Population-wide evaluation of artificial intelligence and radiologist assessment of screening mammograms.

OBJECTIVES: To validate an AI system for standalone breast cancer detection on an entire screening population in comparison to first-reading breast radiologists. MATERIALS AND METHODS: All mammography screenings performed between August 4, 2014, and August 15, 2018, in the Region of Southern Denmark with follow-up within 24 months were eligible. Screenings were assessed as normal or abnormal by breast radiologists through double reading with arbitration. For an AI decision of normal or abnormal, two AI-score cut-off points were applied by matching at mean sensitivity (AIsens) and specificity (AIspec) of first readers. Accuracy measures were sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and recall rate (RR). RESULTS: The sample included 249,402 screenings (149,495 women) and 2033 breast cancers (72.6% screen-detected cancers, 27.4% interval cancers). AIsens had lower specificity (97.5% vs 97.7%; p < 0.0001) and PPV (17.5% vs 18.7%; p = 0.01) and a higher RR (3.0% vs 2.8%; p < 0.0001) than first readers. AIspec was comparable to first readers in terms of all accuracy measures. Both AIsens and AIspec detected significantly fewer screen-detected cancers (1166 (AIsens), 1156 (AIspec) vs 1252; p < 0.0001) but found more interval cancers compared to first readers (126 (AIsens), 117 (AIspec) vs 39; p < 0.0001) with varying types of cancers detected across multiple subgroups. CONCLUSION: Standalone AI can detect breast cancer at an accuracy level equivalent to the standard of first readers when the AI threshold point was matched at first reader specificity. However, AI and first readers detected a different composition of cancers. CLINICAL RELEVANCE STATEMENT: Replacing first readers with AI with an appropriate cut-off score could be feasible. AI-detected cancers not detected by radiologists suggest a potential increase in the number of cancers detected if AI is implemented to support double reading within screening, although the clinicopathological characteristics of detected cancers would not change significantly. KEY POINTS: • Standalone AI cancer detection was compared to first readers in a double-read mammography screening population. • Standalone AI matched at first reader specificity showed no statistically significant difference in overall accuracy but detected different cancers. • With an appropriate threshold, AI-integrated screening can increase the number of detected cancers with similar clinicopathological characteristics.

Salivary duct carcinoma: a Danish national study.

BACKGROUND: To present the first national series of salivary duct carcinoma patients, including survival rates and an analysis of prognostic factors. METHODS: By merging three Danish nationwide registries that encompass an entire population, 34 patients diagnosed with salivary duct carcinoma from 1990 to 2005 were identified. Histological slides were reviewed, and data concerning demographics, tumour site, clinical stage, treatment profiles and follow-up were retrieved. Survival estimates and prognostic factors were evaluated by comparing Kaplan-Meier plots using the Mantel-Haenszel log-rank test. RESULTS: Salivary duct carcinoma showed an incidence of 0.04/100.000 inhabitants/year. Distant recurrence was seen in 52% of patients. Five-year overall survival, disease-specific survival and recurrence-free survival were 32%, 42% and 35%, respectively. Univariate analyses suggested that overall stage (III/IV) and vascular invasion have a negative impact on all survival measures. Involved resection margins correlated with a poorer overall survival and disease-specific survival, whereas adjuvant radiotherapy improved overall survival and recurrence-free survival. CONCLUSIONS: Salivary duct carcinoma incidence averages to two episodes per year in the entire Kingdom of Denmark. With half of patients in this study experiencing distant recurrences and only a third surviving at 5 years, prognosis is dismal. Advanced overall stage, vascular invasion and involved resection margins all seem to correlate with a poorer survival, while adjuvant radiotherapy significantly improved outcome. Extensive T-site surgery, neck dissection and adjuvant radiotherapy are therefore recommended.

Association of Antithrombotic Drug Use With Incident Intracerebral Hemorrhage Location.

