Comparison of evaluation metrics of deep learning for imbalanced imaging data in osteoarthritis studies.

PURPOSE: To compare the evaluation metrics for deep learning methods that were developed using imbalanced imaging data in osteoarthritis studies. MATERIALS AND METHODS: This retrospective study utilized 2996 sagittal intermediate-weighted fat-suppressed knee MRIs with MRI Osteoarthritis Knee Score readings from 2467 participants in the Osteoarthritis Initiative study. We obtained probabilities of the presence of bone marrow lesions (BMLs) from MRIs in the testing dataset at the sub-region (15 sub-regions), compartment, and whole-knee levels based on the trained deep learning models. We compared different evaluation metrics (e.g., receiver operating characteristic (ROC) and precision-recall (PR) curves) in the testing dataset with various class ratios (presence of BMLs vs. absence of BMLs) at these three data levels to assess the model's performance. RESULTS: In a subregion with an extremely high imbalance ratio, the model achieved a ROC-AUC of 0.84, a PR-AUC of 0.10, a sensitivity of 0, and a specificity of 1. CONCLUSION: The commonly used ROC curve is not sufficiently informative, especially in the case of imbalanced data. We provide the following practical suggestions based on our data analysis: 1) ROC-AUC is recommended for balanced data, 2) PR-AUC should be used for moderately imbalanced data (i.e., when the proportion of the minor class is above 5% and less than 50%), and 3) for severely imbalanced data (i.e., when the proportion of the minor class is below 5%), it is not practical to apply a deep learning model, even with the application of techniques addressing imbalanced data issues.

Does a Patient-centered Educational Intervention Affect African-American Access to Knee Replacement? A Randomized Trial.

BACKGROUND: A TKA is the most effective and cost-effective surgical option for moderate to severe osteoarthritis (OA) of the knee. Yet, black patients are less willing to undergo knee replacement surgery than white patients. Decision aids help people understand treatment options and consider the personal importance of possible benefits and harms of treatments, including TKA. QUESTIONS/PURPOSES: We asked: (1) Does a patient-centered intervention consisting of a decision aid for knee OA and motivational interviewing improve the proportion of referrals of blacks with knee OA to orthopaedic surgery? (2) Does the intervention increase patients' willingness to undergo TKA? METHODS: Adults who self-identified as black who were at least 50 years old with moderate to severe knee OA were enrolled from urban primary care clinics in a two-group randomized, controlled trial. A total of 1253 patients were screened for eligibility, and 760 were excluded for not meeting inclusion criteria, declining to participate, or other reasons. Four hundred ninety-three patients were randomized and completed the intervention; three had missing referral data at followup. The mean age of the patients was 61 years, and 51% were women. The majority had an annual household income less than USD 15,000. Participants in the treatment group were shown a decision-aid video and had a brief session with a trained counselor in motivational interviewing. Participants in the control group received an educational booklet about OA that did not mention joint replacement. The two groups had comparable demographic and socioeconomic characteristics. The primary outcome was referral to orthopaedic surgery 12 months after treatment exposure. Receipt of referral was defined as the receipt of a recommendation or prescription from a primary care provider for orthopaedic evaluation. The secondary outcome was change in patient willingness to undergo TKA based on patient self-report. RESULTS: The odds of receiving a referral to orthopaedic surgery did not differ between the two study groups (36%, 90 of 253 of the control group; 32%, 76 of 240 of the treatment group; odds ratio [OR], 0.81; 95% CI, 0.56-1.18; p = 0.277). At 2 weeks followup, there was no difference between the treatment and the control groups in terms of increased willingness to consider TKA relative to baseline (34%, 67 of 200 patients in the treatment group; 33%, 68 of 208 patients in the control group; OR, 1.06; p = 0.779). At 12 months followup, the percent increase in willingness to undergo TKA still did not differ between patients in the treatment and control groups (29%, 49 of 174 in the treatment group; 27%, 51 of 191 in the control group; OR, 1.10; p = 0.679). CONCLUSION: A combination decision aid and motivational interviewing strategy was no better than an educational pamphlet in improving patients' preferences toward joint replacement surgery for knee OA. The type of intervention treatment also did not affect access to surgical evaluation. Other tools that target patient knowledge, beliefs, and attitudes regarding surgical treatments for OA may be further developed and tested in the future. LEVEL OF EVIDENCE: Level I, therapeutic study.

