Stratifying High-Risk Thyroid Nodules Using a Novel Deep Learning System

INTRODUCTION

The current ultrasound scan classification system for thyroid nodules is time-consuming, labor-intensive, and subjective. Artificial intelligence (AI) has been shown to increase the accuracy of predicting the malignancy rate of thyroid nodules. This study aims to demonstrate the state-of-the-art Swin Transformer to classify thyroid nodules.

MATERIALS AND METHODS

Ultrasound images were collected prospectively from patients who received fine needle aspiration biopsy for thyroid nodules from January 2016 to June 2021. One hundred thirty-nine patients with malignant thyroid nodules were enrolled, while 235 patients with benign nodules served as controls. Images were fed to Swin-T and ResNeSt50 models to classify the thyroid nodules.

RESULTS

Patients with malignant nodules were younger and more likely male compared to those with benign nodules. The average sensitivity and specificity of Swin-T were 82.46% and 84.29%, respectively. The average sensitivity and specificity of ResNeSt50 were 72.51% and 77.14%, respectively. Receiver operating characteristics analysis revealed that the area under the curve of Swin-T was higher (AUC=0.91) than that of ResNeSt50 (AUC=0.82). The McNemar test evaluating the performance of these models showed that Swin-T had significantly better performance than ResNeSt50.Swin-T classifier can be a useful tool in helping shared decision-making between physicians and patients with thyroid nodules, particularly in those with high-risk characteristics of sonographic patterns.

Degree of Pelvic Rotation in the Coronal Plane on Postoperative Radiographs Obtained after Total Hip Arthroplasty

There are many published cup anteversion measurements for postoperative total hip arthroplasty (THA), including Liaw’s, Lewinnek’s, and Murray’s methods. However, most measurements ignore the potential pelvic rotation on radiographs except in Liaw’s method. Without considering pelvic rotation, clinicians can miscalculate cup anteversion. Therefore, we aimed to quantify the mean degree of pelvic rotation. Herein, we collected 388 radiographs of 98 postoperative THA hips of 77 patients and measured pelvic rotation by measuring h, the horizontal displacement of the sacrococcygeal junction associated with the upper pole of the symphysis pubis, and ssd, the distance between the sacrococcygeal junction and pubic symphysis. The angle θ of pelvic rotation was defined as θ = arc sin (h/ssd) × (180°/π). The mean degree of pelvic rotation was then calculated. The standard deviation of h was 7.84 mm, and the mean ssd was 158 mm. The potential pelvic rotation was 2.50°. The p-values from the paired t-test were all >0.05 when interobserver and intraobserver errors were assessed. This is the first study to quantify the potential pelvic rotation in the coronal plane on postoperative plain radiographs. The potential pelvic rotation was too large to be neglected during the measurement of cup anteversion.

Cauda equina syndrome caused by the application of DuraSealTM in a microlaminectomy surgery: A case report

BACKGROUND

The management of dural tears is important. While a massive dura can be repaired with absorbable suture lines, cerebrospinal fluid leakage can be attenuated by dural sealant when an unintended tiny durotomy occurs intraoperatively. DuraSeal is often used because it can expand to seal tears. This case emphasizes the need for caution when DuraSeal is used as high expansion can cause complications following microlaminectomy.

CASE SUMMARY

A 77-year-old woman presented with L2/3 and L3/4 lateral recess stenosis. She underwent microlaminectomy, foraminal decompression, and disk height restoration using an IntraSPINE^® device. A tiny incident durotomy occurred intraoperatively and was sealed using DuraSeal^TM. However, decreased muscle power, urinary incontinence, and absence of anal reflexes were observed postoperatively. Emergent magnetic resonance imaging revealed fluid collection causing thecal sac indentation and central canal compression. Surgical exploration revealed that the gel-like DuraSeal had entrapped the hematoma and, consequently, compressed the thecal sac and nerve roots. While we removed all DuraSeal^TM and exposed the nerve root, the patient’s neurological function did not recover postoperatively.

CONCLUSION

DuraSeal expansion must not be underestimated. Changes in neurological status require investigation for cauda equina syndrome due to expansion.

