An Efficient Deep Neural Network to Classify Large 3D Images With Small Objects.

3D imaging enables accurate diagnosis by providing spatial information about organ anatomy. However, using 3D images to train AI models is computationally challenging because they consist of 10x or 100x more pixels than their 2D counterparts. To be trained with high-resolution 3D images, convolutional neural networks resort to downsampling them or projecting them to 2D. We propose an effective alternative, a neural network that enables efficient classification of full-resolution 3D medical images. Compared to off-the-shelf convolutional neural networks, our network, 3D Globally-Aware Multiple Instance Classifier (3D-GMIC), uses 77.98%-90.05% less GPU memory and 91.23%-96.02% less computation. While it is trained only with image-level labels, without segmentation labels, it explains its predictions by providing pixel-level saliency maps. On a dataset collected at NYU Langone Health, including 85,526 patients with full-field 2D mammography (FFDM), synthetic 2D mammography, and 3D mammography, 3D-GMIC achieves an AUC of 0.831 (95% CI: 0.769-0.887) in classifying breasts with malignant findings using 3D mammography. This is comparable to the performance of GMIC on FFDM (0.816, 95% CI: 0.737-0.878) and synthetic 2D (0.826, 95% CI: 0.754-0.884), which demonstrates that 3D-GMIC successfully classified large 3D images despite focusing computation on a smaller percentage of its input compared to GMIC. Therefore, 3D-GMIC identifies and utilizes extremely small regions of interest from 3D images consisting of hundreds of millions of pixels, dramatically reducing associated computational challenges. 3D-GMIC generalizes well to BCS-DBT, an external dataset from Duke University Hospital, achieving an AUC of 0.848 (95% CI: 0.798-0.896).

Enhancing the reliability and accuracy of AI-enabled diagnosis via complementarity-driven deferral to clinicians.

Predictive artificial intelligence (AI) systems based on deep learning have been shown to achieve expert-level identification of diseases in multiple medical imaging settings, but can make errors in cases accurately diagnosed by clinicians and vice versa. We developed Complementarity-Driven Deferral to Clinical Workflow (CoDoC), a system that can learn to decide between the opinion of a predictive AI model and a clinical workflow. CoDoC enhances accuracy relative to clinician-only or AI-only baselines in clinical workflows that screen for breast cancer or tuberculosis (TB). For breast cancer screening, compared to double reading with arbitration in a screening program in the UK, CoDoC reduced false positives by 25% at the same false-negative rate, while achieving a 66% reduction in clinician workload. For TB triaging, compared to standalone AI and clinical workflows, CoDoC achieved a 5-15% reduction in false positives at the same false-negative rate for three of five commercially available predictive AI systems. To facilitate the deployment of CoDoC in novel futuristic clinical settings, we present results showing that CoDoC's performance gains are sustained across several axes of variation (imaging modality, clinical setting and predictive AI system) and discuss the limitations of our evaluation and where further validation would be needed. We provide an open-source implementation to encourage further research and application.

Improving Information Extraction from Pathology Reports using Named Entity Recognition.

Pathology reports are considered the gold standard in medical research due to their comprehensive and accurate diagnostic information. Natural language processing (NLP) techniques have been developed to automate information extraction from pathology reports. However, existing studies suffer from two significant limitations. First, they typically frame their tasks as report classification, which restricts the granularity of extracted information. Second, they often fail to generalize to unseen reports due to variations in language, negation, and human error. To overcome these challenges, we propose a BERT (bidirectional encoder representations from transformers) named entity recognition (NER) system to extract key diagnostic elements from pathology reports. We also introduce four data augmentation methods to improve the robustness of our model. Trained and evaluated on 1438 annotated breast pathology reports, acquired from a large medical center in the United States, our BERT model trained with data augmentation achieves an entity F1-score of 0.916 on an internal test set, surpassing the BERT baseline (0.843). We further assessed the model's generalizability using an external validation dataset from the United Arab Emirates, where our model maintained satisfactory performance (F1-score 0.860). Our findings demonstrate that our NER systems can effectively extract fine-grained information from widely diverse medical reports, offering the potential for large-scale information extraction in a wide range of medical and AI research. We publish our code at https://github.com/nyukat/pathology_extraction.

A Competition, Benchmark, Code, and Data for Using Artificial Intelligence to Detect Lesions in Digital Breast Tomosynthesis.

