Deep learning for sex determination: Analyzing over 200,000 panoramic radiographs.

The objective of this study is to assess the performance of an innovative AI-powered tool for sex determination using panoramic radiographs (PR) and to explore factors affecting the performance of the convolutional neural network (CNN). The study involved 207,946 panoramic dental X-rays and their corresponding reports from 15 clinical centers in São Paulo, Brazil. The PRs were acquired with four different devices, and 58% of the patients were female. Data preprocessing included anonymizing the exams, extracting pertinent information from the reports, such as sex, age, type of dentition, and number of missing teeth, and organizing the data into a PostgreSQL database. Two neural network architectures, a standard CNN and a ResNet, were utilized for sex classification, with both undergoing hyperparameter tuning and cross-validation to ensure optimal performance. The CNN model achieved 95.02% accuracy in sex estimation, with image resolution being a significant influencing factor. The ResNet model attained over 86% accuracy in subjects older than 6 years and over 96% in those over 16 years. The algorithm performed better on female images, and the area under the curve (AUC) exceeded 96% for most age groups, except the youngest. Accuracy values were also assessed for different dentition types (deciduous, mixed, and permanent) and missing teeth. This study demonstrates the effectiveness of an AI-driven tool for sex determination using PR and emphasizes the role of image resolution, age, and sex in determining the algorithm's performance.

Contrast phase recognition in liver computer tomography using deep learning.

Hepatocellular carcinoma (HCC) has become the 4th leading cause of cancer-related deaths, with high social, economical and health implications. Imaging techniques such as multiphase computed tomography (CT) have been successfully used for diagnosis of liver tumors such as HCC in a feasible and accurate way and its interpretation relies mainly on comparing the appearance of the lesions in the different contrast phases of the exam. Recently, some researchers have been dedicated to the development of tools based on machine learning (ML) algorithms, especially by deep learning techniques, to improve the diagnosis of liver lesions in imaging exams. However, the lack of standardization in the naming of the CT contrast phases in the DICOM metadata is a problem for real-life deployment of machine learning tools. Therefore, it is important to correctly identify the exam phase based only on the image and not on the exam metadata, which is unreliable. Motivated by this problem, we successfully created an annotation platform and implemented a convolutional neural network (CNN) to automatically identify the CT scan phases in the HCFMUSP database in the city of São Paulo, Brazil. We improved this algorithm with hyperparameter tuning and evaluated it with cross validation methods. Comparing its predictions with the radiologists annotation, it achieved an accuracy of 94.6%, 98% and 100% in the testing dataset for the slice, volume and exam evaluation, respectively.

THREE-DIMENSIONAL PRINTING IN ORTHOPEDICS: WHERE WE STAND AND WHERE WE ARE HEADING.

Three-dimensional printing is a technology in expansion in the medical field. It also presents many applications in orthopedics. Our review article aims to describe 3D printing, types of 3D printers, and its use in the orthopedic field. 3D models can be created using tomography scans. Those models can then be manipulated, even simulating surgeries. It is possible to print biomodels, which will help us understand deformities and plan surgeries. Orthopedic surgeons must be updated in these disruptive technologies that may help their daily practice. Level of Evidence V, Expert opinion.

A impressão 3D é uma tecnologia em expansão na medicina, possuindo diversas utilidades na ortopedia. O objetivo deste artigo de revisão é descrever o que é a impressão 3D, seus tipos e suas aplicações na ortopedia. Modelos em 3 dimensões podem ser criados a partir da tomografia computadorizada. Estes modelos podem ser manipulados em softwares específicos, onde inclusive cirurgias podem ser simuladas. Utilizando impressoras 3D podemos criar biomodelos que nos ajudam a compreender deformidades e planejar cirurgias. É importante que o ortopedista se mantenha atualizado nestas novas tecnologias disruptivas que podem auxiliar muito no seu dia a dia. Nível de Evidência V, Opinião do especialista.

Three-dimensional printing in orthopedics: where we stand and where we are heading / Impressão 3d na ortopedia: onde estamos e aonde podemos chegar

ABSTRACT Three-dimensional printing is a technology in expansion in the medical field. It also presents many applications in orthopedics. Our review article aims to describe 3D printing, types of 3D printers, and its use in the orthopedic field. 3D models can be created using tomography scans. Those models can then be manipulated, even simulating surgeries. It is possible to print biomodels, which will help us understand deformities and plan surgeries. Orthopedic surgeons must be updated in these disruptive technologies that may help their daily practice. Level of Evidence V, Expert opinion.

