Automatically Diagnosing Skull Fractures Using an Object Detection Method and Deep Learning Algorithm in Plain Radiography Images

Objective@#: Deep learning is a machine learning approach based on artificial neural network training, and object detection algorithm using deep learning is used as the most powerful tool in image analysis. We analyzed and evaluated the diagnostic performance of a deep learning algorithm to identify skull fractures in plain radiographic images and investigated its clinical applicability. @*Methods@#: A total of 2026 plain radiographic images of the skull (fracture, 991; normal, 1035) were obtained from 741 patients. The RetinaNet architecture was used as a deep learning model. Precision, recall, and average precision were measured to evaluate the deep learning algorithm’s diagnostic performance. @*Results@#: In ResNet-152, the average precision for intersection over union (IOU) 0.1, 0.3, and 0.5, were 0.7240, 0.6698, and 0.3687, respectively. When the intersection over union (IOU) and confidence threshold were 0.1, the precision was 0.7292, and the recall was 0.7650. When the IOU threshold was 0.1, and the confidence threshold was 0.6, the true and false rates were 82.9% and 17.1%, respectively. There were significant differences in the true/false and false-positive/false-negative ratios between the anteriorposterior, towne, and both lateral views (p=0.032 and p=0.003). Objects detected in false positives had vascular grooves and suture lines. In false negatives, the detection performance of the diastatic fractures, fractures crossing the suture line, and fractures around the vascular grooves and orbit was poor. @*Conclusion@#: The object detection algorithm applied with deep learning is expected to be a valuable tool in diagnosing skull fractures.

Development of an Automatic Pill Image Data Generation System / 대한의료정보학회지

Objectives@#Since the easiest way to identify pills and obtain information about them is to distinguish them visually, many studies on image processing technology exist. However, no automatic system for generating pill image data has yet been developed. Therefore, we propose a system for automatically generating image data by taking pictures of pills from various angles. This system is referred to as the pill filming system in this paper. @*Methods@#We designed the pill filming system to have three components: structure, controller, and a graphical user interface (GUI). This system was manufactured with black polylactic acid using a 3D printer for lightweight and easy manufacturing. The mainboard controls data storage, and the entire process is managed through the GUI. After one reciprocating movement of the seesaw, the web camera at the top shoots the target pill on the stage. This image is then saved in a specific directory on the mainboard. @*Results@#The pill filming system completes its workflow after generating 300 pill images. The total time to collect data per pill takes 21 minutes and 25 seconds. The generated image size is 1280 × 960 pixels, the horizontal and vertical resolutions are both 96 DPI (dot per inch), and the file extension is .jpg. @*Conclusions@#This paper proposes a system that can automatically generate pill image data from various angles. The pill observation data from various angles include many cases. In addition, the data collected in the same controlled environment have a uniform background, making it easy to process the images. Large quantities of high-quality data from the pill filming system can contribute to various studies using pill images.

Detection and Weak Segmentation of Masses in Gray-Scale Breast Mammogram Images Using Deep Learning

Purpose@#In this paper, we propose deep-learning methodology with which to enhance the mass differentiation performance of convolutional neural network (CNN)-based architecture. @*Materials and Methods@#We differentiated breast mass lesions from gray-scale X-ray mammography images based on regions of interest (ROIs). Our dataset comprised breast mammogram images for 150 cases of malignant masses from which we extracted the mass ROI, and we composed a CNN-based deep learning model trained on this dataset to identify ROI mass lesions. The test dataset was created by shifting some of the training data images. Thus, although both datasets were different, they retained a deep structural similarity. We then applied our trained deep-learning model to detect masses on 8-bit mammogram images containing malignant masses. The input images were preprocessed by applying a scaling parameter of intensity before being used to train the CNN model for mass differentiation. @*Results@#The highest area under the receiver operating characteristic curve was 0.897 (Î 20). @*Conclusion@#Our results indicated that the proposed patch-wise detection method can be utilized as a mass detection and segmentation tool.