Importance: An accurate and robust artificial intelligence (AI) algorithm for detecting cancer in digital breast tomosynthesis (DBT) could significantly improve detection accuracy and reduce health care costs worldwide. Objectives: To make training and evaluation data for the development of AI algorithms for DBT analysis available, to develop well-defined benchmarks, and to create publicly available code for existing methods. Design, Setting, and Participants: This diagnostic study is based on a multi-institutional international grand challenge in which research teams developed algorithms to detect lesions in DBT. A data set of 22 032 reconstructed DBT volumes was made available to research teams. Phase 1, in which teams were provided 700 scans from the training set, 120 from the validation set, and 180 from the test set, took place from December 2020 to January 2021, and phase 2, in which teams were given the full data set, took place from May to July 2021. Main Outcomes and Measures: The overall performance was evaluated by mean sensitivity for biopsied lesions using only DBT volumes with biopsied lesions; ties were broken by including all DBT volumes. Results: A total of 8 teams participated in the challenge. The team with the highest mean sensitivity for biopsied lesions was the NYU B-Team, with 0.957 (95% CI, 0.924-0.984), and the second-place team, ZeDuS, had a mean sensitivity of 0.926 (95% CI, 0.881-0.964). When the results were aggregated, the mean sensitivity for all submitted algorithms was 0.879; for only those who participated in phase 2, it was 0.926. Conclusions and Relevance: In this diagnostic study, an international competition produced algorithms with high sensitivity for using AI to detect lesions on DBT images. A standardized performance benchmark for the detection task using publicly available clinical imaging data was released, with detailed descriptions and analyses of submitted algorithms accompanied by a public release of their predictions and code for selected methods. These resources will serve as a foundation for future research on computer-assisted diagnosis methods for DBT, significantly lowering the barrier of entry for new researchers.

Application of HoloLens-based augmented reality and three-dimensional printed anatomical tooth reference models in dental education.

Tooth anatomy is fundamental knowledge used in everyday dental practice to reconstruct the occlusal surface during cavity fillings. The main objective of this project was to evaluate the suitability of two types of anatomical tooth reference models used to support reconstruction of the occlusal anatomy of the teeth: (1) a three-dimensional (3D)-printed model and (2) a model displayed in augmented reality (AR) using Microsoft HoloLens. The secondary objective was to evaluate three aspects impacting the outcome: clinical experience, comfort of work, and other variables. The tertiary objective was to evaluate the usefulness of AR in dental education. Anatomical models of crowns of three different molars were made using cone beam computed tomography image segmentation, printed with a stereolithographic 3D-printer, and then displayed in the HoloLens. Each participant reconstructed the occlusal anatomy of three teeth. One without any reference materials and two with an anatomical reference model, either 3D-printed or holographic. The reconstruction work was followed by the completion of an evaluation questionnaire. The maximum Hausdorff distances (Hmax) between the superimposed images of the specimens after the procedures and the anatomical models were then calculated. The results showed that the most accurate but slowest reconstruction was achieved with the use of 3D-printed reference models and that the results were not affected by other aspects considered. For this method, the Hmax was observed to be 630 µm (p = 0.004). It was concluded that while AR models can be helpful in dental anatomy education, they are not suitable replacements for physical models.

Improving breast cancer diagnostics with deep learning for MRI.

Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) has a high sensitivity in detecting breast cancer but often leads to unnecessary biopsies and patient workup. We used a deep learning (DL) system to improve the overall accuracy of breast cancer diagnosis and personalize management of patients undergoing DCE-MRI. On the internal test set (n = 3936 exams), our system achieved an area under the receiver operating characteristic curve (AUROC) of 0.92 (95% CI: 0.92 to 0.93). In a retrospective reader study, there was no statistically significant difference (P = 0.19) between five board-certified breast radiologists and the DL system (mean ΔAUROC, +0.04 in favor of the DL system). Radiologists' performance improved when their predictions were averaged with DL's predictions [mean ΔAUPRC (area under the precision-recall curve), +0.07]. We demonstrated the generalizability of the DL system using multiple datasets from Poland and the United States. An additional reader study on a Polish dataset showed that the DL system was as robust to distribution shift as radiologists. In subgroup analysis, we observed consistent results across different cancer subtypes and patient demographics. Using decision curve analysis, we showed that the DL system can reduce unnecessary biopsies in the range of clinically relevant risk thresholds. This would lead to avoiding biopsies yielding benign results in up to 20% of all patients with BI-RADS category 4 lesions. Last, we performed an error analysis, investigating situations where DL predictions were mostly incorrect. This exploratory work creates a foundation for deployment and prospective analysis of DL-based models for breast MRI.

