Full-Text (Deep) Screening | HubMeta Tutorial #12 — Transcript

Full transcript

1. 0:15Welcome back. So, after we have the PDFs
2. 0:18for our articles downloaded, now would
3. 0:21be the time to go through
4. 0:24the full text of our articles. According
5. 0:27to Prisma, this is the next step in our
6. 0:30selection process. So, we technically we
7. 0:33have to go through all of these PDFs, so
8. 0:36one by one. If you go inside deep scan,
9. 0:39it shows side by side your PDF.
10. 0:43You can view it in a new screen.
11. 0:46You know, if you're working on multi
12. 0:48screens, that's really handy. Or if you
13. 0:51can expand it to full size. But then
14. 0:54based on this, you make yes or no
15. 0:56decisions
16. 0:58in comparison with your inclusion
17. 1:01exclusion criteria. Very similar to our
18. 1:04title scan, same filters,
19. 1:08a lot of the same
20. 1:10actions,
21. 1:12like interrater reliability, which you
22. 1:14need again if you have
23. 1:16collaborators which are submitting their
24. 1:18votes, you will see it here. Not much
25. 1:20different from our title scan.
26. 1:24The only difference between deep scan
27. 1:26and title scan is that in this one, you
28. 1:29have the full text of the article, not
29. 1:31just the title and abstract of it.
30. 1:34And again, a good way of doing things in
31. 1:37the new platform is to go to
32. 1:41the assist function inside Peg. And this
33. 1:45time, like we did before, you can set
34. 1:48your criteria again or just use the
35. 1:50criteria that you used before,
36. 1:54set up your
37. 1:55you know, your AI bots, and then
38. 1:59the only difference is that in this
39. 2:02step, instead of title scan, you will
40. 2:04use deep scan. And here it says, "Which
41. 2:08of your papers you want to include?" So,
42. 2:11the ones that are in the full text
43. 2:12stage, obviously,
44. 2:15uh, you know, all papers that have a
45. 2:16PDF, like there are different options to
46. 2:18filter things out. Uh, again,
47. 2:22if you click on the start of screening,
48. 2:24it will show you how many it will cost
49. 2:27you to run this, and then you confirm
50. 2:30and start, and
51. 2:32it will do that for you.
52. 2:34Again, it will go through each of the
53. 2:37papers, this time not just the title and
54. 2:40abstract, but it will
55. 2:42grab the whole text of the paper and
56. 2:45send it to AI
57. 2:48with the inclusion exclusion criteria
58. 2:50and ask it to make a decision of
59. 2:53inclusion or exclusion. And after it's
60. 2:56done, I will show you one from a
61. 2:58previous project. We will go to our deep
62. 3:01scan, and again, just like title scan,
63. 3:04you will see the AI bots and their
64. 3:08consensus vote on whether it should be
65. 3:10included or excluded in the project. And
66. 3:14using these filters, again, I can say,
67. 3:17for example, let's filter all the ones
68. 3:19that everybody agreed should be
69. 3:21accepted, and then uh, with just one
70. 3:24click, we can
71. 3:26uh, send them to the next stage. So,
72. 3:28with that, we are done with the second
73. 3:32part of the pipeline, which was pick.
74. 3:35So, what have we done so far? We have
75. 3:38pulled, we have brought all the articles
76. 3:41that we will potentially include our
77. 3:43systematic
78. 3:45that we will potentially include in our
79. 3:47systematic review into HubMed. We have
80. 3:51applied our inclusion exclusion criteria
81. 3:54and narrowed them down to the set that
82. 3:57actually have to be included. We have
83. 4:00gathered the PDF files for all of them
84. 4:03and then looked at them one more time,
85. 4:05looked at the full text and then
86. 4:08confirmed that they are indeed
87. 4:10includable.
88. 4:12What we have after this stage is a set
89. 4:15of includable, extractable, and
90. 4:19analyzable set of articles. Most of
91. 4:21these final set of articles will
92. 4:23potentially be included in our final set
93. 4:26of included studies and have the data
94. 4:30that we need and we can extract from
95. 4:32that. In the next video, we will move on
96. 4:36to the preference stage, which is all
97. 4:38about extracting information.
98. 4:49>> [music]
99. 4:54[music]