An improvement of current hypercube pooling PCR tests for SARS-CoV-2 detection

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) pandemic can be effectively controlled by rapid and accurate identification of SARS-CoV-2-infected cases through large-scale screening. Hypercube pooling polymerase chain reaction (PCR) is frequently used as a pooling technique because of its high speed and efficiency. We attempted to implement the hypercube pooling strategy and found it had a large quantization effect. This raised two questions: is hypercube pooling with edge = 3 actually the optimal strategy? If not, what is the best edge and dimension? We used a C++ program to calculate the expected number of PCR tests per patient for different values of prevalence, edge, and dimension. The results showed that every edge had a best performance range. Then, using C++ again, we created a program to calculate the optimal edge and dimension required for pooling samples when entering prevalence into our program. Our program will be provided as freeware in the hope that it can help governments fight the SARS-CoV-2 pandemic.

Mathematical model of distal radius orientation

Distal radius orientation is important in evaluating Colles' fracture. In most cases, the wrist was protected by a bandage, splint, or cast. Therefore, it was difficult for the radiology technician to take perfect anteroposterior and lateral view radiographs. In this study, we build a mathematical model and calculate the pronation angle needed to produce dorsal tilt, which is a volar tilt in a perfect lateral view radiograph. The formulas are all incorporated into Excel to facilitate usage.

An Outperforming Artificial Intelligence Model to Identify Referable Blepharoptosis for General Practitioners

The aim of this study is to develop an AI model that accurately identifies referable blepharoptosis automatically and to compare the AI model’s performance to a group of non-ophthalmic physicians. In total, 1000 retrospective single-eye images from tertiary oculoplastic clinics were labeled by three oculoplastic surgeons as having either ptosis, including true and pseudoptosis, or a healthy eyelid. A convolutional neural network (CNN) was trained for binary classification. The same dataset was used in testing three non-ophthalmic physicians. The CNN model achieved a sensitivity of 92% and a specificity of 88%, compared with the non-ophthalmic physician group, which achieved a mean sensitivity of 72% and a mean specificity of 82.67%. The AI model showed better performance than the non-ophthalmic physician group in identifying referable blepharoptosis, including true and pseudoptosis, correctly. Therefore, artificial intelligence-aided tools have the potential to assist in the diagnosis and referral of blepharoptosis for general practitioners.

A deep learning approach to identify blepharoptosis by convolutional neural networks

PURPOSE

Blepharoptosis is a known cause of reversible vision loss. Accurate assessment can be difficult, especially amongst non-specialists. Existing automated techniques disrupt clinical workflow by requiring user input, or placement of reference markers. Neural networks are known to be effective in image classification tasks. We aim to develop an algorithm that can accurately identify blepharoptosis from a clinical photo.

METHODS

A total of 500 clinical photographs from patients with and without blepharoptosis were sourced from a tertiary ophthalmic center in Taiwan. Images were labeled by two oculoplastic surgeons, with an independent third oculoplastic surgeon to adjudicate disagreements. These images were used to train a series of convolutional neural networks (CNNs) to ascertain the best CNN architecture for this particular task.

RESULTS

Of the models that trained on the dataset, most were able to identify ptosis images with reasonable accuracy. We found the best performing model to use the DenseNet121 architecture without pre-training which achieved a sensitivity of 90.1 % with a specificity of 82.4 %, compared to the worst performing model which was used a Resnet34 architecture with pre-training, achieving a sensitivity of 74.1 %, and specificity of 63.6 %. Models with and without pre-training performed similarly (mean accuracy 82.6 % vs. 85.8 % respectively, p = 0.06), though models with pre-training took less time to train (1-minute vs. 16 min, p < 0.01).

CONCLUSIONS

We report the use of AI to accurately diagnose blepharoptosis from a clinical photograph with no external reference markers or user input requirement. Most current-generation CNN architectures performed reasonably on this task, with the DenseNet121, and Resnet18 architectures without pre-training performing best in our dataset.

Survival outcomes of eye-sparing surgery for adenoid cystic carcinoma of lacrimal gland

PURPOSE

To survey adenoid cystic carcinoma of lacrimal glands in Asian population and investigate the predictability in prognosis following the 8^th edition American Joint Committee on Cancer (AJCC) staging guideline.