Semiquantitative Imaging Biomarkers of Knee Osteoarthritis Progression: Data From the Foundation for the National Institutes of Health Osteoarthritis Biomarkers Consortium.

OBJECTIVE: To determine the association between changes in semiquantitative magnetic resonance imaging (MRI) biomarkers over 24 months and radiographic and pain progression over 48 months in knees with mild-to-moderate osteoarthritis (OA). METHODS: We undertook a nested case-control study as part of the Foundation for the National Institutes of Health Biomarkers Consortium Project. We used multivariable logistic regression models to examine the association between change over 24 months in semiquantitative MRI markers and radiographic and pain progression in knee OA. MRIs were read according to the MRI OA Knee Score system. We focused on changes in cartilage, osteophytes, meniscus, bone marrow lesions, Hoffa-synovitis, and effusion-synovitis. RESULTS: The most parsimonious model included changes in cartilage thickness and surface area, effusion-synovitis, Hoffa-synovitis, and meniscal morphology (C statistic 0.740). Compared with no worsening, worsening in cartilage thickness in ≥3 subregions was associated with 2.8-fold (95% confidence interval [95% CI] 1.3-5.9) greater odds of being a case, and worsening in cartilage surface area in ≥3 subregions was associated with 2.4-fold (95% CI 1.3-4.4) greater odds of being a case. Worsening of meniscal morphology in any region was associated with 2.2-fold (95% CI 1.3-3.8) greater odds of being a case. Worsening effusion-synovitis and Hoffa-synovitis were also associated with a greater odds of being a case (odds ratios 2.7 and 2.0, respectively). CONCLUSION: Twenty-four-month changes in cartilage thickness, cartilage surface area, effusion-synovitis, Hoffa-synovitis, and meniscal morphology were independently associated with OA progression, suggesting that these factors may serve as efficacy biomarkers in clinical trials of disease-modifying interventions for knee OA.

Pattern of joint damage in persons with knee osteoarthritis and concomitant ACL tears.

Anterior cruciate ligament (ACL) tears are known to be a risk factor for incident knee osteoarthritis (OA). At the present time, it is unknown whether an incidental ACL tear in those with established knee OA alters the pattern of synovial joint damage. Therefore, our aim was to assess whether ACL tears in persons with knee OA are associated with specific patterns of cartilage loss, meniscal degeneration, and bone marrow lesion (BML) location. We included 160 participants from the progression subcohort of the Osteoarthritis Initiative (OAI) Study, an ongoing 4-year, multicenter study, focusing on knee OA. Regional cartilage morphometry measures including cartilage volume (mm(3)), denuded area, normalized cartilage volume, bone surface area, as well as location of meniscal pathology and BMLs in index knees on the same side were compared between those with and without ACL tears. Of the 160 subjects (51% women, age 62.1 (±9.9), BMI 30.3 (±4.7) kg/m(2)), 14.4% had an ACL tear. After adjusting for age, BMI and gender participants with ACL tears had significantly greater cartilage volume in the posterior lateral femur (P = 0.04) and the central medial tibia (0.001) compared to those without ACL tears. Normalized cartilage volume was not different between those with and without ACL tears. In addition, individuals with ACL tears had significantly larger bone surface areas in the medial tibia (P = 0,006), the central medial tibia (P = 0.008), the posterior lateral femur (P = 0.004), and the posterior medial femur (P = 0.04). Furthermore, participants with ACL tears showed significantly more meniscal derangement in the lateral posterior horn (P = 0.019) and significantly more BMLs in the lateral femur (P = 0.0025). We found clear evidence of predominant lateral tibiofemoral involvement, with OA-associated findings on MRI, including increased denuded area and bone surface area, BMLs, and meniscal derangement in knees of individuals with ACL tears compared to those without.

The association of osteoarthritis risk factors with localized, regional and diffuse knee pain.

