Development and Evaluation of Machine Learning in Whole-Body Magnetic Resonance Imaging for Detecting Metastases in Patients With Lung or Colon Cancer: A Diagnostic Test Accuracy Study.

OBJECTIVES: Whole-body magnetic resonance imaging (WB-MRI) has been demonstrated to be efficient and cost-effective for cancer staging. The study aim was to develop a machine learning (ML) algorithm to improve radiologists' sensitivity and specificity for metastasis detection and reduce reading times. MATERIALS AND METHODS: A retrospective analysis of 438 prospectively collected WB-MRI scans from multicenter Streamline studies (February 2013-September 2016) was undertaken. Disease sites were manually labeled using Streamline reference standard. Whole-body MRI scans were randomly allocated to training and testing sets. A model for malignant lesion detection was developed based on convolutional neural networks and a 2-stage training strategy. The final algorithm generated lesion probability heat maps. Using a concurrent reader paradigm, 25 radiologists (18 experienced, 7 inexperienced in WB-/MRI) were randomly allocated WB-MRI scans with or without ML support to detect malignant lesions over 2 or 3 reading rounds. Reads were undertaken in the setting of a diagnostic radiology reading room between November 2019 and March 2020. Reading times were recorded by a scribe. Prespecified analysis included sensitivity, specificity, interobserver agreement, and reading time of radiology readers to detect metastases with or without ML support. Reader performance for detection of the primary tumor was also evaluated. RESULTS: Four hundred thirty-three evaluable WB-MRI scans were allocated to algorithm training (245) or radiology testing (50 patients with metastases, from primary 117 colon [n = 117] or lung [n = 71] cancer). Among a total 562 reads by experienced radiologists over 2 reading rounds, per-patient specificity was 86.2% (ML) and 87.7% (non-ML) (-1.5% difference; 95% confidence interval [CI], -6.4%, 3.5%; P = 0.39). Sensitivity was 66.0% (ML) and 70.0% (non-ML) (-4.0% difference; 95% CI, -13.5%, 5.5%; P = 0.344). Among 161 reads by inexperienced readers, per-patient specificity in both groups was 76.3% (0% difference; 95% CI, -15.0%, 15.0%; P = 0.613), with sensitivity of 73.3% (ML) and 60.0% (non-ML) (13.3% difference; 95% CI, -7.9%, 34.5%; P = 0.313). Per-site specificity was high (>90%) for all metastatic sites and experience levels. There was high sensitivity for the detection of primary tumors (lung cancer detection rate of 98.6% with and without ML [0.0% difference; 95% CI, -2.0%, 2.0%; P = 1.00], colon cancer detection rate of 89.0% with and 90.6% without ML [-1.7% difference; 95% CI, -5.6%, 2.2%; P = 0.65]). When combining all reads from rounds 1 and 2, reading times fell by 6.2% (95% CI, -22.8%, 10.0%) when using ML. Round 2 read-times fell by 32% (95% CI, 20.8%, 42.8%) compared with round 1. Within round 2, there was a significant decrease in read-time when using ML support, estimated as 286 seconds (or 11%) quicker ( P = 0.0281), using regression analysis to account for reader experience, read round, and tumor type. Interobserver variance suggests moderate agreement, Cohen κ = 0.64; 95% CI, 0.47, 0.81 (with ML), and Cohen κ = 0.66; 95% CI, 0.47, 0.81 (without ML). CONCLUSIONS: There was no evidence of a significant difference in per-patient sensitivity and specificity for detecting metastases or the primary tumor using concurrent ML compared with standard WB-MRI. Radiology read-times with or without ML support fell for round 2 reads compared with round 1, suggesting that readers familiarized themselves with the study reading method. During the second reading round, there was a significant reduction in reading time when using ML support.

Diagnostic accuracy of whole-body MRI versus standard imaging pathways for metastatic disease in newly diagnosed colorectal cancer: the prospective Streamline C trial.