STUDY DESIGN

Retrospective study.

METHODS

The clinical entities and surgical outcomes of the patients who were histologically confirmed with a diagnosis of lacrimal adenoid cystic carcinoma in National Taiwan University Hospital between January 1995 and December 2015 were retrospectively reviewed.

RESULTS

Enrolled were 11 patients. The median follow-up was 7.2 years. Eight patients (72.7%) were diagnosed as T1 or T2 disease, and three patients (27.3%) were diagnosed as T3 or T4 disease according to the AJCC 8^th edition guideline. Eye-sparing surgery with radiotherapy was performed in nine patients. Local recurrence was noted in six patients (54.5%) with median disease-free interval of 23.5 months. Six patients (54.5%) developed distant metastases, including lung, bone, and cranial invasions. Overall survival rate during the study period was 54.6%. Five-year overall survival was 81.8% and ten-year overall survival was 68.2%. The Log-rank test for overall survival and disease-free survival between patients with less than T3 disease (p=0.001) and patients with T3 or T4 disease (p=0.006) revealed significant differences.

CONCLUSION

This study highlighted the aggressive nature of adenoid cystic carcinoma of lacrimal glands. Eye-sparing surgery with adjunctive radiotherapy may achieve relatively optimal disease control in diseases staged T1 or T2, but in advanced disease metastasis and mortality are usually inevitable.

How to evaluate three dimensional angle error from plain radiographs

Evaluating three-dimensional angle error is necessary because we cannot get every patient's CT or MRI at all times. Creating a method that can calculate angle error from plain radiographs is therefore important. Using vector and trigonometric mathematics, we gradually deduct our formula which can calculate angle error from goal angles (the angles we plan to achieve before operation) to result angles (the angles we get after operation) by two perpendicular radiographs. We also encode it into Micorsoft Excel (Redmond Campus, Redmond, Washington, U.S.) so that it becomes more user-friendly. We hope this tool can be used when evaluating TKR, corrective osteotomy, fracture fixation, and so on.

Link clustering reveals structural characteristics and biological contexts in signed molecular networks

Many biological networks are signed molecular networks which consist of positive and negative links. To reveal the distinct features between links with different signs, we proposed signed link-clustering coefficients that assess the similarity of inter-action profiles between linked molecules. We found that positive links tended to cluster together, while negative links usually behaved like bridges between positive clusters. Positive links with higher adhesiveness tended to share protein domains, be associated with protein-protein interactions and make intra-connections within protein complexes. Negative links that were more bridge-like tended to make interconnections between protein complexes. Utilizing the proposed measures to group positive links, we observed hierarchical modules that could be well characterized by functional annotations or known protein complexes. Our results imply that the proposed sign-specific measures can help reveal the network structural characteristics and the embedded biological contexts of signed links, as well as the functional organization of signed molecular networks.

Computerized ellipse method for measuring acetabular version after total hip replacement--a precision study using synthetic and real radiographs

BACKGROUND

Previous work by our group to address the problem of acetabular positioning based on 2D methods resulted in the development of a measurement method with better precision--Liaw's version. This method may help the early diagnosis of acetabular loosening. In the present study, we hypothesized that our computerized ellipse method could improve the precision of measuring acetabular version.

METHODS

We developed our Elliversion software to measure acetabular version. Using total hip replacement (THR) Simulator, 96 radiographs were synthesized with random femoral inclination and 5° to 52° version, half with the femoral head included and half without. These synthetic radiographs and 28 real radiographs were measured with both Elliversion and the trigonometric method twice by one of the authors with a one-week interval between measurements. We then calculated the difference in the repeated measurements. Student's t-test was used for statistical analysis of the measuring error and inter-measurement difference.