OBJECTIVE: To identify determinants of different patterns of knee pain with a focus on risk factors for knee osteoarthritis (OA). DESIGN: The Knee Pain Map is an interviewer-administered assessment that asks subjects to characterize their knee pain as localized, regional, or diffuse. A total of 2677 participants from the Osteoarthritis Initiative were studied. We used multinomial logistic regression to examine the relationship between risk factors for OA and knee pain patterns. We examined the bivariate and multivariate relationships of knee pain pattern with age, body mass index (BMI), sex, race, family history of total joint replacement, knee injury, knee surgery, and hand OA. RESULTS: We compared 2462 knees with pain to 1805 knees without pain. In the bivariate analysis, age, sex, BMI, injury, surgery, and hand OA were associated with at least one pain pattern. In the multivariate model, all of these variables remained significantly associated with at least one pattern. When compared to knees without pain, higher BMI, injury, and surgery were associated with all patterns. BMI had its strongest association with diffuse pain. Older age was less likely to be associated with localized pain while female sex was associated with regional pain. CONCLUSIONS: We have shown that specific OA risk factors are associated with different knee pain patterns. Better understanding of the relationship between OA risk factors and knee pain patterns may help to characterize the heterogeneous subsets of knee OA.

Co-morbid medical and psychiatric illness and substance abuse in HCV-infected and uninfected veterans.

Comorbidities may affect the decision to treat chronic hepatitis C virus (HCV) infection. We undertook this study to determine the prevalence of these conditions in the HCV-infected persons compared with HCV-uninfected controls. Demographic and comorbidity data were retrieved for HCV-infected and -uninfected subjects from the VA National Patient Care Database using ICD-9 codes. Logistic regression was used to determine the odds of comorbid conditions in the HCV-infected subjects. HCV-uninfected controls were identified matched on age, race/ethnicity and sex. We identified 126 926 HCV-infected subjects and 126 926 controls. The HCV-infected subjects had a higher prevalence of diabetes, anaemia, hypertension, chronic obstructive pulmonary disease (COPD)/asthma, cirrhosis, hepatitis B and cancer, but had a lower prevalence of coronary artery disease and stroke. The prevalence of all psychiatric comorbidities and substance abuse was higher in the HCV-infected subjects. In the HCV-infected persons, the odds of being diagnosed with congestive heart failure, diabetes, anaemia, hypertension, COPD/asthma, cirrhosis, hepatitis B and cancer were higher, but lower for coronary artery disease and stroke. After adjusting for alcohol and drug abuse and dependence, the odds of psychiatric illness were not higher in the HCV-infected persons. The prevalence and patterns of comorbidities in HCV-infected veterans are different from those in HCV-uninfected controls. The association between HCV and psychiatric diagnoses is at least partly attributable to alcohol and drug abuse and dependence. These factors should be taken into account when evaluating patients for treatment and designing new intervention strategies.

Elderly hospitalized patients with diastolic heart failure: lack of gender and ethnic differences in 18-month mortality rates.

BACKGROUND: Racial and gender differences in mortality rates have been reported for patients with systolic heart failure. Relatively little is known regarding diastolic heart failure prognosis. METHODS: Our sample consisted of 1058 patients 65 years of age or older who were admitted to 30 hospitals in Northeastern Ohio with a principal diagnosis of heart failure and a left ventricular ejection fraction of >/=50% by echocardiogram. RESULTS: Of the 1058 patients with diastolic heart failure (13% African American and 87% white), African Americans and whites were comparable with respect to history of angina, stroke, being on dialysis, and alcohol use; the proportion of male patients was also comparable. The African American to white adjusted odds ratio for 18-month mortality (all cause) was 1.03 (0.66-1.59). For men versus women (30% vs 70%), the above-mentioned comorbidities were comparable, except women were more likely to have a do not resuscitate status (16% vs 7.3%; p =.000) and to be older (79.5 +/- 8 vs 77 +/- 7; p =.000). Men were more likely to have a history of tobacco use (30% vs 14%; p =.000) and alcohol use (36% vs 15%; p =.000), and a higher serum creatinine level (1.7 +/- 1.2 vs 1.4 +/- 1.1; p =.001). The men to women adjusted odds ratio for 18-month mortality (all cause) was 1.06 (0.76-1.46). CONCLUSION: In this cohort of elderly patients admitted with diastolic heart failure, there were no ethnic or gender differences in 18-month mortality rates.