2023年 · 第03卷第04期

Ptosis is a common ophthalmologic condition, and the diagnosis is primarily based on ocular appearance. The diagnosis of such conditions can be improved using emerging technology such as artificial intelligence-based methods. However, unified data collection and labeling standards have not yet been established. This directly impacts the accuracy of ptosis diagnosis based on appearance alone. Therefore, in the present study, we aimed to establish a procedure to obtain and label images to devise a recommendation system for optimal recognition of ptosis based on ocular appearances. This would help to standardize and facilitate data sharing and serve as a guideline for the development and improvisation of algorithms in artificial intelligence for ptosis.

The trochanter of the femur is a common site for bone tumors. However, locating the specific boundary of bone tumor infiltration and determining the surgical method can be challenging. The objective of this study was to review the diagnosis, treatment, and surgical outcomes of patients with tumors or tumor-like changes in the femoral trochanter after computer-assisted precise tumor resection and hip-preserving reconstruction of the trochanter.

From January 2005 to September 2020, 11 patients with trochanteric tumors (aged: 18–53 years; six males and five females) were treated in Guangzhou First People’s Hospital. The cases included aneurysmal bone cyst (n = 1), giant cell tumor of bone (n = 2), fibrous histiocytoma of bone (n = 1), endochondroma (n = 1), and fibrous dysplasia of bone (n = 6). For patients with trochanteric tumors, computed tomography and magnetic resonance imaging scanning were performed before operation to obtain two-dimensional image data of the lesion. A three-dimensional digital model of bilateral lower limbs was reconstructed by computer technology, the boundary of tumor growth was determined by computer simulation, the process of tumor resection and reconstruction was simulated, and the personalized guide template was designed. During the operation, the personalized guide plate guided the precise resection of the tumor, and the allogeneic bone was trimmed to match the shape of the bone defect.

All 11 patients underwent accurate resection of the tumor or tumor-like lesion and reconstruction of the hip. In eight cases, the lesion was confined to the trochanter, which was fixed with large segment allogeneic bone, autologous iliac bone, and proximal femoral anatomic plate. In three cases, allogeneic bone, autologous iliac bone, and femoral reconstruction nail were used to fix the tumor under the trochanter. Postoperative X-ray examination showed that the repair and reconstruction of the bone defect was effective, and callus bridging between the allogenic bone and autogenous bone was observed 6 months after operation. All patients recovered their walking function 3–6 months after operation. The duration of the follow-up period ranged from 6 months to 6 years. A patient experienced recurrence of endochondroma; pathological examination revealed chondrocytic sarcoma. The remaining 10 patients were treated with segmental resection and reconstruction. The operation time ranged 2.5–4.5 h (average: 3.2 h). Intraoperative blood loss ranged from 300 to 500 ml (average: 368 ml). The local recurrence rate was 9.1%, and the overall survival rate was 100%. The average Musculoskeletal Tumor Society score was 27 (excellent and good for eight and three patients, respectively).

Three-dimensional computer skeleton modeling and simulation-assisted resection and reconstruction of femoral trochanteric tumor is a new surgical technique, which might markedly improve the surgical effect, shorten the surgical time, increase the overall survival rate of patients with tumors, reduce the local recurrence rate, assist in the digitization and programming of femoral trochanteric tumor surgery, and improve surgical accuracy.

Knee arthroscopy is one of the most complex minimally invasive surgeries, and it is routinely performed to treat a range of ailments and injuries to the knee joint. Its complex ergonomic design imposes visualization and navigation constraints, consequently leading to unintended tissue damage and a steep learning curve before surgeons gain proficiency. The lack of robust visual texture and landmark frame features further limits the success of image-guided approaches to knee arthroscopy Feature- and texture-less tissue structures of knee anatomy, lighting conditions, noise, blur, debris, lack of accurate ground-truth label, tissue degeneration, and injury make semantic segmentation an extremely challenging task. To address this complex research problem, this study reported the utility of reconstructed surface reflectance as a viable piece of information that could be used with cutting-edge deep learning technique to achieve highly accurate segmented scenes.

We proposed an intraoperative, two-tier deep learning method that makes full use of tissue reflectance information present within an RGB frame to segment texture-less images into multiple tissue types from knee arthroscopy video frames. This study included several cadaver knees experiments at the Medical and Engineering Research Facility, located within the Prince Charles Hospital campus, Brisbane Queensland. Data were collected from a total of five cadaver knees, three were males and one from a female. The age range of the donors was 56–93 years. Aging-related tissue degeneration and some anterior cruciate ligament injury were observed in most cadaver knees. An arthroscopic image dataset was created and subsequently labeled by clinical experts. This study also included validation of a prototype stereo arthroscope, along with conventional arthroscope, to attain larger field of view and stereo vision. We reconstructed surface reflectance from camera responses that exhibited distinct spatial features at different wavelengths ranging from 380 to 730 nm in the RGB spectrum. Toward the aim to segment texture-less tissue types, this data was used within a two-stage deep learning model.