RESULTS

In the precision study, for synthetic radiographs including the femoral head, the ellipse method was significantly better than the trigonometric method (p < 0.01). For synthetic radiographs without the femoral head, there was no significant difference between the ellipse method and the trigonometric method (p = 0.19). As for the repeated measurements, for synthetic radiographs including the femoral head, the ellipse method was significantly better than the trigonometric method (p = 0.001), whereas for synthetic radiographs without the femoral head, there was no significant difference between the two methods (p = 0.17). For real radiographs, there was no significant difference between the two measuring methods (p = 0.12). However, if we excluded the four poor-quality radiographs, there was a significant difference between the two measuring methods (p = 0.04).

DISCUSSION

We developed a computerized ellipse method for measuring acetabular version on synthetic radiographs and good-quality real radiographs. This method is characterized by its superior precision as compared to the trigonometric method. With the 2D standardized method (Liaw's version), improving the precision of measurement will help earlier diagnosis of acetabular loosening.

Near Point Light Sources for Shape from Shading

In this paper, a linear algorithm3,4 is proposed to recover shape information from multiple images, each of them is taken under the environment that all of the object surfaces are illuminated by a known near point light source. In this method, an approximate range of the distance for the objects to the viewer (e.g. camera) is previously defined. Using this predefined value, the absolute depth map of the objects can be found out.

Multiple kernel learning for dimensionality reduction

In solving complex visual learning tasks, adopting multiple descriptors to more precisely characterize the data has been a feasible way for improving performance. The resulting data representations are typically high-dimensional and assume diverse forms. Hence, finding a way of transforming them into a unified space of lower dimension generally facilitates the underlying tasks such as object recognition or clustering. To this end, the proposed approach (termed MKL-DR) generalizes the framework of multiple kernel learning for dimensionality reduction, and distinguishes itself with the following three main contributions: first, our method provides the convenience of using diverse image descriptors to describe useful characteristics of various aspects about the underlying data. Second, it extends a broad set of existing dimensionality reduction techniques to consider multiple kernel learning, and consequently improves their effectiveness. Third, by focusing on the techniques pertaining to dimensionality reduction, the formulation introduces a new class of applications with the multiple kernel learning framework to address not only the supervised learning problems but also the unsupervised and semi-supervised ones.

Coregulation of transcription factors and microRNAs in human transcriptional regulatory network

BACKGROUND

MicroRNAs (miRNAs) are small RNA molecules that regulate gene expression at the post-transcriptional level. Recent studies have suggested that miRNAs and transcription factors are primary metazoan gene regulators; however, the crosstalk between them still remains unclear.

METHODS

We proposed a novel model utilizing functional annotation information to identify significant coregulation between transcriptional and post-transcriptional layers. Based on this model, function-enriched coregulation relationships were discovered and combined into different kinds of functional coregulation networks.

RESULTS

We found that miRNAs may engage in a wider diversity of biological processes by coordinating with transcription factors, and this kind of cross-layer coregulation may have higher specificity than intra-layer coregulation. In addition, the coregulation networks reveal several types of network motifs, including feed-forward loops and massive upstream crosstalk. Finally, the expression patterns of these coregulation pairs in normal and tumour tissues were analyzed. Different coregulation types show unique expression correlation trends. More importantly, the disruption of coregulation may be associated with cancers.

CONCLUSION

Our findings elucidate the combinatorial and cooperative properties of transcription factors and miRNAs regulation, and we proposes that the coordinated regulation may play an important role in many biological processes.

A health examination system integrated with clinical decision support system

Health examinations play a key role in preventive medicine. We propose a health examination system named Health Examination Automatic Logic System (HEALS) to assist clinical workers in improving the total quality of health examinations. Quality of automated inference is confirmed by the zero inference error where during 6 months and 14,773 cases. Automated inference time is less than one second per case in contrast to 2 to 5 min for physicians. The most significant result of efficiency evaluation is that 3,494 of 4,356 (80.2%) cases take less than 3 min per case for producing a report summary. In the evaluation of effectiveness, novice physicians got 18% improvement in making decisions with the assistance of our system. We conclude that a health examination system with a clinical decision system can greatly reduce the mundane burden on clinical workers and markedly improve the quality and efficiency of health examination tasks.

