Drop the shortcuts: image augmentation improves fairness and decreases AI detection of race and other demographics from medical images.

BACKGROUND: It has been shown that AI models can learn race on medical images, leading to algorithmic bias. Our aim in this study was to enhance the fairness of medical image models by eliminating bias related to race, age, and sex. We hypothesise models may be learning demographics via shortcut learning and combat this using image augmentation. METHODS: This study included 44,953 patients who identified as Asian, Black, or White (mean age, 60.68 years ±18.21; 23,499 women) for a total of 194,359 chest X-rays (CXRs) from MIMIC-CXR database. The included CheXpert images comprised 45,095 patients (mean age 63.10 years ±18.14; 20,437 women) for a total of 134,300 CXRs were used for external validation. We also collected 1195 3D brain magnetic resonance imaging (MRI) data from the ADNI database, which included 273 participants with an average age of 76.97 years ±14.22, and 142 females. DL models were trained on either non-augmented or augmented images and assessed using disparity metrics. The features learned by the models were analysed using task transfer experiments and model visualisation techniques. FINDINGS: In the detection of radiological findings, training a model using augmented CXR images was shown to reduce disparities in error rate among racial groups (-5.45%), age groups (-13.94%), and sex (-22.22%). For AD detection, the model trained with augmented MRI images was shown 53.11% and 31.01% reduction of disparities in error rate among age and sex groups, respectively. Image augmentation led to a reduction in the model's ability to identify demographic attributes and resulted in the model trained for clinical purposes incorporating fewer demographic features. INTERPRETATION: The model trained using the augmented images was less likely to be influenced by demographic information in detecting image labels. These results demonstrate that the proposed augmentation scheme could enhance the fairness of interpretations by DL models when dealing with data from patients with different demographic backgrounds. FUNDING: National Science and Technology Council (Taiwan), National Institutes of Health.

Enabling chronic obstructive pulmonary disease diagnosis through chest X-rays: A multi-site and multi-modality study.

PURPOSE: Chronic obstructive pulmonary disease (COPD) is one of the most common chronic illnesses in the world. Unfortunately, COPD is often difficult to diagnose early when interventions can alter the disease course, and it is underdiagnosed or only diagnosed too late for effective treatment. Currently, spirometry is the gold standard for diagnosing COPD but it can be challenging to obtain, especially in resource-poor countries. Chest X-rays (CXRs), however, are readily available and may have the potential as a screening tool to identify patients with COPD who should undergo further testing or intervention. In this study, we used three CXR datasets alongside their respective electronic health records (EHR) to develop and externally validate our models. METHOD: To leverage the performance of convolutional neural network models, we proposed two fusion schemes: (1) model-level fusion, using Bootstrap aggregating to aggregate predictions from two models, (2) data-level fusion, using CXR image data from different institutions or multi-modal data, CXR image data, and EHR data for model training. Fairness analysis was then performed to evaluate the models across different demographic groups. RESULTS: Our results demonstrate that DL models can detect COPD using CXRs with an area under the curve of over 0.75, which could facilitate patient screening for COPD, especially in low-resource regions where CXRs are more accessible than spirometry. CONCLUSIONS: By using a ubiquitous test, future research could build on this work to detect COPD in patients early who would not otherwise have been diagnosed or treated, altering the course of this highly morbid disease.

Advanced Endoscopy for Thoracic Surgeons.

The thoracic surgeon, well versed in advanced endoscopy, has an array of therapeutic options for foregut pathologic conditions. Peroral endoscopic myotomy (POEM) offers a less-invasive means to treat achalasia, and the authors' preferred approach is described in this article. They also describe variations of POEM, such as G-POEM, Z-POEM, and D-POEM. In addition, endoscopic stenting, endoluminal vacuum therapy, endoscopic internal drainage, and endoscopic suturing/clipping are discussed and can be valuable tools for esophageal leaks and perforations. Endoscopic procedures are advancing rapidly, and thoracic surgeons must maintain at the forefront of these technologies.

Robotic Surgery for Tracheobronchomalacia.

