Managing Risk and Quality of AI in Healthcare: Are Hospitals Ready for Implementation?

Artificial intelligence (AI) provides a unique opportunity to help meet the demands of the future healthcare system. However, hospitals may not be well equipped to handle safe and effective development and/or procurement of AI systems. Furthermore, upcoming regulations such as the EU AI Act may enforce the need to establish new management systems, quality assurance and control mechanisms, novel to healthcare organizations. This paper discusses challenges in AI implementation, particularly potential gaps in current management systems (MS), by reviewing the harmonized standard for AI MS, ISO 42001, as part of a gap analysis of a tertiary acute hospital with ongoing AI activities. Examination of the industry agnostic ISO 42001 reveals a technical debt within healthcare, aligning with previous research on digitalization and AI implementation. To successfully implement AI with quality assurance in mind, emphasis should be put on the foundation and structure of the healthcare organizations, including both workforce and data infrastructure.

Can I trust my fake data - A comprehensive quality assessment framework for synthetic tabular data in healthcare.

BACKGROUND: Ensuring safe adoption of AI tools in healthcare hinges on access to sufficient data for training, testing and validation. Synthetic data has been suggested in response to privacy concerns and regulatory requirements and can be created by training a generator on real data to produce a dataset with similar statistical properties. Competing metrics with differing taxonomies for quality evaluation have been proposed, resulting in a complex landscape. Optimising quality entails balancing considerations that make the data fit for use, yet relevant dimensions are left out of existing frameworks. METHOD: We performed a comprehensive literature review on the use of quality evaluation metrics on synthetic data within the scope of synthetic tabular healthcare data using deep generative methods. Based on this and the collective team experiences, we developed a conceptual framework for quality assurance. The applicability was benchmarked against a practical case from the Dutch National Cancer Registry. CONCLUSION: We present a conceptual framework for quality assuranceof synthetic data for AI applications in healthcare that aligns diverging taxonomies, expands on common quality dimensions to include the dimensions of Fairness and Carbon footprint, and proposes stages necessary to support real-life applications. Building trust in synthetic data by increasing transparency and reducing the safety risk will accelerate the development and uptake of trustworthy AI tools for the benefit of patients. DISCUSSION: Despite the growing emphasis on algorithmic fairness and carbon footprint, these metrics were scarce in the literature review. The overwhelming focus was on statistical similarity using distance metrics while sequential logic detection was scarce. A consensus-backed framework that includes all relevant quality dimensions can provide assurance for safe and responsible real-life applications of synthetic data. As the choice of appropriate metrics are highly context dependent, further research is needed on validation studies to guide metric choices and support the development of technical standards.

Outcome after nonoperative treatment of stable Lisfranc injuries. A prospective cohort study.

BACKGROUND: The aim of this study was to evaluate the outcome after nondisplaced and stable Lisfranc injuries. METHODS: 26 patients with injuries to the Lisfranc joint complex detected on CT scans, but without displacement were tested to be stable using a fluoroscopic stress test. The patients were immobilized in a non-weightbearing short leg cast for 6 weeks. The final follow-up was 55 (IQR 53-60) months after injury. RESULTS: All the Lisfranc injuries were confirmed to be stable on follow-up weightbearing radiographs at a minimum of 3 months after injury. Median American Foot and Ankle Society (AOFAS) midfoot score at 1-year follow-up was 89 (IQR 84-97) and at final follow-up 100 (IQR 90-100); The AOFAS score continued to improve after 1-year (P=.005). The median visual analog scale (VAS) for pain was 0 (IQR 0-0) at the final follow-up. One patient had radiological signs of osteoarthritis at 1-year follow-up. CONCLUSION: Stable Lisfranc injuries treated nonoperatively had an excellent outcome in this study with a median follow-up of 55 months. The AOFAS score continued to improve after 1 year.

Iterative reconstruction improves image quality and reduces radiation dose in trauma protocols; A human cadaver study.

