images using AI to determine if the image is a tumor or not. Software that infers a
tumor and presents it along with its potential was developed. The safety and efficacy
of this method were recognized in December 2018, and approval under the Pharmaceuticals and Medical Devices Act was obtained. The developed software
EndoBRAIN
® uses a support vector machine for machine learning.
EndoBRAIN
® processes the endoscopic image information of the large intestine
taken by Olympus’s ultra-magnifying endoscope Endocyto. EndoBRAIN
® has the
ability to output tumor and non-tumor potentials as numerical values from images
and thus can assist physicians in predicting and diagnosing lesions. EndoBRAIN
®
could learn approximately 60,000 endoscope images based on support vector
machine. The clinical performance test identified neoplastic polyps and
nonneoplastic polyps with an accuracy of 98% and a sensitivity of 97%, which
was comparable to that of specialists and exceeded that of nonspecialists (Japan
Agency for Medical Research and Development 2019).
5.3 Detection of Influenza Follicles with AI
For entrepreneurs, the intersection of healthcare and AI composes one of the
attractive and emerging market. For most clinicians, AI is an unknown world. In
fact, many doctors view it as a threat that could steal their work. However, as we
have discussed so far, the combination of clinical and AI heterogeneity also triggers
innovation from the perspective of business start-ups.
The movement to use AI for diagnosis of familiar diseases is also becoming
active. If we include patients who rely on self-diagnosis and do not go to the hospital,
we may actually have more than 25 million flu cases a year. However, for patients
who are given the therapeutic agent Tamiflu 48 h or more after the onset of influenza,
no noticeable effect is observed. The current examination method has approximately
60% diagnostic accuracy because it is not sufficient unless 24 h or more have passed
since the onset. Note that if a positive case of influenza is incorrectly diagnosed as
negative, the carrier may risk spreading the disease, for example, by commuting to
school (Chartrand, Leeflang, Minion, Brewer, & Pai, 2012).
Aillis Inc. (Japan Agency for Medical Research and Development 2019),
established in November 2017 by emergency doctor Sho Okiyama, aims to support
efficient and effective influenza diagnosis by AI. His encounter with a paper in 2013
motivated Dr. Okiyama to start the company. The paper described a flu-specific
swelling of the pharynx that could help an expert physician identify the flu with
nearly 99% accuracy by examining the flu follicles (Japan Agency for Medical
Research and Development 2019).
In influenza patients, a swelling called flu follicles presents in their throat.
Although there is a similar swelling in the back of the throat of patients with common
cold or even totally healthy people, it was discovered that this follicle has a feature
that appears only in the case of influenza. It is a combination of various features such
as the color tone and luster of the surface, size, and manner of swelling. For decades,
it was known that even objects that inexperienced physicians could recognize as the
Innovation in Health Informatics
17
tumor and presents it along with its potential was developed. The safety and efficacy
of this method were recognized in December 2018, and approval under the Pharmaceuticals and Medical Devices Act was obtained. The developed software
EndoBRAIN
® uses a support vector machine for machine learning.
EndoBRAIN
® processes the endoscopic image information of the large intestine
taken by Olympus’s ultra-magnifying endoscope Endocyto. EndoBRAIN
® has the
ability to output tumor and non-tumor potentials as numerical values from images
and thus can assist physicians in predicting and diagnosing lesions. EndoBRAIN
®
could learn approximately 60,000 endoscope images based on support vector
machine. The clinical performance test identified neoplastic polyps and
nonneoplastic polyps with an accuracy of 98% and a sensitivity of 97%, which
was comparable to that of specialists and exceeded that of nonspecialists (Japan
Agency for Medical Research and Development 2019).
5.3 Detection of Influenza Follicles with AI
For entrepreneurs, the intersection of healthcare and AI composes one of the
attractive and emerging market. For most clinicians, AI is an unknown world. In
fact, many doctors view it as a threat that could steal their work. However, as we
have discussed so far, the combination of clinical and AI heterogeneity also triggers
innovation from the perspective of business start-ups.
The movement to use AI for diagnosis of familiar diseases is also becoming
active. If we include patients who rely on self-diagnosis and do not go to the hospital,
we may actually have more than 25 million flu cases a year. However, for patients
who are given the therapeutic agent Tamiflu 48 h or more after the onset of influenza,
no noticeable effect is observed. The current examination method has approximately
60% diagnostic accuracy because it is not sufficient unless 24 h or more have passed
since the onset. Note that if a positive case of influenza is incorrectly diagnosed as
negative, the carrier may risk spreading the disease, for example, by commuting to
school (Chartrand, Leeflang, Minion, Brewer, & Pai, 2012).
Aillis Inc. (Japan Agency for Medical Research and Development 2019),
established in November 2017 by emergency doctor Sho Okiyama, aims to support
efficient and effective influenza diagnosis by AI. His encounter with a paper in 2013
motivated Dr. Okiyama to start the company. The paper described a flu-specific
swelling of the pharynx that could help an expert physician identify the flu with
nearly 99% accuracy by examining the flu follicles (Japan Agency for Medical
Research and Development 2019).
In influenza patients, a swelling called flu follicles presents in their throat.
Although there is a similar swelling in the back of the throat of patients with common
cold or even totally healthy people, it was discovered that this follicle has a feature
that appears only in the case of influenza. It is a combination of various features such
as the color tone and luster of the surface, size, and manner of swelling. For decades,
it was known that even objects that inexperienced physicians could recognize as the
Innovation in Health Informatics
17