Robotic tracheobronchoplasty (TBP) is a new surgical approach to treat tracheobronchomalacia (TBM), and is based on open TBP, which has undergone various technical iterations since the 1950s. The robotic approach to TBM may allow for more patients to undergo TBP, and in retrospective series has demonstrated equivalent outcomes compared to open TBP if not superior in terms of postoperative complications. Long-term data are eagerly awaited, and we describe our approach in this article.

Robotic Mediastinal Surgery.

The robotic platform can be viewed as an advanced thoracoscopic instrument and can be utilized for any pathology amenable to thoracoscopic surgery. This ultimately comes down to surgeon comfort, but many have demonstrated the robotic approach to be useful in benign and malignant mediastinal disease in all compartments with at least equivalent-if not superior-outcomes compared to sternotomy for many metrics. There are various robotic approaches to the same compartments (such as with thymectomy), and no one robotic approach has proven superior to another. Here we describe our robotic approach to common mediastinal pathology.

Using Artificial Intelligence-Enhanced Sensing and Wearable Technology in Sports Medicine and Performance Optimisation.

Wearable technologies are small electronic and mobile devices with wireless communication capabilities that can be worn on the body as a part of devices, accessories or clothes. Sensors incorporated within wearable devices enable the collection of a broad spectrum of data that can be processed and analysed by artificial intelligence (AI) systems. In this narrative review, we performed a literature search of the MEDLINE, Embase and Scopus databases. We included any original studies that used sensors to collect data for a sporting event and subsequently used an AI-based system to process the data with diagnostic, treatment or monitoring intents. The included studies show the use of AI in various sports including basketball, baseball and motor racing to improve athletic performance. We classified the studies according to the stage of an event, including pre-event training to guide performance and predict the possibility of injuries; during events to optimise performance and inform strategies; and in diagnosing injuries after an event. Based on the included studies, AI techniques to process data from sensors can detect patterns in physiological variables as well as positional and kinematic data to inform how athletes can improve their performance. Although AI has promising applications in sports medicine, there are several challenges that can hinder their adoption. We have also identified avenues for future work that can provide solutions to overcome these challenges.

Global healthcare fairness: We should be sharing more, not less, data.

The availability of large, deidentified health datasets has enabled significant innovation in using machine learning (ML) to better understand patients and their diseases. However, questions remain regarding the true privacy of this data, patient control over their data, and how we regulate data sharing in a way that that does not encumber progress or further potentiate biases for underrepresented populations. After reviewing the literature on potential reidentifications of patients in publicly available datasets, we argue that the cost-measured in terms of access to future medical innovations and clinical software-of slowing ML progress is too great to limit sharing data through large publicly available databases for concerns of imperfect data anonymization. This cost is especially great for developing countries where the barriers preventing inclusion in such databases will continue to rise, further excluding these populations and increasing existing biases that favor high-income countries. Preventing artificial intelligence's progress towards precision medicine and sliding back to clinical practice dogma may pose a larger threat than concerns of potential patient reidentification within publicly available datasets. While the risk to patient privacy should be minimized, we believe this risk will never be zero, and society has to determine an acceptable risk threshold below which data sharing can occur-for the benefit of a global medical knowledge system.

A scoping review of artificial intelligence applications in thoracic surgery.

OBJECTIVES: Machine learning (ML) has great potential, but there are few examples of its implementation improving outcomes. The thoracic surgeon must be aware of pertinent ML literature and how to evaluate this field for the safe translation to patient care. This scoping review provides an introduction to ML applications specific to the thoracic surgeon. We review current applications, limitations and future directions. METHODS: A search of the PubMed database was conducted with inclusion requirements being the use of an ML algorithm to analyse patient information relevant to a thoracic surgeon and contain sufficient details on the data used, ML methods and results. Twenty-two papers met the criteria and were reviewed using a methodological quality rubric. RESULTS: ML demonstrated enhanced preoperative test accuracy, earlier pathological diagnosis, therapies to maximize survival and predictions of adverse events and survival after surgery. However, only 4 performed external validation. One demonstrated improved patient outcomes, nearly all failed to perform model calibration and one addressed fairness and bias with most not generalizable to different populations. There was a considerable variation to allow for reproducibility. CONCLUSIONS: There is promise but also challenges for ML in thoracic surgery. The transparency of data and algorithm design and the systemic bias on which models are dependent remain issues to be addressed. Although there has yet to be widespread use in thoracic surgery, it is essential thoracic surgeons be at the forefront of the eventual safe introduction of ML to the clinic and operating room.

