Park C, et al. Yonsei Med J. 2022 Jan;63(Suppl):S74-S83. https://doi.org/10.3349/ymj.2022.63.S74

In this paper, a smartphone-based lung function test, developed to estimate lung function parameters using a high-resolution time-frequency spectrum from a smartphone built-in microphone is presented. A method of estimation of the forced expiratory volume in 1 s divided by forced vital capacity (FEV₁/FVC) based on the variable frequency complex demodulation method (VFCDM) is first proposed. We evaluated our proposed method on 26 subjects, including 13 healthy subjects and 13 chronic obstructive pulmonary disease (COPD) patients, by comparing with the parameters clinically obtained from pulmonary function tests (PFTs). For the healthy subjects, we found that an absolute error (AE) and a root mean squared error (RMSE) of the FEV₁/FVC ratio were 4.49% ± 3.38% and 5.54%, respectively. For the COPD patients, we found that AE and RMSE from COPD patients were 10.30% ± 10.59% and 14.48%, respectively. For both groups, we compared the results using the continuous wavelet transform (CWT) and short-time Fourier transform (STFT), and found that VFCDM was superior to CWT and STFT. Further, to estimate other parameters, including forced vital capacity (FVC), forced expiratory volume in 1 s (FEV₁), and peak expiratory flow (PEF), regression analysis was conducted to establish a linear transformation. However, the parameters FVC, FEV1, and PEF had correlation factor r values of 0.323, 0.275, and −0.257, respectively, while FEV₁/FVC had an r value of 0.814. The results obtained suggest that only the FEV1/FVC ratio can be accurately estimated from a smartphone built-in microphone. The other parameters, including FVC, FEV1, and PEF, were subjective and dependent on the subject’s familiarization with the test and performance of forced exhalation toward the microphone.

Background: Sarcopenia is defined as muscle wasting, characterized by a progressive loss of muscle mass and function due to ageing. Diagnosis of sarcopenia typically involves both muscle imaging and the physical performance of people exhibiting signs of muscle weakness. Despite its worldwide prevalence, a molecular method for accurately diagnosing sarcopenia has not been established.Methods: We develop an artificial intelligence (AI) diagnosis model of sarcopenia using a published transcriptome dataset comprising patients from multiple ethnicities. For the AI model for sarcopenia diagnosis, we use a transcriptome database comprising 17 339 genes from 118 subjects. Among the 17 339 genes, we select 27 features as the model inputs. For feature selection, we use a random forest, extreme gradient boosting and adaptive boosting. Using the top 27 features, we propose a four‐layer deep neural network, named DSnet‐v1, for sarcopenia diagnosis.Results: Among isolated testing datasets, DSnet‐v1 provides high sensitivity (100%), specificity (94.12%), accuracy (95.83%), balanced accuracy (97.06%) and area under receiver operating characteristics (0.99). To extend the number of patient data, we develop a web application (http://sarcopeniaAI.ml/), where the model can be accessed unrestrictedly to diagnose sarcopenia if the transcriptome is available. A focused analysis of the top 27 genes for their differential or co‐expression with other genes implied the potential existence of race‐specific factors for sarcopenia, suggesting the possibility of identifying causal factors of sarcopenia when a more extended dataset is provided.Conclusions: Our new AI model, DSnet‐v1, accurately diagnoses sarcopenia and is currently available publicly to assist healthcare providers in diagnosing and treating sarcopenia.

The liver morphological changes in relation to fibrosis stage in nonalcoholic fatty liver disease (NAFLD) have not yet been clearly understood. This study was to develop a liver surface nodularity (LSN) quantification program and to compare the fibrosis grades in simple steatosis (SS) and nonalcoholic steatohepatitis (NASH). Thirty subjects (7 normal controls [NC], 12 SS and 11 NASH) were studied. LSN quantification procedure was bias correction, boundary detection, segmentation and LSN measurement. LSN scores among three groups and fibrosis grades compared using Kruskal–Wallis H test. Diagnostic accuracy was determined by calculating the area under the receiver operating characteristics (ROC) curve. Mean LSN scores were NC 1.30 ± 0.09, SS 1.54 ± 0.21 and NASH 1.59 ± 0.23 (p = 0.008). Mean LSN scores according to fibrosis grade (F) were F0 1.30 ± 0.09, F1 1.45 ± 0.17 and F2&F3 1.67 ± 0.20 (p = 0.001). The mean LSN score in F2&F3 is significantly higher than that in F1 (p = 0.019). The AUROC curve to distinguish F1 and F2&F3 was 0.788 (95% CI 0.595–0.981, p = 0.019) at a cut-off LSN score greater than 1.48, and its diagnostic accuracy had 0.833 sensitivity and 0.727 specificity. This study developed LSN program and its clinical application demonstrated that the quantitative LSN scores can help to differentially diagnose fibrosis stage in NAFLD.

The gold standard test for diagnosing dysphagia is the videofluoroscopic swallowing study (VFSS). However, the accuracy of this test varies depending on the specialist’s skill level. We proposed a VFSS-based artificial intelligence (AI) web application to diagnose dysphagia. Video from the VFSS consists of multiframe data that contain approximately 300 images. To label the data, the server separated them into frames during the upload and stored them as a video for analysis. Then, the separated data were loaded into a labeling tool to perform the labeling. The labeled file was downloaded, and an AI model was developed by training with You Only Look Once (YOLOv7). Using a utility called SplitFolders, the entire dataset was divided according to a ratio of training (70%), test (10%), and validation (20%). When a VFSS video file was uploaded to an application equipped with the developed AI model, it was automatically classified and labeled as oral, pharyngeal, or esophageal. The dysphagia of a person was categorized as either penetration or aspiration, and the final analyzed result was displayed to the viewer. The following labeling datasets were created for the AI learning: oral (n = 2355), pharyngeal (n = 2338), esophageal (n = 1480), penetration (n = 1856), and aspiration (n = 1320); the learning results of the YOLO model, which analyzed dysphagia using the dataset, were predicted with accuracies of 0.90, 0.82, 0.79, 0.92, and 0.96, respectively. This is expected to help clinicians more efficiently suggest the proper dietary options for patients with oropharyngeal dysphagia.

Liver biopsy is the reference standard test to differentiate between non-alcoholic steatohepatitis (NASH) and simple steatosis (SS) in non-alcoholic fatty liver disease (NAFLD), but noninvasive diagnostics are warranted. The diagnostic accuracy in NASH using MR imaging modality have not yet been clearly identified. This study was assessed the accuracy of magnetic resonance imaging (MRI) method for diagnosing NASH. Data were extracted from research articles obtained after a literature search from multiple electronic databases. Random-effects meta-analyses were performed to obtain overall effect size of the area under the receiver operating characteristic(ROC) curve, sensitivity, specificity, likelihood ratios(LR), diagnostic odds ratio(DOR) of MRI method in detecting histopathologically-proven SS(or non-NASH) and NASH. Seven studies were analyzed 485 patients, which included 207 SS and 278 NASH. The pooled sensitivity was 87.4% (95% CI, 76.4–95.3) and specificity was 74.3% (95% CI, 62.4–84.6). Pooled positive LR was 2.59 (95% CI, 1.96–3.42) and negative LR was 0.17 (95% CI, 0.07–0.38). DOR was 21.57 (95% CI, 7.27–63.99). The area under the curve of summary ROC was 0.89. Our meta-analysis shows that the MRI-based diagnostic methods are valuable additions in detecting NASH.