RESUMO A impressão 3D é uma tecnologia em expansão na medicina, possuindo diversas utilidades na ortopedia. O objetivo deste artigo de revisão é descrever o que é a impressão 3D, seus tipos e suas aplicações na ortopedia. Modelos em 3 dimensões podem ser criados a partir da tomografia computadorizada. Estes modelos podem ser manipulados em softwares específicos, onde inclusive cirurgias podem ser simuladas. Utilizando impressoras 3D podemos criar biomodelos que nos ajudam a compreender deformidades e planejar cirurgias. É importante que o ortopedista se mantenha atualizado nestas novas tecnologias disruptivas que podem auxiliar muito no seu dia a dia. Nível de Evidência V, Opinião do especialista.

Otobone ®: Three-dimensional printed Temporal Bone Biomodel for Simulation of Surgical Procedures

Introduction: The anatomy of the temporal bone is complex due to the large number of structures and functions grouped in this small bone space, which do not exist in any other region in the human body. With the difficulty of obtaining anatomical parts and the increasing number of ear, nose and throat (ENT) doctors, there was a need to create alternatives as real as possible for training otologic surgeons. Objective: Developing a technique to produce temporal bone models that allow them to maintain the external and internal anatomical features faithful to the natural bone. Methods: For this study, we used a computed tomography (CT) scan of the temporal bones of a 30-year-old male patient, with no structural morphological changes or any other pathology detected in the examination, which was later sent to a 3D printer in order to produce a temporal bone biomodel. Results: After dissection, the lead author evaluated the plasticity of the part and its similarity in drilling a natural bone as grade "4" on a scale of 0 to 5, in which 5 is the closest to the natural bone and 0 the farthest from the natural bone. All structures proposed in the method were found with the proposed color. Conclusion: It is concluded that it is feasible to use biomodels in surgical training of specialist doctors. After dissection of the bone biomodel, it was possible to find the anatomical structures proposed, and to reproduce the surgical approaches most used in surgical practice and training implants (AU)

Otobone ® : Three-dimensional printed Temporal Bone Biomodel for Simulation of Surgical Procedures.

Introduction The anatomy of the temporal bone is complex due to the large number of structures and functions grouped in this small bone space, which do not exist in any other region in the human body. With the difficulty of obtaining anatomical parts and the increasing number of ear, nose and throat (ENT) doctors, there was a need to create alternatives as real as possible for training otologic surgeons. Objective Developing a technique to produce temporal bone models that allow them to maintain the external and internal anatomical features faithful to the natural bone. Methods For this study, we used a computed tomography (CT) scan of the temporal bones of a 30-year-old male patient, with no structural morphological changes or any other pathology detected in the examination, which was later sent to a 3D printer in order to produce a temporal bone biomodel. Results After dissection, the lead author evaluated the plasticity of the part and its similarity in drilling a natural bone as grade "4" on a scale of 0 to 5, in which 5 is the closest to the natural bone and 0 the farthest from the natural bone. All structures proposed in the method were found with the proposed color. Conclusion It is concluded that it is feasible to use biomodels in surgical training of specialist doctors. After dissection of the bone biomodel, it was possible to find the anatomical structures proposed, and to reproduce the surgical approaches most used in surgical practice and training implants.

Computed tomography window affects kidney stones measurements

ABSTRACT Objectives Measurements of stone features may vary according to the non-contrast computed tomography (NCCT) technique. Using magnified bone window is the most accurate method to measure urinary stones. Possible differences between stone measurements in different NCCT windows have not been evaluated in stones located in the kidney. The aim of this study is to compare measurements of kidney stone features between NCCT bone and soft tissue windows in patients submitted to retrograde intrarenal surgery (RIRS). Materials and Methods Preoperative and 90th postoperative day NCCT were performed in 92 consecutive symptomatic adult patients (115 renal units) with kidney stones between 5 mm to 20 mm (< 15 mm in the lower calyx) treated by RIRS. NCCT were evaluated in the magnified bone window and soft tissue window in three axes in a different time by a single radiologist blinded for the measurements of the NCCT other method. Results Stone largest size (7.92±3.81 vs. 9.13±4.08; mm), volume (435.5±472.7 vs. 683.1±665.0; mm3) and density (989.4±330.2 vs. 893.0±324.6; HU) differed between bone and soft-tissue windows, respectively (p<0.0001) 5.2% of the renal units (6/115) were reclassified from residual fragments > 2 mm on soft tissue window to 0-2 mm on bone window. Conclusion Kidney stone measurements vary according to NCCT window. Measurements in soft tissue window NCCT of stone diameter and volume are larger and stone density is lesser than in bone window. These differences may have impact on clinical decisions.