Video Archiving and Communication System (VACS): A Progressive Approach, Design, Implementation, and Benefits for Surgical Videos / 대한의료정보학회지

Objectives@#As endoscopic, laparoscopic, and robotic surgical procedures become more common, surgical videos are increasingly being treated as records and serving as important data sources for education, research, and developing new solutions with recent advances in artificial intelligence (AI). However, most hospitals do not have a system that can store and manage such videos systematically. This study aimed to develop a system to help doctors manage surgical videos and turn them into content and data. @*Methods@#We developed a video archiving and communication system (VACS) to systematically process surgical videos. The VACS consists of a video capture device called SurgBox and a video archiving system called SurgStory. SurgBox automatically transfers surgical videos recorded in the operating room to SurgStory. SurgStory then analyzes the surgical videos and indexes important sections or video frames to provide AI reports. It allows doctors to annotate classified indexing frames, “data-ize” surgical information, create educational content, and communicate with team members. @*Results@#The VACS collects surgical and procedural videos, and helps users manage archived videos. The accuracy of a convolutional neural network learning model trained to detect the top five surgical instruments reached 96%. @*Conclusions@#With the advent of the VACS, the informational value of medical videos has increased. It is possible to improve the efficiency of doctors’ continuing education by making video-based online learning more active and supporting research using data from medical videos. The VACS is expected to promote the development of new AI-based products and services in surgical and procedural fields.

Video Archiving and Communication System (VACS): A Progressive Approach, Design, Implementation, and Benefits for Surgical Videos / 대한의료정보학회지

Objectives@#As endoscopic, laparoscopic, and robotic surgical procedures become more common, surgical videos are increasingly being treated as records and serving as important data sources for education, research, and developing new solutions with recent advances in artificial intelligence (AI). However, most hospitals do not have a system that can store and manage such videos systematically. This study aimed to develop a system to help doctors manage surgical videos and turn them into content and data. @*Methods@#We developed a video archiving and communication system (VACS) to systematically process surgical videos. The VACS consists of a video capture device called SurgBox and a video archiving system called SurgStory. SurgBox automatically transfers surgical videos recorded in the operating room to SurgStory. SurgStory then analyzes the surgical videos and indexes important sections or video frames to provide AI reports. It allows doctors to annotate classified indexing frames, “data-ize” surgical information, create educational content, and communicate with team members. @*Results@#The VACS collects surgical and procedural videos, and helps users manage archived videos. The accuracy of a convolutional neural network learning model trained to detect the top five surgical instruments reached 96%. @*Conclusions@#With the advent of the VACS, the informational value of medical videos has increased. It is possible to improve the efficiency of doctors’ continuing education by making video-based online learning more active and supporting research using data from medical videos. The VACS is expected to promote the development of new AI-based products and services in surgical and procedural fields.

Web-Based Spine Segmentation Using Deep Learning in Computed Tomography Images / 대한의료정보학회지

Application of machine learning to ultrasound images to differentiate follicular neoplasms of the thyroid gland

Purpose@#This study was conducted to evaluate the diagnostic performance of machine learning in differentiating follicular adenoma from carcinoma using preoperative ultrasonography (US). @*Methods@#In this retrospective study, preoperative US images of 348 nodules from 340 patients were collected from two tertiary referral hospitals. Two experienced radiologists independently reviewed each image and categorized the nodules according to the 2015 American Thyroid Association guideline. Categorization of a nodule as highly suspicious was considered a positive diagnosis for malignancy. The nodules were manually segmented, and 96 radiomic features were extracted from each region of interest. Ten significant features were selected and used as final input variables in our in-house developed classifier models based on an artificial neural network (ANN) and support vector machine (SVM). The diagnostic performance of radiologists and both classifier models was calculated and compared. @*Results@#In total, 252 nodules from 245 patients were confirmed as follicular adenoma and 96 nodules from 95 patients were diagnosed as follicular carcinoma. As measures of diagnostic performance, the average sensitivity, specificity, and accuracy of the two experienced radiologists in discriminating follicular adenoma from carcinoma on preoperative US images were 24.0%, 84.0%, and 64.8%, respectively. The sensitivity, specificity, and accuracy of the ANN and SVM-based models were 32.3%, 90.1%, and 74.1% and 41.7%, 79.4%, and 69.0%, respectively. The kappa value of the two radiologists was 0.076, corresponding to slight agreement. @*Conclusion@#Machine learning-based classifier models may aid in discriminating follicular adenoma from carcinoma using preoperative US.