BACKGROUND: Radiation-related cancer risk is an object of concern in CT of trauma patients, as these represent a young population. Different radiation reducing methods, including iterative reconstruction (IR), and spilt bolus techniques have been introduced in the recent years in different large scale trauma centers. PURPOSE: To compare image quality in human cadaver exposed to thoracoabdominal computed tomography using IR and standard filtered back-projection (FBP) at different dose levels. MATERIAL AND METHODS: Ten cadavers were scanned at full dose and a dose reduction in CTDIvol of 5 mGy (low dose 1) and 7.5 mGy (low dose 2) on a Siemens Definition Flash 128-slice computed tomography scanner. Low dose images were reconstructed with FBP and Sinogram affirmed iterative reconstruction (SAFIRE) level 2 and 4. Quantitative image quality was analyzed by comparison of contrast-to-noise ratio (CNR) and signal-to-noise ratio (SNR). Qualitative image quality was evaluated by use of visual grading regression (VGR) by four radiologists. RESULTS: Readers preferred SAFIRE reconstructed images over FBP at a dose reduction of 40% (low dose 1) and 56% (low dose 2), with significant difference in overall impression of image quality. CNR and SNR showed significant improvement for images reconstructed with SAFIRE 2 and 4 compared to FBP at both low dose levels. CONCLUSIONS: Iterative image reconstruction, SAFIRE 2 and 4, resulted in equal or improved image quality at a dose reduction of up to 56% compared to full dose FBP and may be used a strong radiation reduction tool in the young trauma population.

Improved image quality in abdominal computed tomography reconstructed with a novel Deep Learning Image Reconstruction technique - initial clinical experience.

BACKGROUND: A novel Deep Learning Image Reconstruction (DLIR) technique for computed tomography has recently received clinical approval. PURPOSE: To assess image quality in abdominal computed tomography reconstructed with DLIR, and compare with standardly applied iterative reconstruction. MATERIAL AND METHODS: Ten abdominal computed tomography scans were reconstructed with iterative reconstruction and DLIR of medium and high strength, with 0.625 mm and 2.5 mm slice thickness. Image quality was assessed using eight visual grading criteria in a side-by-side comparative setting. All series were presented twice to evaluate intraobserver agreement. Reader scores were compared using univariate logistic regression. Image noise and contrast-to-noise ratio were calculated for quantitative analyses. RESULTS: For 2.5 mm slice thickness, DLIR images were more frequently perceived as equal or better than iterative reconstruction across all visual grading criteria (for both DLIR of medium and high strength, p < 0.001). Correspondingly, DLIR images were more frequently perceived as better (as opposed to equal or in favor of iterative reconstruction) for visual reproduction of liver parenchyma, intrahepatic vascular structures as well as overall impression of image noise and texture (p < 0.001). This improved image quality was also observed for 0.625 mm slice images reconstructed with DLIR of high strength when directly comparing to traditional iterative reconstruction in 2.5 mm slices. Image noise was significantly lower and contrast-to-noise ratio measurements significantly higher for images reconstructed with DLIR compared to iterative reconstruction (p < 0.01). CONCLUSIONS: Abdominal computed tomography images reconstructed using a DLIR technique shows improved image quality when compared to standardly applied iterative reconstruction across a variety of clinical image quality criteria.

Facet Arthropathy Following Disc Replacement Versus Rehabilitation: A Prospective Study With 8-Year Follow-Up.

STUDY DESIGN: A prospective study of patients originally randomized to total disc replacement (TDR) or multidisciplinary rehabilitation. OBJECTIVE: To assess the long-term development of facet arthropathy (FA) after TDR versus nonoperative treatment, and to analyze the association between FA and clinical outcome. SUMMARY OF BACKGROUND DATA: FA may appear or increase following TDR, but the natural course of FA is unclear, and no previous study has evaluated the long-term development of FA following TDR compared with nonoperative treatment. METHODS: The study included 126 patients with chronic low back pain and degenerative changes in the lumbar intervertebral discs. The patients underwent pretreatment and 8-year follow-up magnetic resonance imaging (MRI) and 8-year follow-up computed tomography (CT) of the lumbar spine. The primary outcome measure was FA development (yes/no) on MRI at index level L4/L5 or L5/S1, defined as increased FA grade value from pretreatment to follow-up according to Weishaupt grading system. Secondary outcomes included the association between FA (on MRI and CT) and Oswestry Disability Index (ODI) or back pain as well as reoperations. RESULTS: Increased index level FA grade was more frequent after TDR versus nonoperative treatment (36%, 25/69 vs. 2%, 1/57 of patients, P < 0.001), but was not related to change in ODI or back pain. At follow-up, index level FA grades were higher after TDR versus nonoperative treatment (odds ratio 4.0 MRI and 5.9 CT), but were not related to ODI less than or equal to 22. Four patients (6%) treated with TDR and no patients treated nonoperatively were operated for lateral recess stenosis with posterior decompression at the index level during follow-up. CONCLUSION: Index level FA development was more likely after TDR compared with nonoperative treatment but was not associated with the 8-year clinical outcome. Index level FA may have contributed to reoperations in the TDR group. LEVEL OF EVIDENCE: 2.

Temporary Bridge Plating vs Primary Arthrodesis of the First Tarsometatarsal Joint in Lisfranc Injuries: Randomized Controlled Trial.