Computed tomography window affects kidney stones measurements.

OBJECTIVES: Measurements of stone features may vary according to the non-contrast computed tomography (NCCT) technique. Using magnified bone window is the most accurate method to measure urinary stones. Possible differences between stone measurements in different NCCT windows have not been evaluated in stones located in the kidney. The aim of this study is to compare measurements of kidney stone features between NCCT bone and soft tissue windows in patients submitted to retrograde intrarenal surgery (RIRS). MATERIALS AND METHODS: Preoperative and 90th postoperative day NCCT were performed in 92 consecutive symptomatic adult patients (115 renal units) with kidney stones between 5 mm to 20 mm (< 15 mm in the lower calyx) treated by RIRS. NCCT were evaluated in the magnified bone window and soft tissue window in three axes in a different time by a single radiologist blinded for the measurements of the NCCT other method. RESULTS: Stone largest size (7.92±3.81 vs. 9.13±4.08; mm), volume (435.5±472.7 vs. 683.1±665.0; mm3) and density (989.4±330.2 vs. 893.0±324.6; HU) differed between bone and soft-tissue windows, respectively (p<0.0001) 5.2% of the renal units (6/115) were reclassified from residual fragments > 2 mm on soft tissue window to 0-2 mm on bone window. CONCLUSION: Kidney stone measurements vary according to NCCT window. Measurements in soft tissue window NCCT of stone diameter and volume are larger and stone density is lesser than in bone window. These differences may have impact on clinical decisions.

Size is Not Everything That Matters: Preoperative CT Predictors of Stone Free After RIRS.

OBJECTIVE: To define computed tomography (CT) predictors of residual fragments after retrograde intrarenal surgery (RIRS) for kidney stones up to 20 mm in patients never submitted to surgical procedures for stone removal. METHODS: From August 2016 to August 2017, symptomatic adult patients with kidney stones less than 20 mm treated by RIRS had their pre- and postoperative CT prospectively evaluated in search for predictors of residual stone fragments. Stone size, stone volume, number of stones, stone density, and location were evaluated in preoperative CT and analyzed as predictors for residual stone fragments on 90 POD CT. Stone location was represented by the infundibulopelvic angle (IPA) measured for each stone on preoperative noncontrast CT using multiplanar reconstruction. RESULTS: Ninety-two patients were successfully submitted to RIRS. Bilateral procedures were performed in 23 patients (25%) resulting in 115 renal units operated. Operative time was 54.5 ± 26.7 minutes (mean ± SD) and 96.7% (89/92) of the patients were discharged up to 12 hours after the procedure. Postoperative CT demonstrated stone-free in 86 of 115 (74.8%), 0-2 mm in 10 of 115 (8.7%), and > 2 mm residual fragments in 19 of 115 (16.5%) procedures. Logistic regression analysis revealed steep IPA was a predictor for any residual stone fragment after RIRS for kidney stones < 20 mm (P= .012). ROC curve showed that IPA < 41° was associated with a higher chance of residual fragments after RIRS. CONCLUSION: IPA < 41° is associated with a higher chance of residual fragments after RIRS for kidney stones up to 20 mm.

Use of 3D Printing in Preoperative Planning and Training for Aortic Endovascular Repair and Aortic Valve Disease.

INTRODUCTION: Three-dimensional (3D) printing has become an affordable tool for assisting heart surgeons in the aorta endovascular field, both in surgical planning, education and training of residents and students. This technique permits the construction of physical prototypes from conventional medical images by converting the anatomical information into computer aided design (CAD) files. OBJECTIVE: To present the 3D printing feature on developing prototypes leading to improved aortic endovascular surgical planning, as well as transcatheter aortic valve implantation, and mainly enabling training of the surgical procedure to be performed on patient's specific condition. METHODS: Six 3D printed real scale prototypes were built representing different aortic diseases, taken from real patients, to simulate the correction of the disease with endoprosthesis deployment. RESULTS: In the hybrid room, the 3D prototypes were examined under fluoroscopy, making it possible to obtain images that clearly delimited the walls of the aorta and its details. The endovascular simulation was then able to be performed, by correctly positioning the endoprosthesis, followed by its deployment. CONCLUSION: The 3D printing allowed the construction of aortic diseases realistic prototypes, offering a 3D view from the two-dimensional image of computed tomography (CT) angiography, allowing better surgical planning and surgeon training in the specific case beforehand.

