1 00:00:00,210 --> 00:00:01,080 So now let's talk about 2 00:00:01,080 --> 00:00:03,030 Amazon Comprehend Medical. 3 00:00:03,030 --> 00:00:06,120 So Comprehend Medical detects and returns 4 00:00:06,120 --> 00:00:09,270 useful information in unstructured clinical text. 5 00:00:09,270 --> 00:00:11,640 For example, if your doctor is taking notes, 6 00:00:11,640 --> 00:00:13,110 or if you have discharge summaries, 7 00:00:13,110 --> 00:00:17,340 or test results, or case notes, then it will use NLP. 8 00:00:17,340 --> 00:00:20,910 So it will use natural language processing to detect 9 00:00:20,910 --> 00:00:23,320 all the protected health information 10 00:00:24,347 --> 00:00:28,417 within the document itself as well with a DetectPHI API. 11 00:00:29,460 --> 00:00:31,290 So from an architecture perspective 12 00:00:31,290 --> 00:00:33,900 you would store whatever documents you have in Amazon S3, 13 00:00:33,900 --> 00:00:37,050 and then you would invoke the Comprehend Medical API. 14 00:00:37,050 --> 00:00:39,090 Or you could have Kinesis Data Firehose, 15 00:00:39,090 --> 00:00:40,890 and then analyze that in real time. 16 00:00:40,890 --> 00:00:42,510 Or you can use Amazon Transcribe 17 00:00:42,510 --> 00:00:45,810 to first transcribe the voice into text. 18 00:00:45,810 --> 00:00:47,340 Okay, as we understand from before. 19 00:00:47,340 --> 00:00:49,200 And then once it's in text form 20 00:00:49,200 --> 00:00:52,620 we can pass it to the Amazon Comprehend Medical service. 21 00:00:52,620 --> 00:00:55,770 So let's have a look into the consult to see how that works. 22 00:00:55,770 --> 00:00:58,787 So I am in the Amazon Comprehend Medical service, 23 00:00:58,787 --> 00:01:01,740 and let's launch a real time analysis. 24 00:01:01,740 --> 00:01:04,470 So here it's Input text. 25 00:01:04,470 --> 00:01:07,650 And so imagine that this is doctor's note and he wrote 26 00:01:07,650 --> 00:01:09,750 or she wrote all these things, right? 27 00:01:09,750 --> 00:01:11,580 So the doctor, okay, wrote all this. 28 00:01:11,580 --> 00:01:13,590 Now let's analyze it. 29 00:01:13,590 --> 00:01:16,140 And from this, we can have entities. 30 00:01:16,140 --> 00:01:19,680 And so you can find an age, and an high school teacher, 31 00:01:19,680 --> 00:01:22,530 and then you can find some overlap between some things 32 00:01:22,530 --> 00:01:25,048 such as there was a procedure name and the time 33 00:01:25,048 --> 00:01:26,607 to the procedure name and the date and so on. 34 00:01:26,607 --> 00:01:30,060 And as you can see all this data become quite organized. 35 00:01:30,060 --> 00:01:33,570 And so you have like some generic name of a molecule 36 00:01:33,570 --> 00:01:35,790 and its strength and dosage and route modes 37 00:01:35,790 --> 00:01:37,170 and frequency and so on. 38 00:01:37,170 --> 00:01:39,990 And this allows you to really start structuring 39 00:01:39,990 --> 00:01:42,960 all your health data from unstructured text 40 00:01:42,960 --> 00:01:44,640 using machine learning. 41 00:01:44,640 --> 00:01:46,413 So you can explore this, okay, 42 00:01:47,250 --> 00:01:48,360 but I'm not gonna go into the details of this. 43 00:01:48,360 --> 00:01:51,480 This is a medical field that I'm not very confident with, 44 00:01:51,480 --> 00:01:53,280 but at least you know that thanks to 45 00:01:53,280 --> 00:01:56,940 Amazon Comprehend Medical you can take information 46 00:01:56,940 --> 00:02:00,930 in text form and then get insights out of it. 47 00:02:00,930 --> 00:02:02,040 So I hope you liked it. 48 00:02:02,040 --> 00:02:03,990 And I will see you in the next lecture.