BACKGROUND: Whole-body MRI (WB-MRI) could be an alternative to multimodality staging of colorectal cancer, but its diagnostic accuracy, effect on staging times, number of tests needed, cost, and effect on treatment decisions are unknown. We aimed to prospectively compare the diagnostic accuracy and efficiency of WB-MRI-based staging pathways with standard pathways in colorectal cancer. METHODS: The Streamline C trial was a prospective, multicentre trial done in 16 hospitals in England. Eligible patients were 18 years or older, with newly diagnosed colorectal cancer. Exclusion criteria were severe systemic disease, pregnancy, contraindications to MRI, or polyp cancer. Patients underwent WB-MRI, the result of which was withheld until standard staging investigations were complete and the first treatment decision made. The multidisciplinary team recorded its treatment decision based on standard investigations, then on the WB-MRI staging pathway (WB-MRI plus additional tests generated), and finally on all tests. The primary outcome was difference in per-patient sensitivity for metastases between standard and WB-MRI staging pathways against a consensus reference standard at 12 months, in the per-protocol population. Secondary outcomes were difference in per-patient specificity for metastatic disease detection between standard and WB-MRI staging pathways, differences in treatment decisions, staging efficiency (time taken, test number, and costs), and per-organ sensitivity and specificity for metastases and per-patient agreement for local T and N stage. This trial is registered with the International Standard Randomised Controlled Trial registry, number ISRCTN43958015, and is complete. FINDINGS: Between March 26, 2013, and Aug 19, 2016, 1020 patients were screened for eligibility. 370 patients were recruited, 299 of whom completed the trial; 68 (23%) had metastasis at baseline. Pathway sensitivity was 67% (95% CI 56 to 78) for WB-MRI and 63% (51 to 74) for standard pathways, a difference in sensitivity of 4% (-5 to 13, p=0·51). No adverse events related to imaging were reported. Specificity did not differ between WB-MRI (95% [95% CI 92-97]) and standard pathways (93% [90-96], p=0·48). Agreement with the multidisciplinary team's final treatment decision was 96% for WB-MRI and 95% for the standard pathway. Time to complete staging was shorter for WB-MRI (median, 8 days [IQR 6-9]) than for the standard pathway (13 days [11-15]); a 5-day (3-7) difference. WB-MRI required fewer tests (median, one [95% CI 1 to 1]) than did standard pathways (two [2 to 2]), a difference of one (1 to 1). Mean per-patient staging costs were £216 (95% CI 211-221) for WB-MRI and £285 (260-310) for standard pathways. INTERPRETATION: WB-MRI staging pathways have similar accuracy to standard pathways and reduce the number of tests needed, staging time, and cost. FUNDING: UK National Institute for Health Research.

Diagnostic accuracy of whole-body MRI versus standard imaging pathways for metastatic disease in newly diagnosed non-small-cell lung cancer: the prospective Streamline L trial.