Assessment of Residual Stone Fragments After Retrograde Intrarenal Surgery.

OBJECTIVES: To define the most suitable approach to assess residual stone fragments after retrograde intrarenal surgery (RIRS). METHODS: Ninety-two patients (115 renal units) submitted to RIRS for symptomatic kidney stones >5 mm and <20 mm or <15 mm in the lower Calyx diagnosed by noncontrast CT (NCCT) were prospectively studied. Residual fragments were assessed by endoscopic evaluation (END) at the end of the procedure and by NCCT, ultrasonography (US), and kidney, ureter, and bladder radiograph (KUB) on the 90th postoperative day (POD). NCCT was considered the gold standard for the evaluation of residual fragments after RIRS. RESULTS: The 90th POD NCCT resulted in stone-free status in 74.8% (86/115), 0-2 mm in 8.7% (10/115), and >2 mm residual fragments in 16.5% (19/115) renal units. Stone-free status by END at the end of RIRS was coincident with NCCT in 93.0% of the cases (40/43). There were no cases of residual fragments >2 mm on NCCT if END resulted in stone-free status. In all cases where END resulted in residual fragments >2 mm, US proved to be correct according to NCCT. Neither US nor KUB was able to identify residual fragments between 0 and 2 mm. KUB had only 31.6% (6/19) sensitivity to detect residual fragments >2 mm and did not add sensitivity or specificity to US. CONCLUSIONS: In the follow-up imaging after RIRS, we suggest that if END resulted in residual fragments <2 mm, a 90th POD NCCT should be performed. US may be used if END showed fragments >2 mm.

Use of 3d printing in preoperative planning and training for aortic endovascular repair and aortic valve disease

Abstract Introduction: Three-dimensional (3D) printing has become an affordable tool for assisting heart surgeons in the aorta endovascular field, both in surgical planning, education and training of residents and students. This technique permits the construction of physical prototypes from conventional medical images by converting the anatomical information into computer aided design (CAD) files. Objective: To present the 3D printing feature on developing prototypes leading to improved aortic endovascular surgical planning, as well as transcatheter aortic valve implantation, and mainly enabling training of the surgical procedure to be performed on patient's specific condition. Methods: Six 3D printed real scale prototypes were built representing different aortic diseases, taken from real patients, to simulate the correction of the disease with endoprosthesis deployment. Results: In the hybrid room, the 3D prototypes were examined under fluoroscopy, making it possible to obtain images that clearly delimited the walls of the aorta and its details. The endovascular simulation was then able to be performed, by correctly positioning the endoprosthesis, followed by its deployment. Conclusion: The 3D printing allowed the construction of aortic diseases realistic prototypes, offering a 3D view from the two-dimensional image of computed tomography (CT) angiography, allowing better surgical planning and surgeon training in the specific case beforehand.

Three-dimensional virtual reconstruction as a tool for preoperative planning in the management of complex anorectal fistulas / Reconstrução virtual tridimensional como ferramenta de planejamento pré-operatório de fístulas anais complexas

ABSTRACT The making of three-dimensional virtual models is a promising technology in preoperative planning, but that is not used in the treatment of anorectal fistulas. The objective of this work is to describe the development and initial experience of the construction of a virtual three-dimensional model of the pelvic anatomy of a patient, allowing the exact identification of the relationships between the fistulous tracts of complex anorectal fistulas and the other pelvic structures. An MRI was performed on this patient, and the images were exported to the Vitrea fX Workstation® software. A radiologist did the analysis and segmentation of the images that were then sent to a three-dimensional image processor (Meshlab v. 1.3.3 - ISTI - CNR Research Center, Pisa University, Italy®). The final 3D color image was analyzed by the surgeon and used to guide the catheterization of the fistulous pathways, the internal orifice and to assist in the identification of adjacent structures. The final three-dimensional model presented a high correlation with the intraoperative findings and facilitated the surgical planning.