BACKGROUND AND OBJECTIVES: Few population-based studies have assessed associations between the use of antithrombotic (platelet antiaggregant or anticoagulant) drugs and location-specific risks of spontaneous intracerebral hemorrhage (s-ICH). In this study, we estimated associations between antithrombotic drug use and the risk of lobar vs nonlobar incident s-ICH. METHODS: Using Danish nationwide registries, we identified cases in the Southern Denmark Region of first-ever s-ICH in patients aged 50 years or older between 2009 and 2018. Each verified case was classified as lobar or nonlobar s-ICH and matched to controls in the general population by age, sex, and calendar year. Prior antithrombotic use was ascertained from a nationwide prescription registry. We calculated odds ratios (aORs) for associations between the use of clopidogrel, aspirin, direct oral anticoagulants (DOACs) or vitamin K antagonists (VKA), and lobar and nonlobar ICH in conditional logistic regression analyses that were adjusted for potential confounders. RESULTS: A total of 1,040 cases of lobar (47.9% men, mean age [SD] 75.2 [10.7] years) and 1,263 cases of nonlobar s-ICH (54.2% men, mean age 73.6 [11.4] years) were matched to 41,651 and 50,574 controls, respectively. A stronger association with lobar s-ICH was found for clopidogrel (cases: 7.6%, controls: 3.5%; aOR 3.46 [95% CI 2.45-4.89]) vs aspirin (cases: 22.9%, controls: 20.4%; aOR 2.14 [1.74-2.63; p = 0.019). Corresponding estimates for nonlobar s-ICH were not different between clopidogrel (cases: 5.4%, controls: 3.4%; aOR 2.44 [1.71-3.49]) and aspirin (cases: 20.7%, controls: 19.2%; aOR 1.77 [1.47-2.15]; p = 0.12). VKA use was associated with higher odds of both lobar (cases: 14.3%, controls: 6.1%; aOR 3.66 [2.78-4.80]) and nonlobar (cases: 15.4%, controls: 5.5%; aOR 4.62 [3.67-5.82]) s-ICH. The association of DOAC use with lobar s-ICH (cases: 3.5%, controls: 2.7%; aOR 1.66 [1.02-2.70]) was weaker than that of VKA use (p = 0.006). Corresponding estimates for nonlobar s-ICH were not different between DOACs (cases: 5.1%, controls: 2.4%; aOR 3.44 [2.33-5.08]) and VKAs (p = 0.20). DISCUSSION: Antithrombotics were associated with higher risks of s-ICH, but the strength of the associations varied by s-ICH location and drug, which may reflect differences in the cerebral microangiopathies associated with lobar vs nonlobar hemorrhages and the mechanisms of drug action.

Artificial intelligence for MRI stroke detection: a systematic review and meta-analysis.

OBJECTIVES: This systematic review and meta-analysis aimed to assess the stroke detection performance of artificial intelligence (AI) in magnetic resonance imaging (MRI), and additionally to identify reporting insufficiencies. METHODS: PRISMA guidelines were followed. MEDLINE, Embase, Cochrane Central, and IEEE Xplore were searched for studies utilising MRI and AI for stroke detection. The protocol was prospectively registered with PROSPERO (CRD42021289748). Sensitivity, specificity, accuracy, and area under the receiver operating characteristic (ROC) curve were the primary outcomes. Only studies using MRI in adults were included. The intervention was AI for stroke detection with ischaemic and haemorrhagic stroke in separate categories. Any manual labelling was used as a comparator. A modified QUADAS-2 tool was used for bias assessment. The minimum information about clinical artificial intelligence modelling (MI-CLAIM) checklist was used to assess reporting insufficiencies. Meta-analyses were performed for sensitivity, specificity, and hierarchical summary ROC (HSROC) on low risk of bias studies. RESULTS: Thirty-three studies were eligible for inclusion. Fifteen studies had a low risk of bias. Low-risk studies were better for reporting MI-CLAIM items. Only one study examined a CE-approved AI algorithm. Forest plots revealed detection sensitivity and specificity of 93% and 93% with identical performance in the HSROC analysis and positive and negative likelihood ratios of 12.6 and 0.079. CONCLUSION: Current AI technology can detect ischaemic stroke in MRI. There is a need for further validation of haemorrhagic detection. The clinical usability of AI stroke detection in MRI is yet to be investigated. CRITICAL RELEVANCE STATEMENT: This first meta-analysis concludes that AI, utilising diffusion-weighted MRI sequences, can accurately aid the detection of ischaemic brain lesions and its clinical utility is ready to be uncovered in clinical trials. KEY POINTS: There is a growing interest in AI solutions for detection aid. The performance is unknown for MRI stroke assessment. AI detection sensitivity and specificity were 93% and 93% for ischaemic lesions. There is limited evidence for the detection of patients with haemorrhagic lesions. AI can accurately detect patients with ischaemic stroke in MRI.

