b'collaborations with experts in other medical specialties,or algorithms, an extensive database is critical for machine learning engineering, data science, regulatory process, etc., we can trulyaccuracy and efficiency. Going through thousands of radiographic accelerate the pace of new discoveries and the clinical adoptionimages and annotating each tooth for caries, furcation, vertical of innovation to benefit individual patients while improving oralbone defects, bone, implants, restorations, root canals, etc. in health on a populational level. every image while managing residency and life is very exhausting Jiman Nelson, DMD, MSand time consuming.University of Texas Health and ScienceOvercoming the overwhelming load of building an adequate image Center at Houston database was purely due to having a great research team, faculty, WinnerClinical Researchand co-residents with me to make sure I stayed on track. From Award, Balint Orban Memorialmy co-residents making sure I knew my deadlines for courses Competitionand clinicals, to the research team helping me with some of the workload, I managed to find a balance that could support the daily Deep Learning Based Computer-Aidedrigor of residency. Periodontal Diagnosis through the Online PlatformHow will winning this award affect your work going Can you briefly explain the results of your winningforward?research?Winning this award has further affirmed my understanding of Assessing radiographic bone level is important for periodontalthe raw materials of success. For instance, understanding the diagnosis. The interpretation of radiographic images isprincipals of putting in goal-oriented efforts, with help, cultivates subjective, and accuracy depends on a clinicians experiencedesired results and advancement. This translates into my everyday and knowledge. Artificial intelligence and image analysis canlife, in and out of work, and future goals that I will obtain later on. improve reliability. This study aimed to develop and validate a deep learning model that can accurately diagnose periodontalWhat advice would you give to up-and-coming disease using radiographic bone loss, similar to a trained dentalperiodontists?clinician. Also, an online user interface was developed for clinicalThe advice I would give any up-and-coming periodontist is the use, using the deep learning algorithm, and it was validated forfollowing: First, dont be afraid to tackle hard or difficult work. accuracy and usability.Second, stay positive and look for answers in every result obtained In total, 1,147 periapical and bitewing radiographic imagesto help you get better. Third, always do your research to deliver the were analyzed, and 366 vertical bone loss and 357 furcationhighest quality product possible. Fourth, be optimistic and never involvement sites were present. The performance ofgive up because the toughest times are when you progress the segmentation models is highly acceptable. The AUROC valuesmost in character and experience.of RBL stage assignment for no bone loss, stage I, stage II,Shy Chwen Ni, BDS and stage III were 0.98, 0.89, 0.90 and 0.90, respectively.Harvard School of Dental Medicine, Boston, MA Sensitivity and specificity for different RBL stages are all above 0.8. There was no significant difference in RBLWinnerBasic Science and Clinical percentage measurements of vertical bone loss sites betweenImpact Awards, Research Forum Poster the CAD model and experts. The computer-assigned toothCompetitionnumbers were highly accurate with a 0.99 detection accuracy.Mastication Mimicking Mechanical Stretch of The time required to generate a comprehensive report for eachFibroblast Induces Active Neurotropic FactorsFMS by the experts was about 20.40 minutes, which was significantly longer than that by the CAD model of 2.16 minutes.Can you briefly explain the results of your winning The accuracy of 100 clinical case diagnosis was 0.84. Theresearch?accuracy for extent (0.90), stage (0.90), and grade (0.98) wereOne major role of the periodontal ligament (PDL) is to provide generally high. proprioception, which is achieved by Ruffini nerve endings (RNE). This reliable CAD online platform can be used as an adjunctThe molecular mechanism of how RNE migrates to the developing diagnosis tool in the clinic and is helpful when a large numberPDL remained unclear. Our in vitro study demonstrated that of radiographs must be reviewed for clinical quality control andmechanical stimuli due to tooth occlusion is necessary to recruit the research purposes.RNE into the PDL. Fibroblast cells were stretched under mastication mimicking pattern using a biaxial stretching machine, and there What challenges did you face while conducting yourwas a two-fold increase of brain-derived neurotrophic factors research, and how did you overcome them?The biggest challenge that I faced while conducting my researchContinued on page 42was time management. When training deep learning computers AAP Periospectives| 41'