[18] Wetstein SC, de Jong VMT, Stathonikos N, Opdam M, Dackus G, Pluim JPW, van Diest PJ, Veta M: Deep learning-based breast cancer grading and survival analysis on whole-slide histopathology images. Sci Rep 2022, 12(1):15102. DOI:10.1038/s41598-022-19112-9
[19] Försch S, Klauschen F, Hufnagl P, Roth W: Artificial Intelligence in Pathology. Dtsch Arztebl Int 2021, 118(12):194-204. DOI:10.3238/arztebl.m2021.0011
[20] Roy M, Wang F, Teodoro G, Bhattarai S, Bhargava M, Rekha TS, Aneja R, Kong J: Deep learning based registration of serial whole-slide histopathology images in different stains. J Pathol Inform 2023, 14:100311. DOI:10.1016/j.jpi.2023.100311
[21] Wang X, Yang S, Zhang J, Wang M, Zhang J, Yang W, Huang J, Han X: Transformer-based unsupervised contrastive learning for histopathological image classification. Med Image Anal 2022, 81:102559. DOI:10.1016/j.media.2022.102559
[22] Xu H, Wang M, Shi D, Qin H, Zhang Y, Liu Z, Madabhushi A, Gao P, Cong F, Lu C: When multiple instance learning meets foundation models: Advancing histological whole slide image analysis. Med Image Anal 2025, 101:103456. DOI:10.1016/j.media.2025.103456
[23] Viswanathan K, Maleki Z, Pantanowitz L, Cantley R, Faquin WC: Cytomorphology and diagnostic pitfalls of sebaceous and nonsebaceous salivary gland lymphadenoma: A multi-institutional study. Diagn Cytopathol 2021, 49(1):83-95. DOI:10.1002/dc.24602
[24] Lu MY, Williamson DFK, Chen TY, Chen RJ, Barbieri M, Mahmood F: Data-efficient and weakly supervised computational pathology on whole-slide images. Nat Biomed Eng 2021, 5(6):555-570. DOI:10.1038/s41551-020-00682-w
[25] Ilse M, Tomczak JM, Welling M: Attention-based deep multiple instance learning. Proceedings of the 35th International Conference on Machine Learning (ICML) 2018, 2127-2136. DOI:10.48550/arXiv.1802.04712
[26] Weber A, Schuster D, Heyer J, Becker C, Burkhardt V, Werner M, Spörlein A, Bronsert P, Schulz T: Deep learning-based histopathological classification and subclassification of benign and malignant salivary gland tumors. Eur Arch Otorhinolaryngol 2026, 283(6):3905-3915. DOI:10.1007/s00405-026-10082-6
[27] Li K, Qian Z, Han Y, Chang EI, Wei B, Lai M, Liao J, Fan Y, Xu Y: Weakly supervised histopathology image segmentation with self-attention. Med Image Anal 2023, 86:102791. DOI:10.1016/j.media.2023.102791
[28] Lu Y, Song W, Arachie C, Huang B: Weakly supervised label learning flows. Neural Netw 2025, 182:106892. DOI:10.1016/j.neunet.2024.106892
[29] Neidlinger P, El Nahhas OSM, Muti HS, Lenz T, Hoffmeister M, Brenner H, van Treeck M, Langer R, Dislich B, Behrens HM et al: Benchmarking foundation models as feature extractors for weakly supervised computational pathology. Nat Biomed Eng 2026, 10(6):1113-1123. DOI:10.1038/s41551-025-01516-3
[30] Wang Y, Gao J, Yin Z, Wen Y, Sun M, Han R: Differentiation of benign and malignant parotid gland tumors based on the fusion of radiomics and deep learning features on ultrasound images. Front Oncol 2024, 14:1384105. DOI:10.3389/fonc.2024.1384105
[31] Wang Y, Dai A, Wen Y, Sun M, Gao J, Yin Z, Han R: Prediction of High-risk Capsule Characteristics for Recurrence of Pleomorphic Adenoma in the Parotid Gland Based on Habitat Imaging and Peritumoral Radiomics: A Two-center Study. Acad Radiol 2025, 32(7):4134-4145. DOI:10.1016/j.acra.2025.01.030
