RESUMEN
BACKGROUND: Quantification of myocardial blood flow (MBF) is used for the noninvasive diagnosis of patients with coronary artery disease (CAD). This study compared traditional statistics, machine learning, and deep learning techniques in their ability to diagnose disease using only the rest and stress MBF values. METHODS: This study included 3245 rest and stress rubidium-82 positron emission tomography (PET) studies and matching diagnostic labels from perfusion reports. Standard logistic regression, lasso logistic regression, support vector machine, random forest, multilayer perceptron, and dense U-Net were compared for per-patient detection and per-vessel localization of scars and ischemia. RESULTS: Receiver-operator characteristic area under the curve (AUC) of machine learning models was significantly higher than those of traditional statistics models for per-patient detection of disease (0.92-0.95 vs. 0.87) but not for per-vessel localization of ischemia or scar. Random forest showed the highest AUC = 0.95 among the different models compared. On the final hold-out set for generalizability, random forest showed an AUC of 0.92 for detection and 0.89 for localization of perfusion abnormalities. CONCLUSIONS: For per-vessel localization, simple models trained on segmental data performed similarly to a convolutional neural network trained on polar-map data, highlighting the need to justify the use of complex predictive algorithms through comparison with simpler methods.
Asunto(s)
Cicatriz , Aprendizaje Profundo , Humanos , Cicatriz/diagnóstico por imagen , Tomografía Computarizada por Rayos X , Isquemia , Tomografía de Emisión de PositronesRESUMEN
AIMS: Cutaneous follicular (infundibular-tricholemmal) squamous cell carcinoma (FSCC) is a new World Health Organization entity. We present the largest series of published cases, summarizing clinical data, diagnostic criteria, differential diagnosis, and implications for patient management. METHODS: Cases were identified from 2004 to 2011. Inclusion criteria included discrete attachment(s) of the tumor to the overlying epidermis via follicular infundibula, tricholemmal keratinization, and cellular pleomorphism. Keratoacanthoma and lesions with adjacent bowenoid epidermal dysplasia were excluded. RESULTS: One hundred three cases of FSCC identified. 48.5% demonstrated completely circumscript borders ( in situ for practical purposes), 12.6% uncertain for invasion (overwhelmingly pushing borders), and only 38.8% as clearly invasive. Follicular mucin in acantholytic spaces within tumor epithelium was a distinctive finding in 57.2% of cases. Clinical data indicated predominance in elderly (median 78.5 years) men (70.4%), with preferential head and neck location (81.6%). Many were clinically suspected as squamous cell carcinoma (48.5%). However, a significant minority were clinically diagnosed as basal cell carcinoma (40.8%). This may reflect that FSCC commonly presented as a papule or nodule (51.3%). By contrast, keratoacanthoma was less frequently suggested (17.2%) and still fewer lesions were suspected to be actinic keratosis/Bowen's disease (13.6%). Follow-up in 82 cases (median 26.5 months, range 3-144) identified 5 (6.1%) local recurrences. There was no instance of metastasis in the subgroup of lesions with completely circumscript borders. Three of 45 (6.7%) patients, with follow-up, considered to have tumors with invasive pushing, and/or infiltrative borders developed lymph node metastases. CONCLUSIONS: FSCC is identified as a common skin cancer, incorporating historical entities, such as infundibular carcinoma and tricholemmal carcinoma, with readily identifiable histologic features. Correct diagnosis has implications for patient management; a significant subgroup of lesions show completely circumscript borders that are considered in situ for practical purposes.