CASIO

AI at medical sites!-Casio camera technology that enables AI diagnosis of skin cancer

On November 16 and 17, the 83rd Japan Dermatological Association Tokyo-Eastern Section Joint Academic Conference was held at Keio Plaza Hotel (Tokyo). Many dermatologists gathered at the venue and gathered information and exchanged opinions. This paper reports on the seminar and exhibition booth related to the DZ-D100 dermo camera released by Casio in May.

  • Booth of Casio's dermo camera “ DZ-D100 ''

Half a year has passed since the DRM-D100 dermo camera, optimized for skin observation and registered as a general medical device (Class I), appeared in May. The DZ-D100 supports both normal shooting and close-up shooting without changing the dedicated lens, and can continuously shoot polarized / unpolarized / UV with one shutter.

As a compact camera that can be used with one hand, it can also handle normal photography such as clinical photography, so it is a product that makes it easy to use without being a dermatologist.

  • DZ-D100 has also begun to collect reactions used in the field

Behind the scenes of AI diagnostic support system development aiming at real clinical practice

The luncheon seminar "Development of AI Diagnosis Support from Japan and Utilization of Dermo Camera" was held on March 16, which included a report on the development status of an AI diagnosis support system for skin cancer and two reports on the effective use of this DZ-D100. It became a pillar content.

Speakers were Hiroshi Koga of Shinshu University and Yasuhide Tokawa of Chiba University. The facilitator was Mr. Masaru Tanaka of Tokyo Women's Medical University East Medical Center.

  • Prof. Hiroshi Koga (left), Prof. Masaru Tanaka (center), Prof. Yasuhide Tokawa (right)

Mr. Koga, a top batter, said, “ Image management in the age of practical use of artificial intelligence (AI) and support for skin cancer diagnosis using AI '', image management methods suitable for AI learning, AI in dermatology He talked about global trends in development. Dr. Koga, who has passed the G test of the Japan Deep Learning Association, is one of the physicians who is working directly on developing a medical image diagnosis support system using AI.

Until now, Dr. Koga has been cooperating in the development of the image management tool "D'z IMAGE Viewer" from the viewpoint of "how to efficiently and functionally manage images taken with the dermo camera". The lecture showed that the development of this viewer was the foundation of AI development for medical image diagnosis support systems.

  • Hiroshi Koga from Shinshu University

“ Speaking of the development of medical AI, the creation of research papers and the development of medical equipment have similar rules, but the goals and the tasks to be cleared are different, so we have to do completely different things. '' Koga points out. Therefore, the development of medical devices is relatively easier to be led by companies than by researchers.

According to the functions that the diagnostic assistant AI wants to realize, the classification of medical devices will change, and the development strategy will greatly change depending on the applicable regulations.

  • Dr. Koga explained the difference between developing research and developing medical devices in the development of medical AI, comparing it to soccer and rugby.

To develop medical diagnostic systems and medical devices that utilize AI, individual dermatologists must be able to properly capture, collect, and manage patient images. The term “appropriate” here means that information such as imaging parameters, linked clinical data, and processing details when processing images, etc., that can be used for medical device development, while protecting patient privacy, is retained in detail. Point.

Dr. Koga describes such a process as "when you take a dermatologist examination, it is necessary to submit the required training history, a list of cases, and training records."

What is useful here is the DZ-D100 dermo camera mentioned from the beginning, and the “D'z IMAGE Viewer” that allows you to easily manage images taken with the DZ-D100. By the way, D'z IMAGE Viewer is Casio'sD'z IMAGE STOREIt is available for free download from.

  • Dr. Koga positions the DZ-D100 as a device for clinical sites where individual dermatologists can properly capture, collect, and manage patient images.

He also mentioned the development status of a support system for skin cancer diagnosis. Koga's affiliation, Shinshu University, together with Casio, challenged the ISBI Challenge 2017 in March 2017, and achieved a prize for winning the top prize in the category. At that time, even in My Navi NewsarticleTaken up. In the contest, AI is trained with 2,000 images for learning, 600 images for test are judged, and the average score is played.

  • Development status of skin cancer diagnosis support system. Won first place in the category of ISBI Challenge 2017 in March 2017

At ISIC 2020, which will be held in 2020, a contest focusing on whether AI can detect changing moles will also be held. Although the comparison function of clinical photos is superior, the engine used for clinical photo comparison of D'z IMAGE Viewer seems to be useful for AI learning as well.

  • D'z IMAGE Viewer may play an important role at ISIC 2020 held in 2020

Lastly, Dr. Koga demonstrated the AI ​​diagnostic support device that Casio is developing, and “ Development of a skin cancer diagnostic solution using imaging data '' under joint research by Casio and Shinshu University was announced by AMED (National Research and Development Corporation Japan) He was selected for the "Development Project for Advanced Medical Devices and Systems" by the Japan Agency for Medical Research and Development, and will receive support from the government for research costs for up to five years.

Dr. Koga says, “This project is a“ medical device, ”so as mentioned above, the goal is to make it usable in clinical practice. With the funding from AMED, we are trying to get beyond the Nobel Prize laureate professor Akira Yoshino's "Valley of Death". " I got it.

Melanoma identification using Dermo Camera DZ-D100

Next, a lecture by Prof. Yasuhide Tokawa of Chiba University, entitled "Let's Use Dermo Cameras Effectively!", Introduced the successful operation of the DZ-D100, which has been involved in development from prototypes.

  • Prof. Yasuhide Tokawa of Chiba University

The first thing Dr. Tokawa explained about the DZ-D100 was that the resolution of the edge resolution was extremely excellent. The slides show that a comparison of the chart shooting with other companies' cameras and that the end of the chart looks the same as the center without distortion is obtained.

