In order to bring you the best possible user experience, this site uses Javascript. If you are seeing this message, it is likely that the Javascript option in your browser is disabled. For optimal viewing of this site, please ensure that Javascript is enabled for your browser.

This content is currently on FREE ACCESS

Comparison of whether a beginner can be close to an expert with an artificial neural network in myocardial perfusion imaging

Session Poster session III

Speaker Akiyo Chiba

Event : ICNC, Nuclear Cardiology & Cardiac CT 2019

  • Topic : imaging
  • Sub-topic : Single Photon Emission Computed Tomography (SPECT)
  • Session type : Poster Session

Authors : A Chiba (Nagasaki,JP), T Kudo (Nagasaki,JP), R Ideguchi (Nagasaki,JP), M Altay (Nagasaki,JP), S Koga (Nagasaki,JP), T Yonekura (Nagasaki,JP), A Tsuneto (Nagasaki,JP), M Morikawa (Nagasaki,JP), S Ikeda (Nagasaki,JP), H Kawano (Nagasaki,JP), Y Koide (Nagasaki,JP), M Uetani (Nagasaki,JP), K Maemura (Nagasaki,JP)

A Chiba1 , T Kudo2 , R Ideguchi2 , M Altay2 , S Koga1 , T Yonekura1 , A Tsuneto1 , M Morikawa3 , S Ikeda1 , H Kawano1 , Y Koide4 , M Uetani3 , K Maemura1 , 1Nagasaki University Hospital, Cardiovascular Medicine - Nagasaki - Japan , 2Nagasaki University, Radioisotope Medicine - Nagasaki - Japan , 3Nagasaki University Hospital, Radiological Sciences - Nagasaki - Japan , 4Nagasaki Memorial Hospital, Nagasaki Medical Education Center - Nagasaki - Japan ,

Single Photon Emission Computed Tomography (SPECT)

Background/Introduction: Interpretation of myocardial perfusion imaging (MPI) requires skilled expert reading. Artificial intelligence including artificial neural network (ANN) is expected to help beginners in diagnostic imaging with ANN.
Purpose: With ANN in stress MPI, whether beginners in cardiology diagnostic imaging can get closer results of experts. 
Methods: The subjects were 138 patients who underwent stress MPI (with Tc-labeled agents)at Nagasaki University Hospital. Among them, 45 patients (33%) with multiple vessel disease, 28 patients (20%) with a history of myocardial infarction, and 17 patients (12%) with both multiple vessel disease and a history of myocardial infarction. An expert over 30 years interpretation experience and a beginner within a few years experience interpreted stress MPI with/without software which implemented ANN (cardioREPO) and compared the results. The myocardium was divided into 5 regions of apex, septum, anterior, lateral, and inferior, and the degree of myocardial blood flow was evaluated as 5 stages of 0 to 4, 5 segment model, and Summed Stress Score (SSS), Summed Rest Score (SRS) and Summed Difference Score (SDS). These numbers were obtained with and without help of ANN for both interpreters.
Results: To investigate the influence by ANN, "effect of ANN" expressed as the difference of each score (SSS, SRS, SDS) between with ANN and without ANN ("with" – "without") was calculated both in beginner and expert. "Skill effect" expressed as the difference of each score between expert and beginner ("expert" – "beginner") was also calculated both with ANN and without ANN.
"Effect of ANN" was smaller in expert than in beginner. (For SSS; expert vs. beginner = -0.49 vs. -1.23, p value <0.0001, For SRS; expert vs. beginner SRS - beginner = -0.34 vs. -0.88, p value 0.0003, For SDS; expert vs. beginner = -0.15 vs. -0.36, p value 0.0128). All the effect of ANN showed negative value both of them, which indicates that both interpreters tend to interpret score smaller with ANN than without. With ANN the "Skill effect"became smaller than without. (For SSS; with ANN vs. without = -0.07 vs. -0.8, p value <0.0001?For SRS; with ANN vs. without = -0.17 vs. -0.7, p value 0.0003?For SDS; with ANN vs. without = 0.101 vs. -0.1, p value 0.0128). When the score of expert with ANN was considered as "Standard Expert Score", the score of beginner got closer to "Standard Expert Score" with ANN than without("Standard Expert Score"-beginner with ANN vs. "Standard Expert Score"-beginner without ANN for SSS, SRS and SSS are; -0.07 vs. -0.8, p value <0.0001, -0.17 vs. -0.7, p value 0.0003, 0.101 vs. -0.1, p value 0.0128). However, a beginner may interpret no ischemia for actually ischemia because a beginner is affected strongly by ANN.
Conclusions: In the interpretation of MPI, interpretation of beginner gets closer to expert interpretation with ANN than that without. Interpretation with ANN may be useful in interpretation by beginners.

Get your access to resources

Join now
  • 1ESC Professional Members – access all ESC Congress resources 
  • 2ESC Association Members (Ivory, Silver, Gold) – access your Association’s resources
  • 3Under 40 or in training - with a Combined Membership, access all resources
Join now
logo esc

Our mission: To reduce the burden of cardiovascular disease

Who we are