Prostate cancer metastasis mimicking appendicitis-A rare but important differential diagnosis in PET/CT imaging.

Prostate cancer most commonly metastasizes to lymph nodes, bones, the liver, and the lungs. Prostate cancer carcinomatosis with an affinity for the appendix is not well described in current literature and is usually reported with acute appendicitis as the primary presentation. A 65-year-old male with a history of recurrent prostate cancer presented with an increase in PSA value. 18F-PSMA-1007 PET/CT showed nodular tissue growth and increased PSMA uptake in the prostate, on the appendix, in the umbilicus, and in several intra- and extra pelvic lymph nodes. The patient had no symptomatic complaints at time of referral. Imaging findings of the appendix resembling characteristic findings of acute appendicitis raised doubts about the interpretation of these as inflammatory disease or peritoneal carcinomatosis secondary to prostate cancer. This case demonstrates the importance of correct differentiation between the 2 conditions based on imaging, clinical symptomatology, and patient history to provide proper care in time.

Choice of Assessment and Subsequent Risk of Breast Cancer among Women with False-Positive Mammography Screening.

Women with false-positive mammography screening results have a two- to four-fold higher risk of breast cancer. This study aimed to investigate if the subsequent risk of breast cancer after a false-positive mammography screening is associated with the received diagnostic assessment. The study population consisted of women who underwent false-positive mammography screening from January 2010 to June 2019. They were categorised into seven groups depending on the elements in the assessment (standard care: additional mammography, ultrasound, and if they had a relevant biopsy). Risks of interval cancer, next-round screen-detected cancer, and long-term breast cancer for non-standard care assessments were compared to standard care assessments using Binomial and Cox regression models. We included 44,279 women with a false-positive result. Invasive assessments that lacked an ultrasound or additional mammography were not more associated with an increased risk of subsequent cancers compared to that of 'all three elements'. The few assessments that included 'only ultrasound' or 'only mammography' resulted in higher relative risks of next-round screen-detected cancer of 1.52 (95% CI: 0.93-2.47) and 1.67 (95% CI: 0.54-5.16), respectively, compared to that of standard care. The increased subsequent risk of breast cancer among women with a previous false-positive result was not found to be correlated with the choice of elements in the assessment process.

Breast cancer detection accuracy of AI in an entire screening population: a retrospective, multicentre study.