Distribution and Characteristics of Pancreatic Volume Using Computed Tomography Volumetry / 대한의료정보학회지

Objectives@#Changes in the pancreatic volume (PV) are useful as potential clinical markers for some pancreatic-related diseases. The objective of this study was to measure the volume of the pancreas using computed tomography (CT) volumetry and to evaluate the relationships between sex, age, body mass index (BMI), and sarcopenia. @*Methods@#We retrospectively analyzed the abdominal CT scans of 1,003 subjects whose ages ranged between 10 and 90 years. The pancreas was segmented manually to define the region of interest (ROI) based on CT images, and then the PVs were measured by counting the voxels in all ROIs within the pancreas boundary. Sarcopenia was identified by examination of CT images that determined the crosssectional area of the skeletal muscle around the third lumbar vertebra. @*Results@#The mean volume of the pancreas was 62.648 ± 19.094 cm3. The results indicated a negative correlation between the PV and age. There was a positive correlation between the PV and BMI for both sexes, females, and males (r = 0.343, p < 0.001; r = 0.461, p < 0.001; and r = 0.244, p < 0.001, respectively). Additionally, there was a positive correlation between the PV and sarcopenia for females (r = 0.253, p < 0.001) and males (r = 0.200, p < 0.001). @*Conclusions@#CT pancreas volumetry results may help physicians follow up or predict conditions of the pancreas after interventions for pancreatic-related disease in the future.

Design of Automatic Hand Sanitizer System Compatible with Various Containers / 대한의료정보학회지

Objectives@#Demand for hand sanitizers has surged since the coronavirus broke out and spread around the world. Hand sanitizers are usually applied by squirting the sanitizer liquid when one presses a pump with one’s hand. This causes many people to come into contact with the pump handle, which increases the risk of viral transmission. Some hand sanitizers on the market are automatically pumped. However, because sanitizer containers and pump devices are designed to be compatible only between products produced by the same manufacturer, consumers must also repurchase the container for the liquid if they replace the hand sanitizer. Therefore, this paper suggests the design of an automatic hand sanitizer system compatible with various sanitizer containers. @*Methods@#An automatic hand sanitizer system was designed, which will be presented in two stages describing the instrument structure and control parts. This work focused on using the elasticity of pumps and improving people’s access to devices. @*Results@#We have designed an automatic hand sanitizer system that is compatible with various containers. When one moves one’s hand close to the device sensor, the hand sanitizer container is pumped once. @*Conclusions@#The automatic hand sanitizer device proposed in this paper is ultimately expected to contribute to contactless hand disinfection in public places and virus infection prevention. Additionally, it is economical and eco-friendly by decreasing waste emissions.

Deep Learning in Thyroid Ultrasonography to Predict Tumor Recurrence in Thyroid Cancers

Purpose@#To evaluate a deep learning model to predict recurrence of thyroid tumor using preoperative ultrasonography (US). @*Materials and Methods@#We included representative images from 229 US-based patients (male:female = 42:187; mean age, 49.6 years) who had been diagnosed with thyroid cancer on preoperative US and subsequently underwent thyroid surgery. After selecting each representative transverse or longitudinal US image, we created a data set from the resulting database of 898 images after augmentation. The Python 2.7.6 and Keras 2.1.5 framework for neural networks were used for deep learning with a convolutional neural network. We compared the clinical and histological features between patients with and without recurrence. The predictive performance of the deep learning model between groups was evaluated using receiver operating characteristic (ROC) analysis, and the area under the ROC curve served as a summary of the prognostic performance of the deep learning model to predict recurrent thyroid cancer. @*Results@#Tumor recurrence was noted in 49 (21.4%) among the 229 patients. Tumor size and multifocality varied significantly between the groups with and without recurrence (p < 0.05). The overall mean area under the curve (AUC) value of the deep learning model for prediction of recurrent thyroid cancer was 0.9 ± 0.06. The mean AUC value was 0.87 ± 0.03 in macrocarcinoma and 0.79 ± 0.16 in microcarcinoma. @*Conclusion@#A deep learning model for analysis of US images of thyroid cancer showed the possibility of predicting recurrence of thyroid cancer.