The accuracy of the network was measured using dice coefficient score. The average segmentation accuracy for the tissue-type articular cruciate ligament (ACL) was 0.6625, for the tissue-type bone was 0.84, and for the tissue-type meniscus was 0.565. For the analysis, we excluded extremely poor quality of frames. Here, a frame is considered extremely poor quality when more than 50% of any tissue structures are over- or underexposed due to nonuniform light exposure. Additionally, when only high quality of frames was considered during the training and validation stage, the average bone segmentation accuracy improved to 0.92 and the average ACL segmentation accuracy reached 0.73. These two tissue types, namely, femur bone and ACL, have a high importance in arthroscopy for tissue tracking. Comparatively, the previous work based on RGB data achieved a much lower average accuracy for femur, tibia, ACL, and meniscus of 0.78, 0.50, 0.41, and 0.43 using U-Net and 0.79, 0.50, 0.51, and 0.48 using U-Net++. From this analysis, it is clear that our multispectral method outperforms the previously proposed methods and delivers a much better solution in achieving automatic arthroscopic scene segmentation.

The method was based on the deep learning model and requires a reconstructed surface reflectance. It could provide tissue awareness in an intraoperative manner that has a high potential to improve surgical precisions. It could be applied to other minimally invasive surgeries as an online segmentation tool for training, aiding, and guiding the surgeons as well as image-guided surgeries.

To analyze the characteristics of tongue imaging color parameters in patients treated with percutaneous coronary intervention (PCI) and non-PCI for coronary atherosclerotic heart disease (CHD), and to observe the effects of PCI on the tongue images of patients as a basis for the clinical diagnosis and treatment of patients with CHD.

This study used a retrospective cross-sectional survey to analyze tongue photographs and medical history information from 204 patients with CHD between November 2018 and July 2020. Tongue images of each subject were obtained using the Z-BOX Series traditional Chinese medicine (TCM) intelligent diagnosis instruments, the SMX System 2.0 was used to transform the image data into parameters in the HSV color space, and finally the parameters of the tongue image between patients in the PCI-treated and non-PCI-treated groups for CHD were analyzed.

Among the 204 patients, 112 were in the non-PCI treatment group (38 men and 74 women; average age of (68.76 ± 9.49) years), 92 were in the PCI treatment group (66 men and 26 women; average age of (66.02 ± 10.22) years). In the PCI treatment group, the H values of the middle and tip of the tongue and the overall coating of the tongue were lower (P < 0.05), while the V values of the middle, tip, both sides of the tongue, the whole tongue and the overall coating of the tongue were higher (P < 0.05).

The color parameters of the tongue image could reflect the physical state of patients treated with PCI, which may provide a basis for the clinical diagnosis and treatment of patients with CHD.

Oral cancer is one of the most common types of cancer in men causing mortality if not diagnosed early. In recent years, computer-aided diagnosis (CAD) using artificial intelligence techniques, in particular, deep neural networks have been investigated and several approaches have been proposed to deal with the automated detection of various pathologies using digital images. Recent studies indicate that the fusion of images with the patient’s clinical information is important for the final clinical diagnosis. As such dataset does not yet exist for oral cancer, as far as the authors are aware, a new dataset was collected consisting of histopathological images, demographic and clinical data. This study evaluated the importance of complementary data to histopathological image analysis of oral leukoplakia and carcinoma for CAD.

A new dataset (NDB-UFES) was collected from 2011 to 2021 consisting of histopathological images and information. The 237 samples were curated and analyzed by oral pathologists generating the gold standard for classification. State-of-the-art image fusion architectures and complementary data (Concatenation, Mutual Attention, MetaBlock and MetaNet) using the latest deep learning backbones were investigated for 4 distinct tasks to identify oral squamous cell carcinoma, leukoplakia with dysplasia and leukoplakia without dysplasia. We evaluate them using balanced accuracy, precision, recall and area under the ROC curve metrics.

Experimental results indicate that the best models present balanced accuracy of 83.24% using images, demographic and clinical information with MetaBlock fusion and ResNetV2 backbone. It represents an improvement in performance of 30.68% (19.54 pp) in the task to differentiate samples diagnosed with oral squamous cell carcinoma and leukoplakia with or without dysplasia.

This study indicates that cured demographic and clinical data may positively influence the performance of artificial intelligence models in automated classification of oral cancer.

This study aimed to explore the mortality prediction of patients with cerebrovascular diseases in the intensive care unit (ICU) by examining the important signals during different periods of admission in the ICU, which is considered one of the new topics in the medical field. Several approaches have been proposed for prediction in this area. Each of these methods has been able to predict mortality somewhat, but many of these techniques require recording a large amount of data from the patients, where recording all data is not possible in most cases; at the same time, this study focused only on heart rate variability (HRV) and systolic and diastolic blood pressure.