Measurement of the acetabular cup anteversion on simulated radiographs

Widmer (J Arthroplasty 2004;19:387) reported a protractor for measuring the anteversion of acetabular cups on radiographs but with limited precision. We intended to improve its precision by trigonometric mathematics. We measured the anteversion of the acetabular cups on 336 simulated radiographs using aforementioned 2 methods. The anteversion measured by Widmer's protractor ranged from 7 degrees to 41 degrees (mean +/- SD = 28.0 degrees +/- 9.8 degrees), and our methods, 5 degrees to 51 degrees (27.7 degrees +/- 13.2 degrees). The mean +/- SD of error by Widmer's protractor was 5.2 +/- 2.5 degrees, and our protractor, 0.8 degrees +/- 0.8 degrees (Student t test, P b .0001). The interobserver study showed the difference between measurements less than 2 degrees for each method. Therefore, the smaller error of our method than that of Widmer implicated a potentially precise measurement of the anteversion (level of evidence: diagnostic study, level II).

THR Simulator--the software for generating radiographs of THR prosthesis

Background

Measuring the orientation of acetabular cup after total hip arthroplasty is important for prognosis. The verification of these measurement methods will be easier and more feasible if we can synthesize prosthesis radiographs in each simulated condition. One reported method used an expensive mechanical device with an indeterminable precision. We thus develop a program, THR Simulator, to directly synthesize digital radiographs of prostheses for further analysis.

Under Windows platform and using Borland C++ Builder programming tool, we developed the THR Simulator. We first built a mathematical model of acetabulum and femoral head. The data of the real dimension of prosthesis was adopted to generate the radiograph of hip prosthesis. Then with the ray tracing algorithm, we calculated the thickness each X-ray beam passed, and then transformed to grey scale by mapping function which was derived by fitting the exponential function from the phantom image. Finally we could generate a simulated radiograph for further analysis.

Results

Using THR Simulator, the users can incorporate many parameters together for radiograph synthesis. These parameters include thickness, film size, tube distance, film distance, anteversion, abduction, upper wear, medial wear, and posterior wear. These parameters are adequate for any radiographic measurement research. This THR Simulator has been used in two studies, and the errors are within 2° for anteversion and 0.2 mm for wearing measurement.

Conclusion

We design a program, THR Simulator that can synthesize prosthesis radiographs. Such a program can be applied in future studies for further analysis and validation of measurement of various parameters of pelvis after total hip arthroplasty.

A simplified guide ruler from numeric table method in doing rotational osteotomy

Background

Čobeljić et al. recently reported a numeric table method to provide precise rotational osteotomy which is a well established orthopaedic procedure. The numeric table requires four pages in length that is rather inconvenient during performing an osteotomy operation.

Methods

We thus develop our own method by summarizing the data of the four-page table into a small ruler, which is easy to carry and use in operation room. An electrical version of this ruler is also available. We also build a computer model to verify Čobeljić et al. method.

Results

The error of Čobeljić et al. is between -37% to 16% (mean ± SD = -6% ± 9%). We verify our ruler by calculating the absolute difference between our method and that of Čobeljić et al. The difference is less than 0.1 mm.

Conclusion

Our ruler is convenient for practical use for the rotational osteotomy procedure with equal precision. Further clinical verification is needed to justify its real significance.

A Simple Mathematical Standardized Measurement of Acetabulum Anteversion after Total Hip Arthroplasty

We invented a standardization method to measure the cup's anteversion after total hip arthroplasty without the influence of patient's position. We measured 68 radiographs of 10 patients after total hip replacement (THR) and calculated the error of each measurement, defined as the difference with the average of the same measuring method on the same patient. We also calculated the repeatability standard deviation (RSD) of each method according to the American Society for Testing and Materials, ASTM E691.

Mean absolute inter-examination angle error, defined as the average of the absolute deviations, was 0.75° for standardized anteversion (range 0.03–2.51°), as compared with those without standardization, 2.30° (range 0.04–13.04°). The inter-examination measurement reliability (precision), defined as one RSD, was 0.99° for standardized anteversion, as compared with those without standarization, 3.50°. There is no difference between patients four and five without (p = 0.097). There is a significant difference with standardization (p < 0.0001).

Our study demonstrated that this mathematical method is a precise tool to measure the anteversion of the acetabular cup. We hope that it can be used widely in the future.