BACKGROUND: Unstable Lisfranc injuries are best treated with anatomic reduction and stable fixation. There are controversies regarding which type of stabilization is best. In the present study, we compared primary arthrodesis of the first tarsometatarsal (TMT) joint to temporary bridge plating in unstable Lisfranc injuries. METHODS: Forty-eight patients with Lisfranc injuries were included and followed for 2 years. Twenty-four patients were randomized to primary arthrodesis (PA) of the medial 3 TMT joints, whereas 24 patients were randomized to temporary bridge plate (BP) over the first TMT joint and primary arthrodesis of the second and third TMT joints. The main outcome parameter was the American Orthopaedic Foot & Ankle Society (AOFAS) midfoot scale and the secondary outcome parameters were the 36-Item Short Form Health Survey (SF-36) and visual analog scale for pain (VAS pain). Computed tomography (CT) scans pre- and postoperatively were obtained. Radiographs were obtained at follow-ups. Pedobarographic examination was performed at the 2-year follow-up. Twenty-two of 24 patients in the PA and 23/24 in the BP group completed the 2-year follow-up. RESULTS: The mean AOFAS midfoot score 2 years postoperatively was 89 (SD 9) in the PA group and 85 (SD 15) in the BP group (P = .32). There were no significant differences between the groups with regard to SF-36 or VAS pain scores. The alignment of the first metatarsal was better in the BP group than in the PA group measured by the anteroposterior Meary angle (P = .04). The PA group had a reduced peak pressure under the fifth metatarsal (P = .047). In the BP group, 11/24 patients had radiologic signs of osteoarthritis in the first TMT joint. CONCLUSION: Both treatment groups had good outcome scores. The first metatarsal was better aligned in the BP group; however, there was a high incidence of radiographic osteoarthritis in this group. LEVEL OF EVIDENCE: Therapeutic level I, prospective randomized controlled study.

Lisfranc injuries: Incidence, mechanisms of injury and predictors of instability.

BACKGROUND: In Lisfranc injuries the stability of the tarsometatarsal joints guides the treatment of the injury. Determining the stability, especially in the subtle Lisfranc injuries, can be challenging. The purpose of this study was to identify incidence, mechanisms of injury and predictors for instability in Lisfranc injuries. METHODS: Eighty-four Lisfranc injuries presenting at Oslo University Hospital between September 2014 and August 2015 were included. The diagnosis was based on radiologically verified injuries to the tarsometatarsal joints. Associations between radiographic findings and stability were examined. RESULTS: The incidence of Lisfranc injuries was 14/100,000 person-years, and only 31% were high-energy injuries. The incidence of unstable injuries was 6/100,000 person-years, and these were more common in women than men (P = 0.016). Intraarticular fractures in the two lateral tarsometatarsal joints increased the risk of instability (P = 0.007). The height of the second tarsometatarsal joint was less in the unstable injuries than in the stable injuries (P = 0.036). CONCLUSION: The incidence of Lisfranc injuries in the present study is higher than previously published. The most common mechanism of injury is low-energy trauma. Intraarticular fractures in the two lateral tarsometatarsal joints, female gender and shorter second tarsometatarsal joint height increase the risk of an unstable injury. LEVEL OF EVIDENCE: Level III, cross-sectional study.

Lumbar total disc replacement: predictors for long-term outcome.

PURPOSE: We aimed to identify patient characteristics associated with favourable long-term outcomes after lumbar total disc replacement (TDR). METHODS: We analysed a cohort of 82 patients with degenerative disc and chronic low back pain (LBP) who were treated with TDR and originally participated in a randomised trial comparing TDR and multidisciplinary rehabilitation. Potential predictors were measured at baseline, and the outcomes assessed 8 years after they received allocated treatment. Outcome measures were dichotomised according to whether the participants achieved a clinically important functional improvement (15 points or more on the Oswestry Disability Index, ODI) (primary outcome) and whether they were employed at 8-year follow-up (secondary outcome). Associations between potential predictors and outcomes were modelled using logistic regression. For the secondary outcome, the results were also organised in a prediction matrix and expressed as probabilities. RESULTS: For 71 patients treated with TDR according to protocol, the follow-up time was 8 years. For a subgroup of 11 patients randomised to rehabilitation who crossed over and received TDR, the median postoperative follow-up time was 72 (range 41-88) months. Of all assessed baseline variables, only presence of Modic changes (type 1 and/or 2) was statistically significantly associated with an improvement of ≥ 15 ODI points. The probability of employment at 8-year follow-up was 1% for patients with ≥ 1 year of sick leave, comorbidity, ODI ≥ 50 and ≤ 9 years of education prior to treatment, and 87% for patients with < 1 year of sick leave, no comorbidity, ODI < 50 and higher education. CONCLUSIONS: Patients with Modic changes prior to the TDR surgery were more likely to report a clinically important functional improvement at long-term follow-up. Comorbidity, low level of education, long-term sick leave and high ODI score at baseline were associated with unemployment at long-term follow-up.