Radiological Features for Frailty Assessment in Patients Requiring Emergency Laparotomy.

INTRODUCTION: As the number of elderly patients requiring surgical intervention rises, it is believed that frailty syndrome has a greater impact on perioperative course than on chronological age. The aim of this study was to evaluate the efficacy of various imaging features for frailty assessment in patients undergoing emergency laparotomy. METHODS: The study included all patients that qualified for emergency surgery with preoperative CT scans between 2016 and 2020 in the Second Department of General Surgery. Multiple trauma patients were excluded from the analysis. The modified frailty index and brief geriatric assessment were used in the analysis. CT images were reviewed for the assessment of osteopenia, sarcopenia, sarcopenic obesity, renal volume and abdominal aorta calcification rate. RESULTS: A total of 261 patients were included in the analysis. Multivariate logistic regression identified every next ASA class (OR: 4.161, 95%CI: 1.672-10.355, p = 0.002), intraoperative adverse events (OR: 12.397, 95%CI: 2.166-70.969, p = 0.005) and osteopenia (OR: 4.213, 95%CI: 1.235-14.367, p = 0.022) as a risk factor for 30-day mortality. Our study showed that every next ASA class (OR: 1.952, 95%Cl: 1.171-3.256, p = 0.010) and every point of the BGA score (OR: 1.496, 95%Cl: 1.110-2.016, p = 0.008) are risk factors for major complications. CONCLUSIONS: Osteopenia was the best parameter for perioperative mortality risk stratification in patients undergoing emergency surgical intervention. Sarcopenia (measured as psoas muscle area), sarcopenic obesity, aortic calcifications and mean kidney volume do not predict poor outcomes in those patients. None of the radiological markers appeared to be useful for the prediction of perioperative morbidity.

Development and Validation of a Virtual Reality Simulator for Robot-Assisted Minimally Invasive Liver Surgery Training.

The value of kinematic data for skill assessment is being investigated. This is the first virtual reality simulator developed for liver surgery. This simulator was coded in C++ using PhysX and FleX with a novel cutting algorithm and used a patient data-derived model and two instruments functioning as ultrasonic shears. The simulator was evaluated by nine expert surgeons and nine surgical novices. Each participant performed a simulated metastasectomy after training. Kinematic data were collected for the instrument position. Each participant completed a survey. The expert participants had a mean age of 47 years and 9/9 were certified in surgery. Novices had a mean age of 30 years and 0/9 were certified surgeons. The mean path length (novice 0.76 ± 0.20 m vs. expert 0.46 ± 0.16 m, p = 0.008), movements (138 ± 45 vs. 84 ± 32, p = 0.043) and time (174 ± 44 s vs. 102 ± 42 s, p = 0.004) were significantly different for the two participant groups. There were no significant differences in activating the instrument (107 ± 25 vs. 109 ± 53). Participants considered the simulator realistic (6.5/7) (face validity), appropriate for education (5/7) (content validity) with an effective interface (6/7), consistent motion (5/7) and realistic soft tissue behavior (5/7). This study showed that the simulator differentiates between experts and novices. Simulation may be an effective way to obtain kinematic data.

Quality of Life After Bariatric Surgery-a Systematic Review with Bayesian Network Meta-analysis.

OBJECTIVE: Comprehensive analysis and comparison of HRQoL following different bariatric interventions through systematic review with network meta-analysis. BACKGROUND: Different types of bariatric surgeries have been developed throughout the years. Apart from weight loss and comorbidities remission, improvement of health-related quality of life (HRQoL) is an important outcome of metabolic surgery. METHODS: MEDLINE, EMBASE, and Scopus databases have been searched up to April 2020. Inclusion criteria to the analysis were (1) study with at least 2 arms comparing bariatric surgeries; (2) reporting of HRQoL with a validated tool; (3) follow-up period of 1, 2, 3, or 5 years. Network meta-analysis was conducted using Bayesian statistics. The primary outcome was HRQoL. RESULTS: Forty-seven studies were included in the analysis involving 26,629 patients and 11 different surgeries such as sleeve gastrectomy (LSG), gastric bypass (LRYGB), one anastomosis gastric bypass (OAGB), and other. At 1 year, there was significant difference in HRQoL in favor of LSG, LRYGB, and OAG compared with lifestyle intervention (SMD: 0.44; 95% CrI 0.2 to 0.68 for LSG, SMD: 0.56; 95% CrI 0.31 to 0.8 for LRYGB; and SMD: 0.43; 95% CrI 0.06 to 0.8 for OAGB). At 5 years, LSG, LRYGB, and OAGB showed better HRQoL compared to control (SMD: 0.92; 95% CrI 0.58 to 1.26, SMD: 1.27; 95% CrI 0.94 to 1.61, and SMD: 1.01; 95% CrI 0.63 to 1.4, respectively). CONCLUSIONS: LSG and LRYGB may lead to better HRQoL across most follow-up time points. Long-term analysis shows that bariatric intervention results in better HRQoL than non-surgical interventions.