[32] Guo J, Feng J, Huang Y, Li X, Hu Z, Zhou Q, Xu H: Diagnostic performance of MRI-based radiomics models using machine learning approaches for the triple classification of parotid tumors. Heliyon 2024, 10(17):e36601. DOI:10.1016/j.heliyon.2024.e36601
[33] Li MY, Pan Y, Lv Y, Ma H, Sun PL, Gao HW: Digital pathology and artificial intelligence in renal cell carcinoma focusing on feature extraction: a literature review. Front Oncol 2025, 15:1516264. DOI:10.3389/fonc.2025.1516264
[34] Bourgade R, Elhannani M, Loussouarn D, Guédon A, Péteri R, Allix-Béguec C, Berthier M, Toquet C, Espitia O, Roncato C et al: Deep learning for giant cell arteritis diagnosis on temporal artery biopsy. Comput Biol Med 2025, 196(Pt A):110707. DOI:10.1016/j.compbiomed.2025.110707
[35] Hezil N, Bouridane A, Hamoudi R, Al-Maadeed S, Akbari Y, Abdullakutty F: Selective multimodal deep learning for reliable breast cancer subtype classification from histopathology and genomic data. Med Eng Phys 2026, 147(3). DOI:10.1088/1873-4030/ae449b
[36] Li M, Shen ZL, Xian HC, Zheng ZJ, Yu ZW, Liang XH, Gao R, Tang YL, Zhang Z: A Recognition System for Diagnosing Salivary Gland Neoplasms Based on Vision Transformer. Am J Pathol 2025, 195(2):221-231. DOI:10.1016/j.ajpath.2024.09.010
[37] Zhao L, Zhang Z: A improved pooling method for convolutional neural networks. Sci Rep 2024, 14(1):1589. DOI:10.1038/s41598-024-51258-6
[38] Zhou Y, Jin F, Suo G, Yang J: ResViT-GANNet: a deep learning framework for classifying breast cancer histopathology images using multimodal attention and GAN-based augmentation. BMC Med Imaging 2025, 25(1):401. DOI:10.1186/s12880-025-01940-6
[39] Lysukhin D, Varlamov A, Yakimov B, Porubayeva E, Pachuashvili N, Kovaleva E, Vanushko V, Platonova N, Shirshin E, Mokrysheva N et al: Multiple-Instance Learning for thyroid gland disease classification: A hands-on experience. Comput Biol Med 2025, 184:109424. DOI:10.1016/j.compbiomed.2024.109424
[40] Butt NA, Sarwat D, Noya ID, Tutusaus K, Samee NA, Ashraf I: Benchmarking multiple instance learning architectures from patches to pathology for prostate cancer detection and grading using attention-based weak supervision. Sci Rep 2026, 16(1). DOI:10.1038/s41598-026-39196-x
[41] Vabalas A, Gowen E, Poliakoff E, Casson AJ: Machine learning algorithm validation with a limited sample size. PLoS One 2019, 14(11):e0224365. DOI:10.1371/journal.pone.0224365
[42] Tellez D, Litjens G, Bándi P, Bulten W, Bokhorst JM, Ciompi F, van der Laak J: Quantifying the effects of data augmentation and stain color normalization in convolutional neural networks for computational pathology. Med Image Anal 2019, 58:101544. DOI:10.1016/j.media.2019.101544
[43] Franchet C, Schwob R, Bataillon G, Syrykh C, Péricart S, Frenois FX, Penault-Llorca F, Lacroix-Triki M, Arnould L, Lemonnier J et al: Bias reduction using combined stain normalization and augmentation for AI-based classification of histological images. Comput Biol Med 2024, 171:108130. DOI:10.1016/j.compbiomed.2024.108130
指标
DOI：
Submission ID：
下载次数
已发布
如何引用
下载引用
利益冲突声明
作者声明无任何需要披露的利益冲突。
Copyright
本预印本的版权持有者为作者/资助方。
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.