The ICU data used for the challenge were extracted from the Multiparameter Intelligent Monitoring in Intensive Care II (MIMIC-II) Clinical Database. The proposed algorithm was evaluated using data from 88 cerebrovascular ICU patients, 48 men and 40 women, during their first 48 hours of ICU stay. The electrocardiogram (ECG) signals are related to lead II, and the sampling frequency is 125 Hz. The time of admission and time of death are labeled in all data. In this study, the mortality prediction in patients with cerebral ischemia is evaluated using the features extracted from the return map generated by the signal of HRV and blood pressure. To predict the patient’s future condition, the combination of features extracted from the return mapping generated by the HRV signal, such as angle (α), area (A), and various parameters generated by systolic and diastolic blood pressure, including DBP_Max−Min SBP_SD have been used. Also, to select the best feature combination, the genetic algorithm (GA) and mutual information (MI) methods were used. Paired sample t-test statistical analysis was used to compare the results of two episodes (death and non-death episodes). The P-value for detecting the significance level was considered less than 0.005.

The results indicate that the new approach presented in this paper can be compared with other methods or leads to better results. The best combination of features based on GA to achieve maximum predictive accuracy was m (mean), L_Mean, A, SBP_SVMax, DBP_Max-Min. The accuracy, specificity, and sensitivity based on the best features obtained from GA were 97.7%, 98.9%, and 95.4% for cerebral ischemia disease with a prediction horizon of 0.5–1 hour before death. The d-factor for the best feature combination based on the GA model is less than 1 (d-factor = 0.95). Also, the bracketed by 95 percent prediction uncertainty (95PPU) (%) was obtained at 98.6.

The combination of HRV and blood pressure signals might increase the accuracy of the prediction of the death episode and reduce the minimum hospitalization time of the patient with cerebrovascular diseases to determine the future status.

The increasing prevalence of hepatic steatosis presents a considerable challenge to public health. There is a critical need for the development of novel preventive and screening strategies for this condition. This study evaluated the potential applications of wrist pulse detection technology for the early detection of liver diseases. The pulse time-domain features of a medical exam population with and without hepatic steatosis were assessed to develop a screening model for this disease.

Participants were consecutively recruited from March 2021 to March 2022 in the medical examination centers of the Yueyang Hospital of Integrated Traditional Chinese and Western Medicine and the Shanghai Municipal Hospital of Traditional Chinese Medicine. Clinical data from 255 participants, including general information (sex, age, and body mass index), and data related to glucose and blood lipids (fasting plasma glucose, triglyceride, total cholesterol, high-density lipoprotein, and low-density lipoprotein levels) were collected. Wrist pulse signals were acquired using a pulse detection device, and the pulse time-domain features, including t₁, t₄, t₅, T, w₁, w₂, h₂/h₁, h₃/h₁, and h₅/h₁ were extracted. Participants were assigned to hepatic steatosis and non-hepatic steatosis groups according to their abdominal ultrasound examination results. Their clinical data and pulse time-domain features were compared using chi-square and parametric or non-parametric statistical methods. Three datasets were used to construct screening models for hepatic steatosis based on the random forest algorithm. The datasets for modeling were defined as Dataset 1, containing blood glucose and lipid data and general information; Dataset 2, containing time-domain features and general information; Dataset 3, containing time-domain features, blood glucose and lipid data, and general information. The evaluation metrics, accuracy, precision, recall, F1-score, and areas under the receiver operating characteristic curve (AUC) were compared for each model.

The time-domain features of the two groups differed significantly. The t₁, t₄, t₅, T, h₂/h₁, h₃/h₁, w₁, and w₂ features were higher in the hepatic steatosis group than in the non-hepatic steatosis group (P < 0.05), while the h₅/h₁ features were lower in the hepatic steatosis group than in the non-hepatic steatosis group (P < 0.05). The screening models for hepatic steatosis based on both time-domain features and blood glucose and lipid data outperformed those based on time-domain features or blood markers alone. The accuracy, precision, recall, F1-score, and AUC of the combined model were 81.18%, 80.56%, 76.32%, 79%, and 87.79%, respectively. These proportions were 1.57%, 1.86%, 1.76%, 2%, and 3.54% higher, respectively, than those of the model based on time-domain features alone and 3.14%, 4.2%, 2.64%, 4%, and 6.47% higher, respectively, than those of the model based on blood glucose and lipid alone.

The early screening model for hepatic steatosis using datasets that included pulse time-domain features achieved better performance. The findings suggest that pulse detection technology could be used to inform the development of a mobile medical device or remote home monitoring system to test for hepatitis steatosis.