论文标题

联合学习可以实现罕见癌症边界检测的大数据

Federated Learning Enables Big Data for Rare Cancer Boundary Detection

论文作者

Pati, Sarthak, Baid, Ujjwal, Edwards, Brandon, Sheller, Micah, Wang, Shih-Han, Reina, G Anthony, Foley, Patrick, Gruzdev, Alexey, Karkada, Deepthi, Davatzikos, Christos, Sako, Chiharu, Ghodasara, Satyam, Bilello, Michel, Mohan, Suyash, Vollmuth, Philipp, Brugnara, Gianluca, Preetha, Chandrakanth J, Sahm, Felix, Maier-Hein, Klaus, Zenk, Maximilian, Bendszus, Martin, Wick, Wolfgang, Calabrese, Evan, Rudie, Jeffrey, Villanueva-Meyer, Javier, Cha, Soonmee, Ingalhalikar, Madhura, Jadhav, Manali, Pandey, Umang, Saini, Jitender, Garrett, John, Larson, Matthew, Jeraj, Robert, Currie, Stuart, Frood, Russell, Fatania, Kavi, Huang, Raymond Y, Chang, Ken, Balana, Carmen, Capellades, Jaume, Puig, Josep, Trenkler, Johannes, Pichler, Josef, Necker, Georg, Haunschmidt, Andreas, Meckel, Stephan, Shukla, Gaurav, Liem, Spencer, Alexander, Gregory S, Lombardo, Joseph, Palmer, Joshua D, Flanders, Adam E, Dicker, Adam P, Sair, Haris I, Jones, Craig K, Venkataraman, Archana, Jiang, Meirui, So, Tiffany Y, Chen, Cheng, Heng, Pheng Ann, Dou, Qi, Kozubek, Michal, Lux, Filip, Michálek, Jan, Matula, Petr, Keřkovský, Miloš, Kopřivová, Tereza, Dostál, Marek, Vybíhal, Václav, Vogelbaum, Michael A, Mitchell, J Ross, Farinhas, Joaquim, Maldjian, Joseph A, Yogananda, Chandan Ganesh Bangalore, Pinho, Marco C, Reddy, Divya, Holcomb, James, Wagner, Benjamin C, Ellingson, Benjamin M, Cloughesy, Timothy F, Raymond, Catalina, Oughourlian, Talia, Hagiwara, Akifumi, Wang, Chencai, To, Minh-Son, Bhardwaj, Sargam, Chong, Chee, Agzarian, Marc, Falcão, Alexandre Xavier, Martins, Samuel B, Teixeira, Bernardo C A, Sprenger, Flávia, Menotti, David, Lucio, Diego R, LaMontagne, Pamela, Marcus, Daniel, Wiestler, Benedikt, Kofler, Florian, Ezhov, Ivan, Metz, Marie, Jain, Rajan, Lee, Matthew, Lui, Yvonne W, McKinley, Richard, Slotboom, Johannes, Radojewski, Piotr, Meier, Raphael, Wiest, Roland, Murcia, Derrick, Fu, Eric, Haas, Rourke, Thompson, John, Ormond, David Ryan, Badve, Chaitra, Sloan, Andrew E, Vadmal, Vachan, Waite, Kristin, Colen, Rivka R, Pei, Linmin, Ak, Murat, Srinivasan, Ashok, Bapuraj, J Rajiv, Rao, Arvind, Wang, Nicholas, Yoshiaki, Ota, Moritani, Toshio, Turk, Sevcan, Lee, Joonsang, Prabhudesai, Snehal, Morón, Fanny, Mandel, Jacob, Kamnitsas, Konstantinos, Glocker, Ben, Dixon, Luke V M, Williams, Matthew, Zampakis, Peter, Panagiotopoulos, Vasileios, Tsiganos, Panagiotis, Alexiou, Sotiris, Haliassos, Ilias, Zacharaki, Evangelia I, Moustakas, Konstantinos, Kalogeropoulou, Christina, Kardamakis, Dimitrios M, Choi, Yoon Seong, Lee, Seung-Koo, Chang, Jong Hee, Ahn, Sung Soo, Luo, Bing, Poisson, Laila, Wen, Ning, Tiwari, Pallavi, Verma, Ruchika, Bareja, Rohan, Yadav, Ipsa, Chen, Jonathan, Kumar, Neeraj, Smits, Marion, van der Voort, Sebastian R, Alafandi, Ahmed, Incekara, Fatih, Wijnenga, Maarten MJ, Kapsas, Georgios, Gahrmann, Renske, Schouten, Joost W, Dubbink, Hendrikus J, Vincent, Arnaud JPE, Bent, Martin J van den, French, Pim J, Klein, Stefan, Yuan, Yading, Sharma, Sonam, Tseng, Tzu-Chi, Adabi, Saba, Niclou, Simone P, Keunen, Olivier, Hau, Ann-Christin, Vallières, Martin, Fortin, David, Lepage, Martin, Landman, Bennett, Ramadass, Karthik, Xu, Kaiwen, Chotai, Silky, Chambless, Lola B, Mistry, Akshitkumar, Thompson, Reid C, Gusev, Yuriy, Bhuvaneshwar, Krithika, Sayah, Anousheh, Bencheqroun, Camelia, Belouali, Anas, Madhavan, Subha, Booth, Thomas C, Chelliah, Alysha, Modat, Marc, Shuaib, Haris, Dragos, Carmen, Abayazeed, Aly, Kolodziej, Kenneth, Hill, Michael, Abbassy, Ahmed, Gamal, Shady, Mekhaimar, Mahmoud, Qayati, Mohamed, Reyes, Mauricio, Park, Ji Eun, Yun, Jihye, Kim, Ho Sung, Mahajan, Abhishek, Muzi, Mark, Benson, Sean, Beets-Tan, Regina G H, Teuwen, Jonas, Herrera-Trujillo, Alejandro, Trujillo, Maria, Escobar, William, Abello, Ana, Bernal, Jose, Gómez, Jhon, Choi, Joseph, Baek, Stephen, Kim, Yusung, Ismael, Heba, Allen, Bryan, Buatti, John M, Kotrotsou, Aikaterini, Li, Hongwei, Weiss, Tobias, Weller, Michael, Bink, Andrea, Pouymayou, Bertrand, Shaykh, Hassan F, Saltz, Joel, Prasanna, Prateek, Shrestha, Sampurna, Mani, Kartik M, Payne, David, Kurc, Tahsin, Pelaez, Enrique, Franco-Maldonado, Heydy, Loayza, Francis, Quevedo, Sebastian, Guevara, Pamela, Torche, Esteban, Mendoza, Cristobal, Vera, Franco, Ríos, Elvis, López, Eduardo, Velastin, Sergio A, Ogbole, Godwin, Oyekunle, Dotun, Odafe-Oyibotha, Olubunmi, Osobu, Babatunde, Shu'aibu, Mustapha, Dorcas, Adeleye, Soneye, Mayowa, Dako, Farouk, Simpson, Amber L, Hamghalam, Mohammad, Peoples, Jacob J, Hu, Ricky, Tran, Anh, Cutler, Danielle, Moraes, Fabio Y, Boss, Michael A, Gimpel, James, Veettil, Deepak Kattil, Schmidt, Kendall, Bialecki, Brian, Marella, Sailaja, Price, Cynthia, Cimino, Lisa, Apgar, Charles, Shah, Prashant, Menze, Bjoern, Barnholtz-Sloan, Jill S, Martin, Jason, Bakas, Spyridon

