Umeå University's logo

Background: Otitis media includes several common inflammatory conditions of the middle ear that can have severe complications if left untreated. Correctly identifying otitis media can be difficult and a screening system supported by machine learning would be valuable for this prevalent disease. This study investigated the performance of a convolutional neural network in screening for otitis media using digital otoscopic images labelled by an expert panel.

Methods: Five experienced otologists diagnosed 347 tympanic membrane images captured with a digital otoscope. Images with a majority expert diagnosis (n = 273) were categorized into three screening groups Normal, Pathological and Wax, and the same images were used for training and testing of the convolutional neural network. Expert panel diagnoses were compared to the convolutional neural network classification. Different approaches to the convolutional neural network were tested to identify the best performing model.

Results: Overall accuracy of the convolutional neural network was above 0.9 in all except one approach. Sensitivity to finding ears with wax or pathology was above 93% in all cases and specificity was 100%. Adding more images to train the convolutional neural network had no positive impact on the results. Modifications such as normalization of datasets and image augmentation enhanced the performance in some instances.

Conclusions: A machine learning approach could be used on digital otoscopic images to accurately screen for otitis media.

Background: There is a lack of hearing health care globally, and tele-audiology and mobile technologies have been proposed as important strategies to reduce the shortfall. Objectives: To investigate the accuracy and reliability of smartphone self-test audiometry in adults, in community clinics in low-income settings.

Methods: A prospective, intra-individual, repeated measurements design was used. Sixty-three adult participants (mean age 52 years, range 20-88 years) were recruited from ENT and primary health care clinics in a low-income community in Tshwane, South Africa. Air conduction hearing thresholds for octave frequencies 0.5 to 8 kHz collected with the smartphone self-test in non-sound treated environments were compared to those obtained by reference audiometry.

Results: The overall mean difference between threshold seeking methods (ie, smartphone thresholds subtracted from reference) was -2.2 dB HL (n = 467 thresholds, P = 0.00). Agreement was within 10 dB HL for 80.1% (n = 467 thresholds) of all threshold comparisons. Sensitivity for detection hearing loss >40 dB HL in one ear was 90.6% (n = 84 ears), and specificity 94.2% (n = 84 ears).

Conclusion: Smartphone self-test audiometry can provide accurate and reliable air conduction hearing thresholds for adults in community clinics in low-income settings.

OBJECTIVE: To validate a calibrated smartphone-based hearing test in a sound booth environment and in primary health-care clinics.

DESIGN: A repeated-measure within-subject study design was employed whereby air-conduction hearing thresholds determined by smartphone-based audiometry was compared to conventional audiometry in a sound booth and a primary health-care clinic environment.

STUDY SAMPLE: A total of 94 subjects (mean age 41 years ± 17.6 SD and range 18-88; 64% female) were assessed of whom 64 were tested in the sound booth and 30 within primary health-care clinics without a booth.

RESULTS: In the sound booth 63.4% of conventional and smartphone thresholds indicated normal hearing (≤15 dBHL). Conventional thresholds exceeding 15 dB HL corresponded to smartphone thresholds within ≤10 dB in 80.6% of cases with an average threshold difference of -1.6 dB ± 9.9 SD. In primary health-care clinics 13.7% of conventional and smartphone thresholds indicated normal hearing (≤15 dBHL). Conventional thresholds exceeding 15 dBHL corresponded to smartphone thresholds within ≤10 dB in 92.9% of cases with an average threshold difference of -1.0 dB ± 7.1 SD.

CONCLUSIONS: Accurate air-conduction audiometry can be conducted in a sound booth and without a sound booth in an underserved community health-care clinic using a smartphone.