Hitachi India Pvt. Ltd. (“Hitachi India”) and Hitachi Ltd today announced that they have developed a technology to extract precisely designated information from electronic medical records with International Institute of Information Technology, Hyderabad (“IIIT-H”). To apply the technology makes it possible to analyse electronic medical records including various mixed-forms such as free-form texts written by doctors or laboratory data written in given formats. It is expected that more advanced medical services will be realised by that, such as prediction of the incidence rates of disease in the future.
Hitachi India, Hitachi and IIIT-H won 1st prize with an extraction precision of 86.8 per cent in CLEF e-Health 2014 Task2, a technical competition on medical information extraction, which was held by a European Research Forum CLEF from Jan. to May 2014. The technology will be presented at CLEF Conference 2014 in UK from 15th to 18th Sep. 2014.
To predict the incidence rates of disease, it is necessary to analyse medical record texts such as doctors’ remarks as well as the input information in the given format such as blood test results. Many research institutes have been developing the technology to extract necessary information because the medical record texts are written in different ways by doctors.
Ichiro Iino, managing director, Hitachi India said, “We are proud that we won 1st prize in this competition.Hitachi India, Hitachi and IIIT-H have collectively worked as a strong team to achieve this technological innovation. We firmly believe that it would help build a sustainable society with reduced health risk. We are honoured to achieve this mark, which will act as a catalyst.”
This competition has been held since 2013 aiming at enabling patients to easily read medical records containing many medical terms, and it is divided into 3 tasks based on the technical objectives. Ten research organisations including a team of Hitachi India, Hitachi and IIIT-H, participated in Task2, which competes in the precision to extract information such as body location, severity, etc. of disease, from electronic medical records. The team won 1st prize with the precision of 86.8 per cent in the competition.
In order to improve extraction precision, machine learning technologies were installed, which can construct rules to determine answers from preset learning data. Moreover, the extraction precisionis improved by utilising the information on document structures such as the text information of chapter, section, etc. as well as main texts.
In order to increase the precision more, 2 post-processing modules are added. The first module converts the extracted information into more precise one by considering contexts around it. The next module determines the meanings of ambiguous phrases by using appropriate dictionaries to the type of input medical records. These modules improved the precision to extract body location information from 60 per cent to 74 per cent.
In the development of the technology, the team adopted Apache cTAKES, an open source program for medical text analysis, which improved efficiency of development, and enabled further expansion of features.
Prof. P.J. Narayanan, Director, International Institute of Information Technology, Hyderabad said, “We are delighted to achieve this breakthrough in association with Hitachi in the field of Technology and Healthcare. This Technology will help a great deal in prevention or reduction of disease risks. It was a learning experience to work with Hitachi, which shares the same passion for producing futuristic technology as we do. We hope to create such fruitful alliances in future too.”
Hitachi India and Hitachi will continue to improve this extraction technology, and develop new technologies leading to the creation of advanced medical services.