论文摘要

尽管机器学习(ML)在众多域中表现出了希望,但人们对样本外数据的概括性存在担忧。目前,通过中央共享来自多个站点的大量数据,并且重要的是多样化的数据来解决。但是,由于各种局限性,这种集中化对于规模(甚至是不可行)都是有挑战性的。 Federated ML(FL)仅通过共享数值模型更新,提供了训练准确且可推广的ML模型的替代方法。在这里,我们介绍了最大的佛罗里达州研究日期的发现,涉及来自6大洲的71个医疗机构的数据,以生成一种自动肿瘤边界检测器,以实现胶质母细胞瘤罕见的疾病,利用文献中有史以来最大的此类此类患者数据集(25,256例MRI扫描,来自6,314名患者)。我们证明了比公开训练的模型提高了33%的改善,以描绘出手术靶向肿瘤,而整个肿瘤的范围比肿瘤的范围提高了23%。我们预计我们的研究至:1)在大量和多样化的数据中启用更多有关医疗保健的研究,确保对罕见疾病的有意义的成果和代表性不足的人群的有意义的结果,2)促进通过对我们的最终公共释放的共识模型的性能优化胶质母细胞瘤的进一步定量分析,以及在这种规模的复杂性上的效果,3)需要数据共享。

Although machine learning (ML) has shown promise in numerous domains, there are concerns about generalizability to out-of-sample data. This is currently addressed by centrally sharing ample, and importantly diverse, data from multiple sites. However, such centralization is challenging to scale (or even not feasible) due to various limitations. Federated ML (FL) provides an alternative to train accurate and generalizable ML models, by only sharing numerical model updates. Here we present findings from the largest FL study to-date, involving data from 71 healthcare institutions across 6 continents, to generate an automatic tumor boundary detector for the rare disease of glioblastoma, utilizing the largest dataset of such patients ever used in the literature (25,256 MRI scans from 6,314 patients). We demonstrate a 33% improvement over a publicly trained model to delineate the surgically targetable tumor, and 23% improvement over the tumor's entire extent. We anticipate our study to: 1) enable more studies in healthcare informed by large and diverse data, ensuring meaningful results for rare diseases and underrepresented populations, 2) facilitate further quantitative analyses for glioblastoma via performance optimization of our consensus model for eventual public release, and 3) demonstrate the effectiveness of FL at such scale and task complexity as a paradigm shift for multi-site collaborations, alleviating the need for data sharing.

扫码加入交流群

加入微信交流群

微信交流群二维码

扫码加入学术交流群,获取更多资源