BACKGROUND: Whole-body magnetic resonance imaging (WB-MRI) could be an alternative to multi-modality staging of non-small-cell lung cancer (NSCLC), but its diagnostic accuracy, effect on staging times, number of tests needed, cost, and effect on treatment decisions are unknown. We aimed to prospectively compare the diagnostic accuracy and efficiency of WB-MRI-based staging pathways with standard pathways in NSCLC. METHODS: The Streamline L trial was a prospective, multicentre trial done in 16 hospitals in England. Eligible patients were 18 years or older, with newly diagnosed NSCLC that was potentially radically treatable on diagnostic chest CT (defined as stage IIIb or less). Exclusion criteria were severe systemic disease, pregnancy, contraindications to MRI, or histologies other than NSCLC. Patients underwent WB-MRI, the result of which was withheld until standard staging investigations were complete and the first treatment decision made. The multidisciplinary team recorded its treatment decision based on standard investigations, then on the WB-MRI staging pathway (WB-MRI plus additional tests generated), and finally on all tests. The primary outcome was difference in per-patient sensitivity for metastases between standard and WB-MRI staging pathways against a consensus reference standard at 12 months, in the per-protocol population. Secondary outcomes were difference in per-patient specificity for metastatic disease detection between standard and WB-MRI staging pathways, differences in treatment decisions, staging efficiency (time taken, test number, and costs) and per-organ sensitivity and specificity for metastases and per-patient agreement for local T and N stage. This trial is registered with the International Standard Randomised Controlled Trial registry, number ISRCTN50436483, and is complete. FINDINGS: Between Feb 26, 2013, and Sept 5, 2016, 976 patients were screened for eligibility. 353 patients were recruited, 187 of whom completed the trial; 52 (28%) had metastasis at baseline. Pathway sensitivity was 50% (95% CI 37-63) for WB-MRI and 54% (41-67) for standard pathways, a difference of 4% (-7 to 15, p=0·73). No adverse events related to imaging were reported. Specificity did not differ between WB-MRI (93% [88-96]) and standard pathways (95% [91-98], p=0·45). Agreement with the multidisciplinary team's final treatment decision was 98% for WB-MRI and 99% for the standard pathway. Time to complete staging was shorter for WB-MRI (13 days [12-14]) than for the standard pathway (19 days [17-21]); a 6-day (4-8) difference. The number of tests required was similar WB-MRI (one [1-1]) and standard pathways (one [1-2]). Mean per-patient costs were £317 (273-361) for WBI-MRI and £620 (574-666) for standard pathways. INTERPRETATION: WB-MRI staging pathways have similar accuracy to standard pathways, and reduce the staging time and costs. FUNDING: UK National Institute for Health Research.

Paroxysmal nocturnal haemoglobinuria (PNH) manifesting on CT as a pathologic segment of small bowel.

Paroxysmal nocturnal haemoglobinuria, a rare, acquired, life-threatening disease of the blood, is characterised by a triad of haemolysis previously believed to occur mainly at night, bone marrow dysfunction, and thrombophilia. Paroxysmal nocturnal haemoglobinuria is customarily regarded to manifest clinically as haemolytic anaemia and haemoglobinuria experienced as reddened urine in the morning, pancytopenia, and thrombosis. We describe a case in which an abnormal segment of small bowel as visualised on computed tomography was the principal sign of the disease process on presentation.

Carpal erosions in children with juvenile idiopathic arthritis: repeatability of a newly devised MR-scoring system.

BACKGROUND: Juvenile idiopathic arthritis (JIA) is characterized by synovial inflammation, with potential risk of developing progressive joint destruction. Personalized state-of-the-art treatment depends on valid markers for disease activity to monitor response; however, no such markers exist. OBJECTIVE: To evaluate the reliability of scoring of carpal bone erosions on MR in children with JIA using two semi-quantitative scoring systems. MATERIALS AND METHODS: A total of 1,236 carpal bones (91 MR wrist examinations) were scored twice by two independent pediatric musculoskeletal radiologists. Bony erosions were scored according to estimated bone volume loss using a 0-4 scale and a 0-10 scale. An aggregate erosion score comprising the sum total carpal bone volume loss was calculated for each examination. RESULTS: The 0-4 scoring system resulted in good intra-reader agreement and moderate to good inter-observer agreement in the assessment of individual bones. Fair and moderate agreement were achieved for inter-reader and intra-reader agreement, respectively, using the 0-10 scale. Intra- and particularly inter-reader aggregate score variability were much less favorable, with wide limits of agreement. CONCLUSION: Further analysis of erosive disease patterns compared with normal subjects is required, and to facilitate the development of an alternative means of quantifying disease.

Magnetic resonance imaging of the immature skeleton.

Magnetic resonance (MR) is unique in its ability to allow assessment of bone marrow, epiphyseal, physeal, and articular cartilage as well as tendons and ligaments. An understanding of skeletal maturation and the accompanying changes on MR is of utmost importance in pediatric radiology. In particular, it is important to recognize the normal spectrum related to ossification and marrow transformation. This review will include a brief description of main indications and common pitfalls in musculoskeletal MR in children. Also, we will focus on the MR appearance of the growing pediatric skeleton on the most commonly used sequences.

MRI assessment of tenosynovitis in children with juvenile idiopathic arthritis: inter- and intra-observer variability.

