ABSTRACT
KEY POINTS: We proposed a hierarchical framework including an unsupervised candidate image selection and a weakly supervised patch image detection based on multiple instance learning (MIL) to effectively estimate eosinophil quantities in tissue samples from whole slide images. MIL is an innovative approach that can help deal with the variability in cell distribution detection and enable automated eosinophil quantification from sinonasal histopathological images with a high degree of accuracy. The study lays the foundation for further research and development in the field of automated histopathological image analysis, and validation on more extensive and diverse datasets will contribute to real-world application.
ABSTRACT
PURPOSE: This study adopted the cervical and ocular vestibular-evoked myogenic potential (cVEMP and oVEMP) tests in Meniere's disease (MD) patients to correlate them with vestibular endolymphatic hydrops (EH) on MR images. METHODS: A total of 25 patients with unilateral definite MD identified by positive cochlear hydrops on MR images were enrolled. All patients underwent audiometry, cVEMP test and oVEMP test, followed by MR imaging for confirmation. RESULTS: A significantly declining sequence of abnormality rates in MD patients was identified from the audiometry (92%), cVEMP test (52%) to the oVEMP test (40%), which was consistent with a significantly decreasing order of prevalence of EH on MR images running from the cochlea (100%), saccule (56%) to the utricle (52%). The cVEMP test for detecting the saccular hydrops revealed a sensitivity of 62%, while the oVEMP test for assessing the utricular hydrops showed a sensitivity of 70%. However, correlating VEMP results with vestibular hydrops did not show any significant relationship. In addition, mean hearing level (MHL) at four frequencies (500, 1000, 2000, and 3000 Hz) of Grade I cochlear hydrops (51 ± 19 dB) did not significantly differ from Grade II cochlear hydrops (53 ± 19 dB). CONCLUSION: Limitations of the updated MR imaging for visualizing the hydrops comprised: (1) failure to correlate vestibular hydrops with VEMP results, and (2) failure to correlate grade of cochlear hydrops with MHL. The reason is probably because updated MR imaging fails to identify distorted contour of the cochlea/utricle/saccule. Further advanced technique using ultrahigh resolution of fine structures in the inner ear compartments is essential to promote a wider use of MR imaging.
Subject(s)
Magnetic Resonance Imaging , Meniere Disease , Humans , Vestibular Evoked Myogenic Potentials , Meniere Disease/diagnosis , Meniere Disease/diagnostic imaging , Endolymphatic Hydrops/diagnostic imaging , Audiometry , Male , Female , Adult , Middle Aged , AgedABSTRACT
Objective/hypothesis: This study correlated audiological results with magnetic resonance (MR) images to predict positive cochlear endolymphatic hydrops (EH) on MR images in patients with Meniere's disease (MD). Study design: Retrospective study. Methods: Twenty definite MD patients with positive cochlear EH on MR images were assigned to Group A. Another 20 definite MD patients with negative cochlear EH on MR images were assigned to Group B. All patients underwent an inner ear test battery followed by MR imaging using HRDROPS-Mi2 technique. Results: The mean hearing levels (MHLs) at frequencies of 125, 250, 500 and 1000 Hz revealed significantly worse in Group A than Group B. Significantly deteriorated MHLs were noted from Grades 0 to II at low frequency (125, 250, and 500 Hz), but not at mid-frequency and high frequency. The respective cutoff hearing thresholds at frequencies of 125, 250, and 500 Hz were 27.5, 32.5, and 40 dBHL, which help predict positive cochlear EH on MR images. By using the sum (27.5 + 32.5 + 40 = 100 dBHL) of cutoff thresholds from three low frequencies as a cutoff value, Group A (80%) showed significantly more ears with sum of low-frequency hearing threshold >100 dBHL than Group B (30%). Conclusion: When sum of three low-frequency (125, 250, and 500 Hz) hearing levels is >100 dBHL, positive cochlear EH may be shown on MR images in definite MD patients. In contrast, those MD patients with sum of three low-frequency hearing levels <100 dBHL, MR imaging should be postponed because resolution of EH may cause negative MR images. Level of evidence: 4.
ABSTRACT
BACKGROUND/PURPOSE: This study utilized the recently advanced technique in MR imaging to establish its role on diagnosing all types of endolymphatic hydrops (EH). METHODS: Twenty-two patients (26 ears) with clinical EH were admitted due to acute hearing loss and/or vertiginous attack. Each patient underwent an inner ear test battery comprising audiometry, cervical and ocular vestibular-evoked myogenic potential tests, and caloric test, followed by MR imaging for confirmation. RESULTS: Of the 22 clinical EH patients, 12 patients were referred to primary EH (Meniere's disease), 8 patients were secondary EH (including delayed EH in 5), and 2 patients were EH of embryopathic origin. MR imaging of 26 affected ears demonstrated EH in the cochlea only for 14 ears, in the utricle and saccule only for 1 ear, and in all three endorgans for 3 ears, accounting for a sensitivity of 69% (18/26). The 8 affected ears showing negative MR images were EH patients with hearing recovery 1, just after vertiginous attack 3, and chronic low-tone hearing loss 4. In contrast, 3 out of 18 unaffected ears demonstrated asymptomatic EH in the cochlea, representing a specificity of 83% (15/18). CONCLUSION: The sensitivity and specificity of MR imaging for confirming all types of EH were 69% and 83%, respectively. Although diagnostic criteria can identify primary and delayed EH, MR imaging may provide a supplementary tool for diagnosing secondary, embryopathic, or asymptomatic EH, if patients are not with hearing recovery, chronic low-tone hearing loss, or just after vertiginous attack.