Artificial intelligence system reduces false-positive findings in the interpretation of breast ultrasound exams.

Though consistently shown to detect mammographically occult cancers, breast ultrasound has been noted to have high false-positive rates. In this work, we present an AI system that achieves radiologist-level accuracy in identifying breast cancer in ultrasound images. Developed on 288,767 exams, consisting of 5,442,907 B-mode and Color Doppler images, the AI achieves an area under the receiver operating characteristic curve (AUROC) of 0.976 on a test set consisting of 44,755 exams. In a retrospective reader study, the AI achieves a higher AUROC than the average of ten board-certified breast radiologists (AUROC: 0.962 AI, 0.924 ± 0.02 radiologists). With the help of the AI, radiologists decrease their false positive rates by 37.3% and reduce requested biopsies by 27.8%, while maintaining the same level of sensitivity. This highlights the potential of AI in improving the accuracy, consistency, and efficiency of breast ultrasound diagnosis.

An artificial intelligence system for predicting the deterioration of COVID-19 patients in the emergency department.

During the coronavirus disease 2019 (COVID-19) pandemic, rapid and accurate triage of patients at the emergency department is critical to inform decision-making. We propose a data-driven approach for automatic prediction of deterioration risk using a deep neural network that learns from chest X-ray images and a gradient boosting model that learns from routine clinical variables. Our AI prognosis system, trained using data from 3661 patients, achieves an area under the receiver operating characteristic curve (AUC) of 0.786 (95% CI: 0.745-0.830) when predicting deterioration within 96 hours. The deep neural network extracts informative areas of chest X-ray images to assist clinicians in interpreting the predictions and performs comparably to two radiologists in a reader study. In order to verify performance in a real clinical setting, we silently deployed a preliminary version of the deep neural network at New York University Langone Health during the first wave of the pandemic, which produced accurate predictions in real-time. In summary, our findings demonstrate the potential of the proposed system for assisting front-line physicians in the triage of COVID-19 patients.

MarkIt: A Collaborative Artificial Intelligence Annotation Platform Leveraging Blockchain For Medical Imaging Research.

Current research on medical image processing relies heavily on the amount and quality of input data. Specifically, supervised machine learning methods require well-annotated datasets. A lack of annotation tools limits the potential to achieve high-volume processing and scaled systems with a proper reward mechanism. We developed MarkIt, a web-based tool, for collaborative annotation of medical imaging data with artificial intelligence and blockchain technologies. Our platform handles both Digital Imaging and Communications in Medicine (DICOM) and non-DICOM images, and allows users to annotate them for classification and object detection tasks in an efficient manner. MarkIt can accelerate the annotation process and keep track of user activities to calculate a fair reward. A proof-of-concept experiment was conducted with three fellowship-trained radiologists, each of whom annotated 1,000 chest X-ray studies for multi-label classification. We calculated the inter-rater agreement and estimated the value of the dataset to distribute the reward for annotators using a crypto currency. We hypothesize that MarkIt allows the typically arduous annotation task to become more efficient. In addition, MarkIt can serve as a platform to evaluate the value of data and trade the annotation results in a more scalable manner in the future. The platform is publicly available for testing on https://markit.mgh.harvard.edu.

High compliance to ERAS protocol does not improve overall survival in patients treated for resectable advanced gastric cancer.