BACKGROUND: Artificial intelligence (AI) systems are proposed as a replacement of the first reader in double reading within mammography screening. We aimed to assess cancer detection accuracy of an AI system in a Danish screening population. METHODS: We retrieved a consecutive screening cohort from the Region of Southern Denmark including all participating women between Aug 4, 2014, and August 15, 2018. Screening mammograms were processed by a commercial AI system and detection accuracy was evaluated in two scenarios, Standalone AI and AI-integrated screening replacing first reader, with first reader and double reading with arbitration (combined reading) as comparators, respectively. Two AI-score cut-off points were applied by matching at mean first reader sensitivity (AIsens) and specificity (AIspec). Reference standard was histopathology-proven breast cancer or cancer-free follow-up within 24 months. Coprimary endpoints were sensitivity and specificity, and secondary endpoints were positive predictive value (PPV), negative predictive value (NPV), recall rate, and arbitration rate. Accuracy estimates were calculated using McNemar's test or exact binomial test. RESULTS: Out of 272,008 screening mammograms from 158,732 women, 257,671 (94.7%) with adequate image data were included in the final analyses. Sensitivity and specificity were 63.7% (95% CI 61.6%-65.8%) and 97.8% (97.7-97.8%) for first reader, and 73.9% (72.0-75.8%) and 97.9% (97.9-98.0%) for combined reading, respectively. Standalone AIsens showed a lower specificity (-1.3%) and PPV (-6.1%), and a higher recall rate (+ 1.3%) compared to first reader (p < 0.0001 for all), while Standalone AIspec had a lower sensitivity (-5.1%; p < 0.0001), PPV (-1.3%; p = 0.01) and NPV (-0.04%; p = 0.0002). Compared to combined reading, Integrated AIsens achieved higher sensitivity (+ 2.3%; p = 0.0004), but lower specificity (-0.6%) and PPV (-3.9%) as well as higher recall rate (+ 0.6%) and arbitration rate (+ 2.2%; p < 0.0001 for all). Integrated AIspec showed no significant difference in any outcome measures apart from a slightly higher arbitration rate (p < 0.0001). Subgroup analyses showed higher detection of interval cancers by Standalone AI and Integrated AI at both thresholds (p < 0.0001 for all) with a varying composition of detected cancers across multiple subgroups of tumour characteristics. CONCLUSIONS: Replacing first reader in double reading with an AI could be feasible but choosing an appropriate AI threshold is crucial to maintaining cancer detection accuracy and workload.

Association Between Statin Use and Intracerebral Hemorrhage Location: A Nested Case-Control Registry Study.

BACKGROUND AND OBJECTIVES: A causal relationship between statin use and intracerebral hemorrhage (ICH) is uncertain. We hypothesized that an association between long-term statin exposure and ICH risk might vary for different ICH locations. METHODS: We conducted this analysis using linked Danish nationwide registries. Within the Southern Denmark Region (population 1.2 million), we identified all first-ever cases of ICH between 2009 and 2018 in persons aged ≥55 years. Patients with medical record-verified diagnoses were classified as having a lobar or nonlobar ICH and matched for age, sex, and calendar year to general population controls. We used a nationwide prescription registry to ascertain prior statin and other medication use that we classified for recency, duration, and intensity. Using conditional logistic regression adjusted for potential confounders, we calculated adjusted ORs (aORs) and corresponding 95% CIs for the risk of lobar and nonlobar ICH. RESULTS: We identified 989 patients with lobar ICH (52.2% women, mean age 76.3 years) who we matched to 39,500 controls and 1,175 patients with nonlobar ICH (46.5% women, mean age 75.1 years) who we matched to 46,755 controls. Current statin use was associated with a lower risk of lobar (aOR 0.83; 95% CI, 0.70-0.98) and nonlobar ICH (aOR 0.84; 95% CI, 0.72-0.98). Longer duration of statin use was also associated with a lower risk of lobar (<1 year: aOR 0.89; 95% CI, 0.69-1.14; ≥1 year to <5 years aOR 0.89; 95% CI 0.73-1.09; ≥5 years aOR 0.67; 95% CI, 0.51-0.87; p for trend 0.040) and nonlobar ICH (<1 year: aOR 1.00; 95% CI, 0.80-1.25; ≥1 year to <5 years aOR 0.88; 95% CI 0.73-1.06; ≥5 years aOR 0.62; 95% CI, 0.48-0.80; p for trend <0.001). Estimates stratified by statin intensity were similar to the main estimates for low-medium intensity therapy (lobar aOR 0.82; nonlobar aOR 0.84); the association with high-intensity therapy was neutral. DISCUSSION: We found that statin use was associated with a lower risk of ICH, particularly with longer treatment duration. This association did not vary by hematoma location.

Major Cardiovascular Events After Spontaneous Intracerebral Hemorrhage by Hematoma Location.