Development of an Optimized Deep Learning Model for Medical Imaging

Deep learning has recently become one of the most actively researched technologies in the field of medical imaging. The availability of sufficient data and the latest advances in algorithms are important factors that influence the development of deep learning models. However, several other factors should be considered in developing an optimal generalized deep learning model. All the steps, including data collection, labeling, and pre-processing and model training, validation, and complexity can affect the performance of deep learning models. Therefore, appropriate optimization methods should be considered for each step during the development of a deep learning model. In this review, we discuss the important factors to be considered for the optimal development of deep learning models.

Development of an Automated Image Analyzer for Microvessel Density Measurement in Bone Marrow Biopsies

Angiogenesis is important for the proliferation and survival of multiple myeloma (MM) cells. Bone marrow (BM) microvessel density (MVD) is a useful marker of angiogenesis and an increase in MVD can be used as a marker of poor prognosis in MM patients. We developed an automated image analyzer to assess MVD from images of BM biopsies stained with anti-CD34 antibodies using two color models. MVD was calculated by merging images from the red and hue channels after eliminating non-microvessels. The analyzer results were compared with those obtained by two experienced hematopathologists in a blinded manner using the 84 BM samples of MM patients. Manual assessment of the MVD by two hematopathologists yielded mean±SD values of 19.4±11.8 and 20.0±11.8. The analyzer generated a mean±SD of 19.5±11.2. The intraclass correlation coefficient (ICC) and Bland-Altman plot of the MVD results demonstrated very good agreement between the automated image analyzer and both hematopathologists (ICC=0.893 [0.840–0.929] and ICC=0.906 [0.859–0.938]). This automated analyzer can provide time- and labor-saving benefits with more objective results in hematology laboratories.

Development of an Optimized Deep Learning Model for Medical Imaging

Deep learning has recently become one of the most actively researched technologies in the field of medical imaging. The availability of sufficient data and the latest advances in algorithms are important factors that influence the development of deep learning models. However, several other factors should be considered in developing an optimal generalized deep learning model. All the steps, including data collection, labeling, and pre-processing and model training, validation, and complexity can affect the performance of deep learning models. Therefore, appropriate optimization methods should be considered for each step during the development of a deep learning model. In this review, we discuss the important factors to be considered for the optimal development of deep learning models.

Applying Deep Learning in Medical Images: The Case of Bone Age Estimation / 대한의료정보학회지

OBJECTIVES: A diagnostic need often arises to estimate bone age from X-ray images of the hand of a subject during the growth period. Together with measured physical height, such information may be used as indicators for the height growth prognosis of the subject. We present a way to apply the deep learning technique to medical image analysis using hand bone age estimation as an example. METHODS: Age estimation was formulated as a regression problem with hand X-ray images as input and estimated age as output. A set of hand X-ray images was used to form a training set with which a regression model was trained. An image preprocessing procedure is described which reduces image variations across data instances that are unrelated to age-wise variation. The use of Caffe, a deep learning tool is demonstrated. A rather simple deep learning network was adopted and trained for tutorial purpose. RESULTS: A test set distinct from the training set was formed to assess the validity of the approach. The measured mean absolute difference value was 18.9 months, and the concordance correlation coefficient was 0.78. CONCLUSIONS: It is shown that the proposed deep learning-based neural network can be used to estimate a subject's age from hand X-ray images, which eliminates the need for tedious atlas look-ups in clinical environments and should improve the time and cost efficiency of the estimation process.