Characterization of Outcomes by Surgical Management of Lung Neuroendocrine Tumors Associated With Cushing Syndrome.

Importance: Ectopic adrenocorticotropic hormone secretion from lung tumors causing Cushing syndrome are associated with high rates of morbidity. Optimal management remains obscure because knowledge is based on rare reports with few patients. Objective: To characterize the outcomes of lung neuroendocrine tumors associated with Cushing syndrome. Design, Setting, and Participants: An observational case series review from 1982 to 2020 was conducted in a single institution referral center. Kaplan-Meier analysis estimated disease-free survival (DFS). Participants underwent curative-intent surgery for a lung neuroendocrine tumor causing Cushing syndrome. Exposures: Lobectomy or pneumonectomy vs sublobar resection. Main Outcomes and Measures: Disease-free survival, disease persistence/recurrence. Results: Of the 68 patients, the median age was 41 years (range, 17-80 years), 42.6% (29 of 68) were male, 81.8% (54 of 66) were White, with a mean follow-up after surgery was 16 months (range, 0.1-341 months). Lobectomy was the most common procedure (48 of 68 [70.6%]), followed by wedge resection (16 of 68 [23.5%]) and segmentectomy (3 of 68 [4.4%]). Video-assisted thoracoscopic surgery was performed in 19 of 68 (27.9%) of patients. Surgical morbidity was 19.1% (13 of 68), and perioperative mortality was 1.5% (1 of 68). Lymph node positivity was 37% (22 of 59) when evaluable. The overall incidence of persistence/recurrence was 16.2% (11 of 68) with a median time to recurrence of 55 months (range, 18-152 months). The median DFS was reached in 12.7 years (0.1-334 months). There were no statistical differences in DFS based on tumor size, stage (8th edition TNM), whether full systematic lymphadenectomy was performed or not, nodal status, or surgical approach. Conclusions and Relevance: In this case series study, neuroendocrine pulmonary tumors associated with Cushing syndrome had increased nodal metastasis, higher recurrence, and lower DFS than quiescent bronchopulmonary carcinoid tumors, but many patients experienced favorable outcomes. This observation is underscored by the discordance of TNM-stage classifications vs prognosis. Observing no difference in surgical techniques, the implication may be that a lung-sparing approach could suffice. These results may reflect the intrinsic importance of the hormone physiology instead of the carcinoid biologic factors.

Mouse models for mesothelioma drug discovery and development.

INTRODUCTION: Mesothelioma is an aggressive mesothelial lining tumor. Available drug therapies include chemotherapeutic agents, targeted molecular therapies, and immune system modulators. Mouse models were instrumental in the discovery and evaluation of such therapies, but there is need for improved understanding of the role of inflammation, tumor heterogeneity, mechanisms of carcinogenesis, and the tumor microenvironment. Novel mouse models may provide new insights and drive drug therapy discovery that improves efficacy. AREAS COVERED: This review concerns available mouse models for mesothelioma drug discovery and development including the advantages and disadvantages of each. Gaps in current knowledge of mesothelioma are highlighted, and future directions for mouse model research are considered. EXPERT OPINION: Soon, CRISPR-Cas gene-editing will improve understanding of mesothelioma mechanisms foundational to the discovery and testing of efficacious therapeutic targets. There are at least two likely areas of upcoming methodology development. One is concerned with precise modeling of inflammation - is it a causal process whereby inflammatory signals contribute to tumor initiation, or is it a secondary passenger process driven by asbestos exposure effects? The other area of methods improvement regards the availability of humanized immunocompromised mice harboring patient-derived xenografts. Combining human tumors in an environment with human immune cells will enable rapid innovation in immuno-oncology therapeutics.

Patient motivations for non-adherence to lung cancer screening in a military population.

