Analyzing medical images, like X-Rays, MRI or CT Scans, as well as ultrasound images or videos is hard. That is why it takes a lot of training to become a proficient radiologist, with the ability to spot changes in normal human tissues and make correct diagnosis. But, as with many other fields, artificial intelligence has also arrived in the medical sector. It is able to do a lot of things, as we are seeing in this trilogy of articles, but one of the most interesting is the possibility of doing diagnostics in medical imaging.

A properly trained AI model, that has been shown hundreds or thousands of digital medical images and videos, like those from modern X-Ray machines, or MRI, CT Scan or PET scans, can be able to spot anomalies in the tissues. Sometimes even those small anomalies that a human would not notice or see.

Actually, almost with everything. The fact that an AI algorithm has to be trained first to be effective, is the key point here. Just like a doctor has be trained in different areas, the same can be done to an AI. In fact, an AI can be trained far more specifically than any human could, because all the AI models rely on two things: the model itself, and the data it has been trained on/is still being trained with each new patient. This allows for a kind of specialization that is difficult to achieve as a person. We can’t deal with immense amounts of data as easily as a machine can. But artificial intelligence has much more application than only image diagnosis. Here are some of them:

In today’s healthcare, medical imaging has an ever more important role when diagnosing a patient’s condition. The systems have gotten so advanced and have such good resolution, that diagnostics are made with more certainty than ever. But, in contrast to the imaging “of old” (where “old” is just a couple of years back), today most of the medical imaging devices generate digital files, instead of having to develop film or the like.

On top of that, most of them do not generate only static images, but also video files. For these images there is the DICOM format (Digital Imaging and Communications in Medicine) which is a standard protocol for the management and transmission of medical images and related data. In fact, DICOM also supports several video formats as well as multiframe images, like the ones in CT scans. However, this is where the problems arise. Most vendors have a particular interpretation of their image formats, in order to lock the hospital to that particular brand, avoiding too much compatibility with the rest of the competing vendors.

The resulting scenario is one in which the medical professional cannot always use a PACS (Picture Archiving and Communications System) to access the images from his or her office, but has to go physically to the machine in question that did the images, and watch them there. The fact that DICOM uses no compression in the images to avoid artifacts (that could be misleading the doctor or radiologist) is one of the reasons. When recording video this also intensifies, as these files are so huge, that they usually cannot be transmitted over the hospital network in an efficient way. Hence the doctor has to go to the machine and watch the exploration video in situ. And this is far from ideal.

But there is a solution to this: adding a secondary video PACS can solve the problem of viewing medical video directly from the doctor’s office, without having to change the hardware or software in the hospital. It just needs to add the secondary video PACS, which integrates with the current PACS and, as we will see in a bit, also with the HIS. Videomed PACS is such a secondary video PACS.

What the secondary PACS does is gathering all the video from the machines by direct connection to the network and storing it centrally. Then all the formats are unified to a single one, usually a high quality MP4. This proxy file then can be viewed with ease and needing a low network bandwidth, from anywhere in the hospital, just be connecting to the PACS. If the doctor or surgeon needs to see the original file, this is also possible over a high speed network connection but, instead of involving the PACS, which may be used by all of the hospital constantly, the video is provided by the secondary PACS instead, taking load from the main PACS.

On top of this, the system anonymizes all the patient data for privacy and compliance reasons, so it never leaves the HIS or main PACS. And it is possible to specify retention periods for the videos (due to the space they take up in the system), so they are converted into lower resolution copies after a time. This makes for efficient storage, and keeps the system tidy, without losing any information.

And on top of this, platforms like Videomed PACS make use of artificial intelligence analyzers in order to catalogue the videos automatically, helping to speed the process up further, putting everything into place without human intervention. And, as we will see in part II of this series, it can integrate diagnostic AI analyzers as well, providing even more help for radiologists or surgeons, who can rely on a “second opinion” of the AI system.