RESUMO A criação de modelos virtuais tridimensionais é uma tecnologia promissora no planejamento pré-operatorio, entretanto não é utilizada no tratamento de fistulas anais. O objetivo desse trabalho é descrever o desenvolvimento e a experiência inicial da construção de um modelo virtual tridimensional da anatomia pélvica de um paciente, que permite a identificação exata das relações entre os tratos fistulosos de fistulas anais complexas e as demais estruturas pélvicas. O paciente realizou uma ressonância magnética e as imagens foram exportadas para o programa Vitrea fX software Workstation®. Um radiologista realizou a analise e segmentação das imagens que, em seguida, foram enviadas para um processador de imagens tridimensionais (Meshlab v. 1.3.3 - ISTI - CNR research center, Pisa University, Italy®). A imagem 3D colorida final foi analisada pelo cirurgião e utilizada para guiar a cateterização dos trajetos fistulosos e orifício interno e para auxiliar na identificação das estruturas adjacentes. O modelo tridimensional final apresentou alta correlação com os achados intraoperatorios e facilitou o planejamento cirúrgico.

Identification and differentiation of Candida species from pediatric patients by random amplified polymorphic DNA.

Thirty-four Candida isolates were analyzed by random amplified polymorphic DNA using the primer OPG-10:24 Candida albicans; 4 Candida tropicalis; 2 Candida parapsilosis; 2 Candida dubliniensis; 1 Candida glabrata and 1 Candida krusei. The UPGMA-Pearson correlation coefficient was used to calculate the genetic distance between the different Candida groupings. Samples were classified as identical (correlation of 100%); highly related samples (90%); moderately related samples (80%) and unrelated samples (< 70%). The results showed that the RAPD proposed was capable of classifying the isolates coherently (such that same species were in the same dendrogram), except for two isolates of Candida parapsilosis and the positive control (Netherlands, 1973), probably because they are now recognized as three different species. Concerning the only fluconazole-resistant Candida tropicalis isolate with a genotype that was different to the others, the data were insufficient to affirm that the only difference was the sensitivity to fluconazole. We concluded that the Random Amplified Polymorphic DNA proposed might be used to confirm Candida species identified by microbiological methods.

Identification and differentiation of Candida species from pediatric patients by random amplified polymorphic DNA

Thirty-four Candida isolates were analyzed by random amplified polymorphic DNA using the primer OPG-10:24 Candida albicans; 4 Candida tropicalis; 2 Candida parapsilosis; 2 Candida dubliniensis; 1 Candida glabrata and 1 Candida krusei. The UPGMA-Pearson correlation coefficient was used to calculate the genetic distance between the different Candida groupings. Samples were classified as identical (correlation of 100 percent); highly related samples (90 percent); moderately related samples (80 percent) and unrelated samples (< 70 percent). The results showed that the RAPD proposed was capable of classifying the isolates coherently (such that same species were in the same dendrogram), except for two isolates of Candida parapsilosis and the positive control (Netherlands, 1973), probably because they are now recognized as three different species. Concerning the only fluconazole-resistant Candida tropicalis isolate with a genotype that was different to the others, the data were insufficient to affirm that the only difference was the sensitivity to fluconazole. We concluded that the Random Amplified Polymorphic DNA proposed might be used to confirm Candida species identified by microbiological methods.

Trinta e quatro isolados de Candida foram analisados por amplificação aleatória de DNA polimórfico (primer OPG-10): 24 Candida albicans, 4 Candida tropicalis, 2 Candida parapsilosis, 2 Candida dubliniensis, 1 Candida glabrata e 1 Candida krusei. O coeficiente de correlação de Pearson-UPGMA calculou a distância genética entre os diferentes agrupamentos de Candida: amostras idênticas (100 por cento de correlação), amostras muito relacionadas (90 por cento), moderadamente relacionadas (80 por cento), e não relacionadas (< 70 por cento). Os resultados demonstram que a amplificação aleatória de DNA polimórfico proposta é capaz de classificar os isolados de forma coerente, ficando os de mesma espécie em um mesmo dendograma, com exceção dos dois isolados de Candida parapsilosis e o controle positivo (Holanda, 1973), provavelmente por serem atualmente classificadas em três espécies diferentes. Quanto ao único isolado de Candida tropicalis resistente ao fluconazol com genótipo diferente dos outros, os dados não são suficientes para afirmar que a única característica distinta fosse a sensibilidade ao fluconazol. Concluímos que a amplificação aleatória de DNA polimórfico proposta poderia ser usada para a confirmação das espécies de Candida identificadas nos testes microbiológicos.