A Novel Implantable Cerebrospinal Fluid Reservoir : A Pilot Study

OBJECTIVE: The purpose of this pilot study was to examine the safety and function of the newly developed cerebrospinal fluid (CSF) reservoir called the V-Port. METHODS: The newly developed V-Port consists of a non-collapsible reservoir outlined with a titanium cage and a connector for the ventricular catheter to be assembled. It is designed to be better palpated and more durable to multiple punctures than the Ommaya reservoir. A total of nine patients diagnosed with leptomeningeal carcinomatosis were selected for V-Port insertion. Each patient was followed up for evaluation for a month after the operation. RESULTS: The average operation time for V-Port insertion was 42 minutes and the average incision size was 6.6 cm. The surgical technique of V-Port insertion was found to be intuitive by all neurosurgeons who participated in the pilot study. There was no obstruction or leakage of the V-Port during intrathecal chemotherapy or CSF drainage. Also, there were no complications including post-operative intracerebral hemorrhage, infection and skin problems related to the V-Port. CONCLUSION: V-Port is a safe and an easy to use implantable CSF reservoir that addresses problems of other implantable CSF reservoirs. Further multicenter clinical trial is needed to prove the safety and the function of the V-Port.

A Fast 3-Dimensional Magnetic Resonance Imaging Reconstruction for Surgical Planning of Uterine Myomectomy

BACKGROUND: Uterine myoma is the most common benign gynecologic tumor in reproductive-aged women. During myomectomy for women who want to preserve fertility, it is advisable to detect and remove all myomas to decrease the risk of additional surgery. However, finding myomas during surgery is often challenging, especially for deep-seated myomas. Therefore, three-dimensional (3D) preoperative localization of myomas can be helpful for the surgical planning for myomectomy. However, the previously reported manual 3D segmenting method takes too much time and effort for clinical use. The objective of this study was to propose a new method of rapid 3D visualization of uterine myoma using a uterine template. METHODS: Magnetic resonance images were listed according to the slide spacing on each plane of the multiplanar reconstruction, and images that were determined to be myomas were selected by simply scrolling the mouse down. By using the selected images, a 3D grid with a slide spacing interval was constructed and filled on its plane and finally registered to a uterine template. RESULTS: The location of multiple myomas in the uterus was visualized in 3D and this proposed method is over 95% faster than the existing manual-segmentation method. Not only the size and location of the myomas, but also the shortest distance between the uterine surface and the myomas, can be calculated. This technique also enables the surgeon to know the number of total, removed, and remaining myomas on the 3D image. CONCLUSION: This proposed 3D reconstruction method with a uterine template enables faster 3D visualization of myomas.

Feasibility of a New Desktop Motion Analysis System with a Video Game Console for Assessing Various Three-Dimensional Wrist Motions

BACKGROUND: The restriction of wrist motion results in limited hand function, and the evaluation of the range of wrist motion is related to the evaluation of wrist function. To analyze and compare the wrist motion during four selected tasks, we developed a new desktop motion analysis system using the motion controller for a home video game console. METHODS: Eighteen healthy, right-handed subjects performed 15 trials of selective tasks (dart throwing, hammering, circumduction, and winding thread on a reel) with both wrists. The signals of light-emitting diode markers attached to the hand and forearm were detected by the optic receptor in the motion controller. We compared the results between both wrists and between motions with similar motion paths. RESULTS: The parameters (range of motion, offset, coupling, and orientations of the oblique plane) for wrist motion were not significantly different between both wrists, except for radioulnar deviation for hammering and the orientation for thread winding. In each wrist, the ranges for hammering were larger than those for dart throwing. The offsets and the orientations of the oblique plane were not significantly different between circumduction and thread winding. CONCLUSIONS: The results for the parameters of dart throwing, hammering, and circumduction of our motion analysis system using the motion controller were considerably similar to those of the previous studies with three-dimensional reconstruction with computed tomography, electrogoniometer, and motion capture system. Therefore, our system may be a cost-effective and simple method for wrist motion analysis.