BACKGROUND: Lung cancer remains the leading cause of cancer deaths in the United States, and lung cancer screening has been shown to decrease this mortality. Adherence to lung cancer screening is paramount to realize the mortality benefit, and reported adherence rates vary widely. Few reports address non-adherence to screening, and our study sought to understand the non-compliant patients in our military population. METHODS: This Institutional Review Board approved retrospective review of patients enrolled in our screening program from 2013-2019 identified patients who failed to obtain a subsequent Low Dose CT scan (LDCT) within 15 months of their prior scan. Attempts were made to contact these patients and elucidate motivations for non-adherence via telephone. RESULTS: Of the 242 patients enrolled, 183 (76%) patients were adherent to the protocol. Significant predictors of non-adherence versus adherence were younger age (P=0.008), female sex (P=0.005), and enlisted officer rank (P=0.03). There was no difference with regards to race, smoking status, pack-years, negative screens, lung-RADS level, or nodule size. 31 (52%) non-adherent patients were contacted, and 24 (77%) reported their reason for non-adherence was lack of follow-up for a LDCT. Twenty (64%) were interested in re-enrollment. Of the total screening cohort, 15 interventions were performed, with lung cancer identified in 5 (2%)-a 67% false positive rate. One stage IV lung cancer was found in a non-adherent patient who re-enrolled. CONCLUSIONS: Lack of perceived contact for follow-up was expressed as the primary reason for non-compliance in our screening program. Compliance is critical to the efficacy of any screening modality, and adherence rates to lung cancer screening may be increased through improved contact with patients via multiple avenues (i.e., phone, email, and letter). There is benefit in contacting non-adherent patients as high rates of re-enrollment are possible.

An Example of 3-D Printing for Expeditionary Medicine: The Air Force Retractor.

INTRODUCTION: Future challenges for the U.S. Department of Defense (DoD) include operating in a contested environment against near-peer adversaries. Providing casualty care in an anti-access/area denial (A2/AD) region would be potentially challenged by impaired logistical ability to reconstitute supplies or adapt to evolving needs. Additive manufacturing (AM), also known as 3D printing, offers an ability to regenerate stocked items as well as modify them or even create novel products de novo. MATERIALS AND METHODS: This article discusses relevant technology and applications for deployed medical forces. Additionally, as proof of concept, we outline our design process for a novel 3D-printed surgical retractor. RESULTS: Our "Air Force Retractor" prototype highlights the potential of using AM to create or augment surgical instruments downrange. CONCLUSIONS: This paper encourages further investigation of the use of AM/3D printing downrange to create surgical instruments and medical supplies in austere, A2/AD, and other logistically challenging environments. Not only would this support regeneration of supplies, but also modification and even creation of novel products to adapt to changing needs. If 3D files could be created of common surgical instruments for print on designated resins downrange, and FDA approval obtained, an online library of files could be created for easy access to DoD members across the globe to support our nation's commitment to provide the best possible care for service members any time, any place.

Quantitative imaging features of pretreatment CT predict volumetric response to chemotherapy in patients with colorectal liver metastases.

OBJECTIVES: This study investigates whether quantitative image analysis of pretreatment CT scans can predict volumetric response to chemotherapy for patients with colorectal liver metastases (CRLM). METHODS: Patients treated with chemotherapy for CRLM (hepatic artery infusion (HAI) combined with systemic or systemic alone) were included in the study. Patients were imaged at baseline and approximately 8 weeks after treatment. Response was measured as the percentage change in tumour volume from baseline. Quantitative imaging features were derived from the index hepatic tumour on pretreatment CT, and features statistically significant on univariate analysis were included in a linear regression model to predict volumetric response. The regression model was constructed from 70% of data, while 30% were reserved for testing. Test data were input into the trained model. Model performance was evaluated with mean absolute prediction error (MAPE) and R2. Clinicopatholologic factors were assessed for correlation with response. RESULTS: 157 patients were included, split into training (n = 110) and validation (n = 47) sets. MAPE from the multivariate linear regression model was 16.5% (R2 = 0.774) and 21.5% in the training and validation sets, respectively. Stratified by HAI utilisation, MAPE in the validation set was 19.6% for HAI and 25.1% for systemic chemotherapy alone. Clinical factors associated with differences in median tumour response were treatment strategy, systemic chemotherapy regimen, age and KRAS mutation status (p < 0.05). CONCLUSION: Quantitative imaging features extracted from pretreatment CT are promising predictors of volumetric response to chemotherapy in patients with CRLM. Pretreatment predictors of response have the potential to better select patients for specific therapies. KEY POINTS: • Colorectal liver metastases (CRLM) are downsized with chemotherapy but predicting the patients that will respond to chemotherapy is currently not possible. • Heterogeneity and enhancement patterns of CRLM can be measured with quantitative imaging. • Prediction model constructed that predicts volumetric response with 20% error suggesting that quantitative imaging holds promise to better select patients for specific treatments.