BACKGROUND: There is sparse knowledge about grading tenosynovitis using MRI. OBJECTIVE: The purpose of this study was to assess the reliability of a tenosynovitis MRI scoring system in juvenile idiopathic arthritis. MATERIALS AND METHODS: Children with juvenile idiopathic arthritis and wrist involvement were enrolled in two paediatric centres, from October 2006 to January 2010. The extensor (compartments II, IV and VI) and flexor tendons were assessed for the presence of tenosynovitis on T1-weighted postcontrast fat-saturated MR images and were scored from 0 (normal) to 2 (moderate to severe) by two observers independently. Intra- and interobserver agreement was assessed. RESULTS: Ninety children (age range: 5-18.5 years) were included, of whom 34 had tenosynovitis involving extensors and 28 had tenosynovitis involving flexors. A total of 360 tendon areas were analysed, of which 114 had tenosynovitis (86/270 extensors and 28/90 flexors). Intra-reader 1 agreement was excellent for the extensors (k = 0.82-0.91) and for the flexors (k = 0.85); intra-reader 2 agreement was moderate to good for the extensors (k = 0.51-0.72) and good for the flexors (k = 0.64). Inter-reader agreement was good for the extensors (k = 0.69-0.73) and moderate for the flexors (k = 0.49). CONCLUSION: The proposed MRI scoring system for the assessment of wrist tenosynovitis in juvenile idiopathic arthritis appears feasible with an observer agreement sufficient for clinical use.

Monobloc and bipartition in craniofacial surgery.

BACKGROUND: The frontofacial monobloc advancement with osteogenic distraction is increasingly used as a surgical treatment for children with complex craniosynostosis-associated syndromes. However, the subfrontal osteotomy cuts to free the facial skeleton from the skull base require extradural retraction of the frontal lobes. The purpose of this study was to determine the frequency and degree of radiologically identifiable frontal lobe changes and whether any such changes affected the patients' outcome. METHODS: The clinical records and preoperative and postoperative computed tomography imaging of all patients undergoing monobloc frontofacial distraction advancement (with or without bipartition) were reviewed. A retrospective medical notes review was undertaken to assess any patient or surgically related factors that might predispose to frontal lobe changes and evaluate outcome from surgery. Where available, magnetic resonance imaging scans were reviewed to compare outcome with that on computed tomography. RESULTS: Fifty cases were identified as suitable for the study. Eighteen patients (36%) had no frontal lobe changes. Thirty-two cases (64%) did have changes that appeared related to the position of maximum retraction during subfrontal osteotomy cuts. There were no changes in the incidence/extent of these changes over time or of any link to the patients' diagnosis, age at surgery, phenotype severity, surgery type, or any surgical or postoperative adverse events. We found no evidence that these changes were responsible for neurologic problems (eg, epilepsy) or reduced cognitive function. CONCLUSIONS: This study reveals a high incidence of frontal lobe changes demonstrable on neuroimaging following the frontofacial monobloc procedure reflecting the retraction points during surgery. Although no postoperative disability was reported, it is clearly important to consider more detailed neuropsychologic testing and review current surgical techniques to ensure that such changes are kept to a minimum.

MRI of the wrist in juvenile idiopathic arthritis: erosions or normal variants? A prospective case-control study.

BACKGROUND: Bony depressions at the wrist resembling erosions are frequently seen on MRI in healthy children. The accuracy of MRI in detecting early bony destruction is therefore questionable. We compared findings on MRI of the wrist in healthy children and those with juvenile idiopathic arthritis (JIA) to investigate markers for true disease. MATERIALS AND METHODS: We compared the number and localisation of bony depressions at the wrist in 85 healthy children and 68 children with JIA, ages 5-15 years. The size of the wrist was assessed from a radiograph of the wrist performed on the same day as the MRI. RESULTS: No significant difference in the number of bony depressions in the carpal bones was seen between healthy children and children with JIA at any age. Depressions are found in similar locations in the two groups, except for a few sites, where bony depressions were seen exclusively in the JIA group, particularly at the CMC joints. The wrist was significantly smaller in children with JIA (P < 0.001). CONCLUSIONS: Using adult scoring systems and standard MR sequences in the assessment of bone destruction in children may lead to overstaging or understaging of disease. At present, standard MRI sequences cannot easily be used for assessment of early signs of erosions in children.