A Novel Implantable Cerebrospinal Fluid Reservoir : A Pilot Study

OBJECTIVE: The purpose of this pilot study was to examine the safety and function of the newly developed cerebrospinal fluid (CSF) reservoir called the V-Port.METHODS: The newly developed V-Port consists of a non-collapsible reservoir outlined with a titanium cage and a connector for the ventricular catheter to be assembled. It is designed to be better palpated and more durable to multiple punctures than the Ommaya reservoir. A total of nine patients diagnosed with leptomeningeal carcinomatosis were selected for V-Port insertion. Each patient was followed up for evaluation for a month after the operation.RESULTS: The average operation time for V-Port insertion was 42 minutes and the average incision size was 6.6 cm. The surgical technique of V-Port insertion was found to be intuitive by all neurosurgeons who participated in the pilot study. There was no obstruction or leakage of the V-Port during intrathecal chemotherapy or CSF drainage. Also, there were no complications including post-operative intracerebral hemorrhage, infection and skin problems related to the V-Port.CONCLUSION: V-Port is a safe and an easy to use implantable CSF reservoir that addresses problems of other implantable CSF reservoirs. Further multicenter clinical trial is needed to prove the safety and the function of the V-Port.

Efficacy of JOINS on Cartilage Protection in Knee Osteoarthritis: Prospective Randomized Controlled Trial

PURPOSE: In this study, we compared the clinical efficacy of JOINS (SKI306X, SK Chemicals) with placebo on cartilage protection using magnetic resonance imaging (MRI). MATERIALS AND METHODS: Sixty-nine patients were randomized to the JOINS group (200 mg, three times daily for 1 year; n=33) or the placebo group (n=36). Changes in cartilage volume and thickness were measured using MRI. Changes in the delayed gadolinium-enhanced MRI of cartilage (dGEMRIC) index, subchondral bone marrow abnormality scores, and clinical scores including knee pain visual analog scale (VAS) score and Korean Western Ontario and McMaster Universities Osteoarthritis Index (K-WOMAC) were also evaluated. RESULTS: Changes in cartilage thickness and volume and subarticular bone marrow abnormality scores were not different between groups. Changes in the dGEMRIC index in the lateral tibial plateau were greater in the JOINS group than in the placebo group (19.64±114.33 msec vs. −57.77±123.30 msec; p=0.011). Significantly greater changes in VAS were observed in the JOINS group than in the placebo group (−26.00±12.25 vs. −12.47±21.54; p=0.002) and K-WOMAC (−15.42 ± 7.73 vs. −8.15±13.71; p=0.003). CONCLUSIONS: Compared with placebo, JOINS had superior clinical efficacy in regard to cartilage protection.

Computed tomography-based quantitative assessment of lower extremity lymphedema following treatment for gynecologic cancer / 부인종양

OBJECTIVE: To develop an algorithmic quantitative skin and subcutaneous tissue volume measurement protocol for lower extremity lymphedema (LEL) patients using computed tomography (CT), to verify the usefulness of the measurement techniques in LEL patients, and to observe the structural characteristics of subcutaneous tissue according to the progression of LEL in gynecologic cancer. METHODS: A program for algorithmic quantitative analysis of lower extremity CT scans has been developed to measure the skin and subcutaneous volume, muscle compartment volume, and the extent of the peculiar trabecular area with a honeycombed pattern. The CT venographies of 50 lower extremities from 25 subjects were reviewed in two groups (acute and chronic lymphedema). RESULTS: A significant increase in the total volume, subcutaneous volume, and extent of peculiar trabecular area with a honeycombed pattern except quantitative muscle volume was identified in the more-affected limb. The correlation of CT-based total volume and subcutaneous volume measurements with volumetry measurement was strong (correlation coefficient: 0.747 and 0.749, respectively). The larger extent of peculiar trabecular area with a honeycombed pattern in the subcutaneous tissue was identified in the more-affected limb of chronic lymphedema group. CONCLUSION: CT-based quantitative assessments could provide objective volume measurements and information about the structural characteristics of subcutaneous tissue in women with LEL following treatment for gynecologic cancer.