OBJECTIVE: To provide quick diagnostic insights to medical practitioners into echocardiograms by only analyzing the echocardiogram workflows (defined as the sequence of modalities examined). METHODS: We define a dictionary of workflows, called subflows, which are commonly encountered in echocardiography workflows but are mutually exclusive. We represent each workflow as a mixture of dictionary subflows and learn discriminative models for various cardiac diseases using Support Vector Machines. Using these discriminative models, we can predict occurrences of diseases for any, yet unseen, echocardiogram workflow. RESULTS: Working with a corpus of 2300 echocardiograms workflows, we build a dictionary of 172 subflows. Using the associated reports (expert created) we identify the ground-truth diagnoses. We then build discriminative models for 7 different cardiac diseases. Using just the workflow as input, these models can predict diseases on average with over 75% accuracy. CONCLUSIONS: Mining collection of echocardiography workflows, for the first time, we are able to predict diseases without even looking at the image contents.
OBJECTIVE: To provide quick diagnostic insights to medical practitioners into echocardiograms by only analyzing the echocardiogram workflows (defined as the sequence of modalities examined). METHODS: We define a dictionary of workflows, called subflows, which are commonly encountered in echocardiography workflows but are mutually exclusive. We represent each workflow as a mixture of dictionary subflows and learn discriminative models for various cardiac diseases using Support Vector Machines. Using these discriminative models, we can predict occurrences of diseases for any, yet unseen, echocardiogram workflow. RESULTS: Working with a corpus of 2300 echocardiograms workflows, we build a dictionary of 172 subflows. Using the associated reports (expert created) we identify the ground-truth diagnoses. We then build discriminative models for 7 different cardiac diseases. Using just the workflow as input, these models can predict diseases on average with over 75% accuracy. CONCLUSIONS: Mining collection of echocardiography workflows, for the first time, we are able to predict diseases without even looking at the image contents.
Authors: William A Zoghbi; Maurice Enriquez-Sarano; Elyse Foster; Paul A Grayburn; Carol D Kraft; Robert A Levine; Petros Nihoyannopoulos; Catherine M Otto; Miguel A Quinones; Harry Rakowski; William J Stewart; Alan Waggoner; Neil J Weissman Journal: J Am Soc Echocardiogr Date: 2003-07 Impact factor: 5.251
Authors: Melvin D Cheitlin; William F Armstrong; Gerard P Aurigemma; George A Beller; Fredrick Z Bierman; Jack L Davis; Pamela S Douglas; David P Faxon; Linda D Gillam; Thomas R Kimball; William G Kussmaul; Alan S Pearlman; John T Philbrick; Harry Rakowski; Daniel M Thys Journal: J Am Coll Cardiol Date: 2003-09-03 Impact factor: 24.094
Authors: Harald P Kühl; Marcus Schreckenberg; Dierk Rulands; Markus Katoh; Wolfgang Schäfer; Georg Schummers; Arno Bücker; Peter Hanrath; Andreas Franke Journal: J Am Coll Cardiol Date: 2004-06-02 Impact factor: 24.094
Authors: Helmut Baumgartner; Judy Hung; Javier Bermejo; John B Chambers; Arturo Evangelista; Brian P Griffin; Bernard Iung; Catherine M Otto; Patricia A Pellikka; Miguel Quiñones Journal: Eur J Echocardiogr Date: 2008-12-08
Authors: R O Bonow; B Carabello; A C de Leon; L H Edmunds; B J Fedderly; M D Freed; W H Gaasch; C R McKay; R A Nishimura; P T O'Gara; R A O'Rourke; S H Rahimtoola; J L Ritchie; M D Cheitlin; K A Eagle; T J Gardner; A Garson; R J Gibbons; R O Russell; T J Ryan; S C Smith Journal: Circulation Date: 1998-11-03 Impact factor: 29.690