Importance: Survivors of spontaneous (ie, nontraumatic and with no known structural cause) intracerebral hemorrhage (ICH) have an increased risk of major cardiovascular events (MACEs), including recurrent ICH, ischemic stroke (IS), and myocardial infarction (MI). Only limited data are available from large, unselected population studies assessing the risk of MACEs according to index hematoma location. Objective: To examine the risk of MACEs (ie, the composite of ICH, IS, spontaneous intracranial extra-axial hemorrhage, MI, systemic embolism, or vascular death) after ICH based on ICH location (lobar vs nonlobar). Design, Setting, and Participants: This cohort study identified 2819 patients in southern Denmark (population of 1.2 million) 50 years or older hospitalized with first-ever spontaneous ICH from January 1, 2009, to December 31, 2018. Intracerebral hemorrhage was categorized as lobar or nonlobar, and the cohorts were linked to registry data until the end of 2018 to identify the occurrence of MACEs and separately recurrent ICH, IS, and MI. Outcome events were validated using medical records. Associations were adjusted for potential confounders using inverse probability weighting. Exposure: Location of ICH (lobar vs nonlobar). Main Outcomes and Measures: The main outcomes were MACEs and separately recurrent ICH, IS, and MI. Crude absolute event rates per 100 person-years and adjusted hazard ratios (aHRs) with 95% CIs were calculated. Data were analyzed from February to September 2022. Results: Compared with patients with nonlobar ICH (n = 1255; 680 [54.2%] men and 575 [45.8%] women; mean [SD] age, 73.5 [11.4] years), those with lobar ICH (n = 1034; 495 [47.9%] men and 539 [52.1%] women, mean [SD] age, 75.2 [10.7] years) had higher rates of MACEs per 100 person-years (10.84 [95% CI, 9.51-12.37] vs 7.91 [95% CI, 6.93-9.03]; aHR, 1.26; 95% CI, 1.10-1.44) and recurrent ICH (3.74 [95% CI, 3.01-4.66] vs 1.24 [95% CI, 0.89-1.73]; aHR, 2.63; 95% CI, 1.97-3.49) but not IS (1.45 [95% CI, 1.02-2.06] vs 1.77 [95% CI, 1.34-2.34]; aHR, 0.81; 95% CI, 0.60-1.10) or MI (0.42 [95% CI, 0.22-0.81] vs 0.64 [95% CI, 0.40-1.01]; aHR, 0.64; 95% CI, 0.38-1.09). Conclusions and Relevance: In this cohort study, spontaneous lobar ICH was associated with a higher rate of subsequent MACEs than nonlobar ICH, primarily due to a higher rate of recurrent ICH. This study highlights the importance of secondary ICH prevention strategies in patients with lobar ICH.

[Artificial intelligence for cancer detection in breast cancer screening].

Artificial intelligence (AI) has the potential to increase quality and efficiency of breast cancer screening. Recent studies have provided comparative data on AI versus human performance in cancer detection with encouraging results, and commercially available AI systems are used worldwide as a clinical tool for mammography screening. There are, however, a number of methodological concerns in relation to the evaluation of AI systems. This review discusses these aspects as well as the opportunities and challenges of clinical validation and implementation in breast cancer screening practice.

Polymorphous low-grade adenocarcinoma: A Danish national study.

OBJECTIVES: To present a national series of polymorphous low-grade adenocarcinoma (PLGA) patients, including survival rates and an analysis of prognostic factors. MATERIALS AND METHODS: By merging three Danish nationwide registries, 73 patients diagnosed with PLGA from 1990 to 2005 were identified. Histological slides were reviewed and data concerning demographics, tumour site, clinical stage, treatment profiles and follow-up were retrieved. Survival estimates and prognostic factors were evaluated by comparing Kaplan-Meier plots using the Mantel-Haenszel log-rank test. RESULTS: Of the 73 patients, 47 (64%) were female. Median age was 58years. The most common location was the palate (73%). Median latency was five months. Recurrence was seen in 13% of patients. Overall survival (OS), disease-specific survival (DSS) and recurrence-free survival (RFS) rates after 10years were 73%, 99% and 83%, respectively. Univariate analyses suggested that free resection margins significantly improve RFS. CONCLUSION: PLGA usually has an excellent survival outcome even in cases of advanced stage disease and locoregional recurrence. Primary choice of treatment should be complete surgical excision. Although there is no convincing evidence for the efficacy of adjuvant radiotherapy, it should still be considered, particularly in cases of involved resection margins and advanced stage disease. Late recurrences are common and respond well to salvage therapy.