INTRODUCTION: The ERAS (Enhanced Recovery after Surgery) protocol revolutionized perioperative care for gastrointestinal surgical procedures. However, little is known about the association between adherence to the ERAS protocol in gastric cancer surgery and the oncological outcome. AIM: To explore the relation between adherence to the ERAS protocol and the oncological outcome in gastric cancer patients. MATERIAL AND METHODS: We performed a retrospective analysis of a prospectively collected database of patients treated for gastric cancer between 2013 and 2016. All patients were treated perioperatively with a 14-item ERAS protocol. Every patient underwent regular follow-up every 3 months for 3 years after surgery. 80% compliance to the ERAS protocol was the goal during perioperative care. Based on the level of compliance, patients were divided into group 1 and group 2 (compliance of ≥ 80% and < 80%, respectively). RESULTS: Compliance to the ERAS protocol was not a risk factor for diminished overall survival - probability of 3-year survival was 63% in group 1 and 56% in group 2 (p = 0.75). The proportional Cox model revealed that only stage III gastric cancer was a risk factor of poor prognosis in patients operated on for gastric cancer (HR = 7.89, 95% CI: 2.96-20.89; p = 0.0001). CONCLUSIONS: High adherence to the ERAS protocol did not improve overall survival in our 3-year observation. Only the stage of the disease, according to the AJCC classification, was identified as a risk factor for poor prognosis.

An artificial intelligence system for predicting the deterioration of COVID-19 patients in the emergency department.

During the coronavirus disease 2019 (COVID-19) pandemic, rapid and accurate triage of patients at the emergency department is critical to inform decision-making. We propose a data-driven approach for automatic prediction of deterioration risk using a deep neural network that learns from chest X-ray images and a gradient boosting model that learns from routine clinical variables. Our AI prognosis system, trained using data from 3,661 patients, achieves an area under the receiver operating characteristic curve (AUC) of 0.786 (95% CI: 0.745-0.830) when predicting deterioration within 96 hours. The deep neural network extracts informative areas of chest X-ray images to assist clinicians in interpreting the predictions and performs comparably to two radiologists in a reader study. In order to verify performance in a real clinical setting, we silently deployed a preliminary version of the deep neural network at New York University Langone Health during the first wave of the pandemic, which produced accurate predictions in real-time. In summary, our findings demonstrate the potential of the proposed system for assisting front-line physicians in the triage of COVID-19 patients.

Radiological Society of North America (RSNA) 3D Printing Special Interest Group (SIG) clinical situations for which 3D printing is considered an appropriate representation or extension of data contained in a medical imaging examination: abdominal, hepatobiliary, and gastrointestinal conditions.

BACKGROUND: Medical 3D printing has demonstrated value in anatomic models for abdominal, hepatobiliary, and gastrointestinal conditions. A writing group composed of the Radiological Society of North America (RSNA) Special Interest Group on 3D Printing (SIG) provides appropriateness criteria for abdominal, hepatobiliary, and gastrointestinal 3D printing indications. METHODS: A literature search was conducted to identify all relevant articles using 3D printing technology associated with a number of abdominal pathologic processes. Each included study was graded according to published guidelines. RESULTS: Evidence-based appropriateness guidelines are provided for the following areas: intra-hepatic masses, hilar cholangiocarcinoma, biliary stenosis, biliary stones, gallbladder pathology, pancreatic cancer, pancreatitis, splenic disease, gastric pathology, small bowel pathology, colorectal cancer, perianal fistula, visceral trauma, hernia, abdominal sarcoma, abdominal wall masses, and intra-abdominal fluid collections. CONCLUSION: This document provides initial appropriate use criteria for medical 3D printing in abdominal, hepatobiliary, and gastrointestinal conditions.

Simulation and Training of Needle Puncture Procedure with a Patient-Specific 3D Printed Gluteal Artery Model.

The puncture of the gluteal artery (GA) is a rare and difficult procedure. Less experienced clinicians do not always have the opportunity to practice and prepare for it, which creates a need for novel training tools. We aimed to investigate the feasibility of developing a 3D-printed, patient-specific phantom of the GA and its surrounding tissues to determine the extent to which the model can be used as an aid in needle puncture planning, simulation, and training. Computed tomography angiography scans of a patient with an endoleak to an internal iliac artery aneurysm with no intravascular antegrade access were processed. The arterial system, including the superior GA with its division branches, and pelvic area bones were 3D printed. The 3D model was embedded in the buttocks-shaped, patient-specific mold and cast. The manufactured, life-sized phantom was used to simulate the GA puncture procedure and was validated by 13 endovascular specialists. The printed GA was visible in the fluoroscopy, allowing for a needle puncture procedure simulation. The contrast medium was administered, simulating a digital subtraction angiography. Participating doctors suggested that the model could make a significant impact on preprocedural planning and resident training programs. Although the results are promising, we recommend that further studies be used to adjust the design and assess its clinical value.

Evaluation of the learning curve of transanal total mesorectal excision: single-centre experience.