MRI of the wrist in juvenile idiopathic arthritis: proposal of a paediatric synovitis score by a consensus of an international working group. Results of a multicentre reliability study.

BACKGROUND: MRI is a sensitive tool for the evaluation of synovitis in juvenile idiopathic arthritis (JIA). OBJECTIVE: The purpose of this study was to introduce a novel MRI-based score for synovitis in children and to examine its inter- and intraobserver variability in a multi-centre study. MATERIALS AND METHODS: Wrist MRI was performed in 76 children with JIA. On postcontrast 3-D spoiled gradient-echo and fat-suppressed T2-weighted spin-echo images, joint recesses were scored for the degree of synovial enhancement, effusion and overall inflammation independently by two paediatric radiologists. Total-enhancement and inflammation-synovitis scores were calculated. RESULTS: Interobserver agreement was poor to moderate for enhancement and inflammation in all recesses, except in the radioulnar and radiocarpal joints. Intraobserver agreement was good to excellent. For enhancement and inflammation scores, mean differences (95 % CI) between observers were -1.18 (-4.79 to 2.42) and -2.11 (-6.06 to 1.83). Intraobserver variability (reader 1) was 0 (-1.65 to 1.65) and 0.02 (-1.39 to 1.44). CONCLUSION: Intraobserver agreement was good. Except for the radioulnar and radiocarpal joints, interobserver agreement was not acceptable. Therefore, the proposed scoring system requires further refinement.

MRI assessment of bone marrow in children with juvenile idiopathic arthritis: intra- and inter-observer variability.

BACKGROUND: Bone marrow oedema (BMO) is included in MRI-based scoring systems of disease activity in adults with rheumatoid arthritis. Similar systems in juvenile idiopathic arthritis (JIA) are lacking. OBJECTIVE: To assess the reproducibility in a multi-centre setting of an MRI BMO scoring system in children with JIA. MATERIALS AND METHODS: Seventy-six wrist MRIs were read twice, independently, by two experienced paediatric radiologists. BMO was defined as ill-defined lesions within the trabecular bone, returning high and low signal on T2- and T1-weighted images respectively, with or without contrast enhancement. BMO extension was scored for each of 14 bones at the wrist from 0 (none) to 3 (extensive). RESULTS: The intra-observer agreement was moderate to excellent, with weighted kappa ranging from 0.85 to 1.0 and 0.49 to 1.0 (readers 1 and 2 respectively), while the inter-observer agreement ranged from 0.41 to 0.79. The intra- and inter-observer intraclass correlation coefficients were excellent and satisfactory, respectively. CONCLUSION: The scoring system was reliable and may be used for grading bone marrow abnormality in JIA. The relatively large variability in aggregate scores, particularly between readers, underscores the need for thorough standardisation.

Magnetic resonance imaging and radiographic assessment of carpal depressions in children with juvenile idiopathic arthritis: normal variants or erosions?

OBJECTIVE: Magnetic resonance imaging (MRI) is playing an increasingly important role in the diagnosis and followup of juvenile idiopathic arthritis (JIA). Carpal depressions are commonly observed in healthy children and in patients with JIA. The aim of our study was to further characterize these depressions in patients with JIA. METHODS: A total of 29 MRI wrist examinations were analyzed. Depressions were classified according to morphology as either tubular or focal. Features including the presence of a vessel related to the depression, evidence of synovitis, bone marrow edema, or loss of joint space on a radiograph taken on the same day were recorded for each depression. RESULTS: A total of 173 depressions were identified in 145 carpal bones. Forty percent were capitate depressions. A third were focal depressions and two-thirds were tubular. About 10% of tubular depressions and 30% of focal depressions were associated with features suggesting true erosions, with the remainder likely to represent vascular channels and normal variants. CONCLUSION: Radiologists and clinicians should undertake caution when assessing carpal depressions on MRI because the vast majority are likely to represent normal variants.