A new tool for measuring cup orientation in total hip arthroplasties from plain radiographs

Orientation of the hip cup is important in total hip arthroplasties. Orientation includes abduction (inclination) and anteversion. Anteversion can be considered as true (anatomic) and planar (radiographic) anteversion. Some measurement methods either are too complicated or are less precise. We developed a new protractor to measure cup orientation using postoperative anteroposterior radiographs centered at the hip. The new protractor measures true and planar anteversion and abduction easily and precisely. We verified its accuracy using a software simulator and simulated 45 radio- graphs of total hip arthroplasties with 15 different anteversions ranging from 15 degrees -29 degrees and 45 actual radiographs of total hip arthroplasties. We then measured the planar ante- version with our method and the method of Lewinnek et al. Maximal errors were 3 degrees and 2.61 degrees , respectively, and mean errors were 0.96 degrees and 1.2 degrees , respectively. The standard deviations were 0.74 degrees with our method and 0.57 degrees with the method of Lewinnek et al. For the real radiographs, the mean of absolute difference between the two methods was 1.34 degrees , and the standard deviation was 1.13 degrees . We found no difference between the two methods and no difference in our findings compared with those of Pradhan.

Fast block matching algorithm based on the winner-update strategy

Block matching is a widely used method for stereo vision, visual tracking, and video compression. Many fast algorithms for block matching have been proposed in the past, but most of them do not guarantee that the match found is the globally optimal match in a search range. This paper presents a new fast algorithm based on the winner-update strategy which utilizes an ascending lower bound list of the matching error to determine the temporary winner. Two lower bound lists derived by using partial distance and by using Minkowski's inequality are described. The basic idea of the winner-update strategy is to avoid, at each search position, the costly computation of the matching error when there exists a lower bound larger than the global minimum matching error. The proposed algorithm can significantly speed up the computation of the block matching because: 1) computational cost of the lower bound we use is less than that of the matching error itself; 2) an element in the ascending lower bound list will be calculated only when its preceding element has already been smaller than the minimum matching error computed so far; 3) for many search positions, only the first several lower bounds in the list need to be calculated. Our experiments have shown that, when applying to motion vector estimation for several widely-used test videos, 92% to 98% of operations can be saved while still guaranteeing the global optimality. Moreover, the proposed algorithm can be easily modified either to meet the limited time requirement or to provide an ordered list of best candidate matches. Our source codes of the proposed algorithm are available at http://smart.iis.sinica.edu.tw/html/winup.html.

Hierarchical color image region segmentation for content-based image retrieval system

In this work, we propose a model of a content-based image retrieval system by using the new idea of combining a color segmentation with relationship trees and a corresponding tree-matching method. We retain the hierarchical relationship of the regions in an image during segmentation. Using the information of the relationships and features of the regions, we can represent the desired objects in images more accurately. In retrieval, we compare not only region features but also region relationships.

Displacement field estimation and image segmentation using block matching enhanced by a neural network

The block-matching method plays an important role in displacement field estimation due to its simplicity, achievement of long-range motion, and robustness to noise. In this paper, a single-layer feedback neural network model is proposed that enhances block matching, estimates the displacement field, and simultaneously performs image segmentation from consecutive images. In this paper, image segmentation is defined as partitioning each image into a set of moving objects and the background. For any two consecutive images, a neural network is created that learns the connection relationship of the pixels in an object from the displacement field and stores the relationship in the network. A modified block matching is used to compute a more accurate displacement field by utilizing the segmentation information embedded in the neural network. The displacement vector at the edge of an object or occluding boundary is hard to estimate, but the proposed model performs satisfactorily because it learns and uses the connection information. Furthermore, a flood-fill algorithm is used to compute the dense displacement field more efficiently and correctly than the exhaustive search does. The most important aspect of this paper is that image segmentation is performed simultaneously with the displacement-field estimation by the neural-network model. The novel idea of the work is to embed the segmentation information (connection relations) in the neural network and to perform the displacement-field estimation and image segmentation simultaneously. Two methods for retrieving segmentation information from the neural network with any two consecutive images are also presented.