INTRODUCTION: Transanal total mesorectal excision (TaTME) has been recently proposed to overcome the difficulties of the standard TME approach, allowing better visualization and dissection of the mesorectal fascia. Although TaTME seems very promising, the evidence and body of knowledge on achieving proficiency in performing it are still sparse. AIM: To evaluate the learning curve of TaTME based on a single centre's experience. MATERIAL AND METHODS: Consecutive patients undergoing TaTME since 2014 in a tertiary referral department were included in the study. All procedures were performed by one experienced surgeon. CUSUM curve analyses were performed to evaluate learning curves. RESULTS: Sixty-six patients underwent TaTME. After analysis of postoperative morbidity rate, intraoperative adverse effects and operative time, we estimated that 40 cases are needed to achieve TaTME proficiency. Subsequently, patients were divided into two groups: before (40 patients) and after overcoming the learning curve (26 patients). Group 1 had higher readmission (p = 0.041) and complication rates (p = 0.019). There were no statistically significant differences in terms of intraoperative adverse effects, length of stay or pathological quality of the specimen. CONCLUSIONS: Transanal total mesorectal excision is a promising yet technically demanding procedure and requires at least 40 cases to complete the learning curve. More data are needed to introduce it as a standard procedure for low rectal cancer treatment.

Decision-making based on 3D printed models in laparoscopic liver resections with intraoperative ultrasound: a prospective observational study.

OBJECTIVES: The aim of this study was to evaluate impact of 3D printed models on decision-making in context of laparoscopic liver resections (LLR) performed with intraoperative ultrasound (IOUS) guidance. METHODS: Nineteen patients with liver malignances (74% were colorectal cancer metastases) were prospectively qualified for LLR or radiofrequency ablation in a single center from April 2017 to December 2018. Models were 3DP in all cases based on CT and facilitated optical visualization of tumors' relationships with portal and hepatic veins. Planned surgical extent and its changes were tracked after CT analysis and 3D model inspection, as well as intraoperatively using IOUS. RESULTS: Nineteen patients were included in the analysis. Information from either 3DP or IOUS led to changes in the planned surgical approach in 13/19 (68%) patients. In 5/19 (26%) patients, the 3DP model altered the plan of the surgery preoperatively. In 4/19 (21%) patients, 3DP independently changed the approach. In one patient, IOUS modified the plan post-3DP. In 8/19 (42%) patients, 3DP model did not change the approach, whereas IOUS did. In total, IOUS altered surgical plans in 9 (47%) cases. Most of those changes (6/9; 67%) were caused by detection of additional lesions not visible on CT and 3DP. CONCLUSIONS: 3DP can be helpful in planning complex and major LLRs and led to changes in surgical approach in 26.3% (5/19 patients) in our series. 3DP may serve as a useful adjunct to IOUS. KEY POINTS: • 3D printing can help in decision-making before major and complex resections in patients with liver cancer. • In 5/19 patients, 3D printed model altered surgical plan preoperatively. • Most surgical plan changes based on intraoperative ultrasonography were caused by detection of additional lesions not visible on CT and 3D model.

Investigating Risk Factors for Complications after Ileostomy Reversal in Low Anterior Rectal Resection Patients: An Observational Study.

Introduction: Defunctioning ileostomy has been widely used in patients undergoing low anterior rectal resection to reduce the rate of postoperative leakage. It is still not clear whether interval between primary procedure and ileostomy reversal has an impact on treatment outcomes. Methods: In our prospective observational study we reviewed 164 consecutive cases of patients who underwent total mesorectal excision with primary anastomosis. Univariate and multivariate regression models were used to search for risk factors for prolonged length of stay and complications after defunctioning ileostomy reversal. Receiver operating characteristic curves were utilized to set cut-off points for prolonged length of stay and perioperative morbidity. Results: In total, 132 patients were included in the statistical analysis. The median interval between primary procedure and defunctioning ileostomy reversal was 134 (range: 17-754) days, while median length of stay was 5 days (4-6 interquartile range (IQR)). Prolonged length of stay cut-off was established at 6 days. Regression models revealed that interval between primary surgery and stoma closure as well as complications after primary procedure are risk factors for complications after defunctioning ileostomy reversal. Prolonged length of stay has been found to be related primarily to interval between primary surgery and stoma closure. Conclusions: In our study interval between primary surgery and stoma closure along with complication occurrence after primary procedure are risk factors for perioperative morbidity and prolonged length of stay (LOS) after ileostomy reversal. The effort should be made to minimize the interval to ileostomy reversal. However, randomized studies are necessary to avoid the bias which appears in this observational study and confirm our findings.