  • Prof. Tokawa highly appreciates the resolution of the edge resolution of the DZ-D100

Dr. Tokawa points out the following seven features of the DZ-D100 and emphasizes that it also supports AI diagnosis.

1. You can take clinical photos.
2. A wide range of dermo photography is possible (1.4 to 3.4 times the area of ​​conventional products).
3. One button for simultaneous shooting of polarized / unpolarized / violet light (UV) and video.
4. High resolution to the periphery, easy to focus on raised lesions.
5. Fluctuation test and diagnosis of scabies can be done easily.
6. Projection to large screen is possible with HDMI terminal.
7. And compact and light.

In the lecture, these characteristics were explained in an easy-to-understand manner using slides. For example, what can we know from simultaneous imaging of polarized, unpolarized, and violet (UV) light?

Shiny white lines appear in the polarized image as artifacts (noise), and a phenomenon called Blink sign, which is not reflected in the non-polarized image, can be easily confirmed. If you switch between polarized and unpolarized when shooting a movie, you can see the Blink sign more clearly in the movie. In addition, in a purple light (UV) image, surrounding pigment spots are clearly visible, and it is easy to identify a lesion having a high possibility of melanoma. Small lesions make diagnosis difficult, but DZ-D100 makes it easier to find small melanomas.

  • Use of DZ-D100 makes it easy to check the Blink sign, and it is easy to distinguish pigment spots by UV imaging

Also important for the effective use of the DZ-D100 is the Melanoma discrimination Chaos and Clues algorithm, which Dr. Tokawa calls the "simplified Kitler method". Roughly speaking, if a pigmented skin lesion is asymmetric and fits at least one of the nine cues, a biopsy is needed (unless pattern analysis reveals a clear diagnosis of seborrheic keratosis) With the content, shooting with the DZ-D100 will lead to not missing such signs of melanoma.

  • FIG. 2 is a diagram of a simplified Kitler method showing a Chaos and Clues algorithm for discriminating melanoma. Dr. Tokawa states that it is important to tap into this algorithm and observe lesions, including exceptions.

Experience booth where you can try the DZ-D100 and various attachments

At the exhibition hall, the sponsoring companies set up booths and exhibited related products and research results. The Casio booth, which launches the DZ-D100 greatly, attracted the attention of visitors by giving them access to the actual equipment and options of the DZ-D100.

  • The Casio booth crowded with many doctors. The doctors who listened eagerly to the explanation of the staff in charge were impressive

At the Casio booth, hearing the voices of doctors who actually use the DZ-D100 at medical sites has helped with development and improvement. The options on display at the booth have been developed to solve doctors' problems, and will continue to actively communicate with medical professionals.

  • DZ-D100 and two options. “ Small-diameter adapter '' (left) for imaging narrow affected areas such as between fingers, and “ three-dimensional adapter '' (right) for imaging tall lesions

  • DZ-D100 attached to microscope using "Microscope Adapter"

At the booth, a "wide wavelength band camera" was exhibited for reference. It is equipped with near-ultraviolet LEDs, near-infrared LEDs, and far-infrared cameras in addition to visible light LEDs, and aims to shoot things that cannot be seen with the eyes.

If you shoot with near ultraviolet light, you can check for skin spots and check for sunscreen. The use of near-infrared light makes it easy to see the veins in the hands and feet, and allows for transparent photography of drinks and packaging bags. In far-infrared, it corresponds to imaging of heat distribution (thermography).

The release time and estimated price are undecided, but development seems to be progressing based on the DZ-D100.

  • "Wide wavelength band camera" for reference exhibit based on the DZ-D100 housing

At the corner of the booth, an AI diagnosis support device was also exhibited for reference. As mentioned in Professor Koga's presentation, this is a project in which Casio and Shinshu University will receive support from AMED.

When the photograph of the lesion is read, the name of a possible disease is presented together with the probability. Lesions can be benign as well as malignant, and doctors use the potential of AI to help them diagnose and prevent oversight and misdiagnosis.

  • Demonstration of AI diagnostic support device under development that was exhibited for reference

Conduct dermosco peak quiz attending with smartphones

Hands-on seminar "Smartphone must-have! Dermosco peak quiz 2019" was held. It is a program to learn how to see the lesion seen with a dermoscope, and to have participants enjoy and improve their diagnostic skills through quiz-style mini-lectures.

Dr. Tanaka first gave a mini-lecture on “melanocyte lesions” and Dr. Tokawa gave a mini-lecture on “non-melanocyte lesions” followed by a quiz. This time, participants adopted their own smartphone or tablet, read the QR code in the handouts, access the dedicated website and adopted the format that they face.

  • One frame explaining dermosco peak quiz

  • You can see that students are seriously working on quizzes aiming to improve dermoscopy skills with smartphones and tablets in hand

Watching participants take a serious look at mini-lectures and quizzes tells us that dermoscopy is no longer an essential diagnostic technique in dermatology. The high participation of young teachers suggests that this will be the standard in the near future.

***

The DZ-D100 is gaining recognition as a tool that plays an important role in dermoscopy diagnosis and its learning because it can accurately and efficiently capture images of diseased parts. However, in order for all the teachers who have introduced the software to be able to use them well, it will be necessary to improve both the hardware and software.

The development of diagnostic support by AI has a long way to go. With the understanding and cooperation of the teachers, we hope that by going step by step, we will definitely contribute to the development of medical care in Japan.

(PR) Courtesy of Casio Computer


Source link

Do you like this article??

Show More

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button