Extended resection of sarcomas involving the mediastinum: a 15-year experience†.

OBJECTIVES: Sarcomas involving the mediastinum are rare and aggressive tumours with a poor prognosis. We reviewed our experience to evaluate outcomes and survival for patients who undergo surgical resection. METHODS: A retrospective review of 35 patients who underwent resection of sarcomas that involved the mediastinum at a single institution over a 15-year period was conducted. RESULTS: Fifteen patients (43%) had sarcomas originating from the great vessels, 15 (43%) the heart and 5 (15%) the chest wall. The most common subtype was leiomyosarcoma (n = 10, 29%). Twenty patients (57%) underwent resection of primary tumours, 11 (31%) of metastases and 4 (11%) of recurrences. Sixteen patients (46%) had Stage IV disease. An R0 resection was achieved in 17 patients (49%). Fourteen (40%) had an R1 resection and 4 (11%) an R2. Thirty cases (85%) were performed on cardiopulmonary bypass (CPB). Twenty-three patients (65%) required resection of adjacent structures. Eight cases (23%) were performed on an urgent basis. Morbidity and mortality were 37 and 6%, respectively. Overall median survival was 33 months (range 7-171 months). Median overall survival for Stage IV disease was 5 months compared with 27 months for all others (P = 0.26). Two-, 5- and 10-year survivals for the cohort were 50, 35 and 25%, respectively. Incomplete resection was associated with poor long-term survival. Median survival was 101 months for R0, 21 months for R1 and 9 months for R2 resection. CONCLUSIONS: CPB and concomitant resection of intrathoracic structures may be required for local tumour control in sarcomas that involve the mediastinum, and should not be considered contraindications to resection as they may offer the best opportunity for prolonged survival in selected cases.

Mycotic Thoracic Aortic Aneurysm After Intravesical Bacillus Calmette-Guérin Treatment.

Bacillus Calmette-Guérin (BCG) is widely used for the treatment of bladder cancer. We report a case of BCG-associated mycotic thoracic aortic aneurysm that was successfully treated with resection and antimycobacterial therapy.

Laparoscopic-assisted Peritoneal Shunt Insertion for Ventriculoperitoneal and Lumboperitoneal Shunt Placement: An Institutional Experience of 53 Consecutive Cases.

PURPOSE: The purpose of this study was to describe operative times, complication rates, and outcomes following laparoscopic placement of the peritoneal catheter in ventriculoperitoneal (VP) and lumboperitoneal (LP) shunt insertion. METHODS: A retrospective review was performed of those who underwent laparoscopic-assisted VP or LP shunt insertion from July 2007 to August 2011. RESULTS: The study included 53 consecutive patients (35 women and 18 men). Mean age was 51 years (range, 16 to 83 y), mean BMI was 27.6 (range, 16 to 54), and 35.8% of the patients had previous abdominal surgery. Mean operative time for VP shunt placement was 68.2 ± 19.0 minutes, and for LP shunt placement 84 ± 12.4 minutes. There were no intraoperative complications, and conversion to minilaparotomy was 0%. There were 2 distal catheter-associated complications. CONCLUSIONS: Laparoscopic-assisted VP/LP shunt placement is associated with a low incidence of distal catheter malfunction. Direct visualization of shunt placement into the peritoneal cavity is a major advantage making it a viable alternative over traditional techniques.