Evaluation of focal cartilage lesions of the knee using MRI T2 mapping and delayed Gadolinium Enhanced MRI of Cartilage (dGEMRIC).

BACKGROUND: Assessment of degenerative changes of the cartilage is important in knee cartilage repair surgery. Magnetic Resonance Imaging (MRI) T2 mapping and delayed Gadolinium Enhanced MRI of Cartilage (dGEMRIC) are able to detect early degenerative changes. The hypothesis of the study was that cartilage surrounding a focal cartilage lesion in the knee does not possess degenerative changes. METHODS: Twenty-eight consecutive patients included in a randomized controlled trial on cartilage repair were evaluated using MRI T2 mapping and dGEMRIC before cartilage treatment was initiated. Inclusion was based on disabling knee problems (Lysholm score of ≤ 75) due to an arthroscopically verified focal femoral condyle cartilage lesion. Furthermore, no major malalignments or knee ligament injuries were accepted. Mean patient age was 33 ± 9.6 years, and the mean duration of knee symptoms was 49 ± 60 months. The MRI T2 mapping and the dGEMRIC measurements were performed at three standardized regions of interest (ROIs) at the medial and lateral femoral condyle, avoiding the cartilage lesion RESULTS: The MRI T2 mapping of the cartilage did not demonstrate significant differences between condyles with or without cartilage lesions. The dGEMRIC results did not show significantly lower values of the affected condyle compared with the opposite condyle and the contra-lateral knee in any of the ROIs. The intraclass correlation coefficient (ICC) of the dGEMRIC readings was 0.882. CONCLUSION: The MRI T2 mapping and the dGEMRIC confirmed the arthroscopic findings that normal articular cartilage surrounded the cartilage lesion, reflecting normal variation in articular cartilage quality. STUDY IDENTIFIER: NCT00885729 , registered April 17 2009.

Efficacy of strength and aerobic exercise on patient-reported outcomes and structural changes in patients with knee osteoarthritis: study protocol for a randomized controlled trial.

BACKGROUND: Despite an extensive literature on treatment interventions for patients with knee osteoarthritis, studies comparing the efficacy of different exercise interventions and living the life as usual on quality of life, cartilage quality and cost-effectiveness are lacking. The aim of the present study is to compare the efficacy of two different exercise programs compared to a control group in individuals with established radiographic and symptomatic knee osteoarthritis on self-reported knee-related quality of life, knee pain, physical function, and cartilage quality. METHODS/DESIGN: A three-armed randomized controlled trial involving two exercise interventions and a control group of individuals doing as they usually do is described. The patients will have mild to moderate radiographic osteoarthritis according to the Kellgren and Lawrence classification (grade 2-3), and fulfill the American College of Rheumatology clinical criteria, be aged between 45 and 65 years, and have no other serious physical or mental illnesses. The patients will be randomly allocated to a strength exercise group; a cycling group, or a control group. The primary outcome is the Knee injury and Osteoarthritis Outcome Score knee-related quality of life subscale. Secondary outcomes include all five Knee Injury and Osteoarthritis Outcome Score subscales, morphological evaluation of cartilage including focal thickness, subchondral bone marrow edema, proteoglycan content and collagen degradation (measured using magnetic resonance imaging clinical sequences, T2 mapping and T1ρ), specific serum biomarkers, isokinetic muscle strength, maximal oxygen uptake, quality of life (EuroQol 5D), and self-efficacy (Arthritis Self-Efficacy Scale). A sample size calculation on the primary outcome showed that 207 individuals, 69 in each group, is needed to detect a clinically relevant difference of 10 points with 80% power and a significance level of 5%. Assessments will be conducted at baseline, 14 weeks, 1 year and 2 years post-randomization. The interventions will be a 14 weeks exercise program. DISCUSSION: Although exercise therapy has been found to be effective in knee osteoarthritis, the knowledge of the underlying mechanisms for why exercise works is lacking. This study will contribute with knowledge on the efficacy of strength exercise versus cycling on patient-reported outcomes, cartilage quality and cost-effectiveness. TRIAL REGISTRATION: Clinicaltrial.gov Identifier: NCT01682980.